feat(tls): support cross-host mTLS with lacme 1.2

fix: revalidate frontend assets across builds
feat: convert large pastes to attachments
2026-08-13 15:32:24 -06:00 · 2026-08-13 02:19:45 -07:00 · 2026-08-12 19:00:31 -07:00 · 2026-08-12 00:27:38 -07:00 · 2026-08-11 23:30:44 -07:00 · 2026-08-11 22:03:10 -07:00
527 changed files with 148940 additions and 16705 deletions
@@ -42,6 +42,9 @@
 # CONSOLE_HTTPS_PORT=8443          # Caddy (dashboard HTTPS)
 # POSTGRES_PORT=5432               # exposed for bare-metal host joins
 # POSTGRES_BIND=127.0.0.1          # set 0.0.0.0 to let another machine join
+# TURNSTONE_HOST_IP=127.0.0.1      # dev-stack bind address for cross-host joins
+# TURNSTONE_CONSOLE_HTTP_BIND=127.0.0.1  # production TLS-overlay ACME/API bind
+# TURNSTONE_ACME_EXTERNAL_URL=http://192.0.2.1:8090/acme  # routable ACME base; bracket IPv6; include /acme

 # -- Workspace ----------------------------------------------------------------
 # Bind-mount a host directory the model can read/write at /workspace:
@@ -55,7 +55,7 @@
    {
      "description": "LLM SDKs — always review manually",
      "groupName": "LLM SDKs",
-      "matchPackageNames": ["openai", "anthropic", "mcp"],
+      "matchPackageNames": ["openai", "httpx2", "anthropic", "mcp"],
      "schedule": ["before 9am on Monday"],
      "automerge": false
    },
@@ -14,8 +14,8 @@ jobs:
  lint:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-      - uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1 # v6
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7
        with:
          python-version: "3.14"
      - run: pip install pre-commit
@@ -25,8 +25,8 @@ jobs:
  typecheck:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-      - uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1 # v6
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7
        with:
          python-version: "3.14"
      - run: pip install mypy
@@ -35,29 +35,31 @@ jobs:

  test:
    runs-on: ubuntu-latest
-    # Cap a hung run at 20 min instead of riding GitHub's 6-hour default
-    # (a flaky-hang run otherwise streams -v output for hours).
-    timeout-minutes: 20
+    # Cap a hung run at 30 min instead of riding GitHub's 6-hour default
+    # (a flaky-hang run otherwise streams -v output for hours).  Was 20;
+    # the suite's growth (~9.7k tests, coverage-instrumented, 3-version
+    # matrix) started brushing the old cap on healthy runs.
+    timeout-minutes: 30
    strategy:
      matrix:
        python-version: ["3.11", "3.12", "3.13"]
    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-      - uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1 # v6
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7
        with:
          python-version: ${{ matrix.python-version }}
      # Node is required by tests/test_renderer_js.py — without
      # explicit setup, that suite silently skips if the runner
      # image happens not to ship Node, masking regressions in
      # the browser-side renderer.
-      - uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6
+      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
        with:
          node-version: "24"
      - run: pip install -e ".[test]"
      # -v lists each test id as it starts (pytest prints the nodeid at
      # logstart), so a hang names the culprit on the last line instead of
      # riding the job timeout with only a trail of "..." dots.
-      - run: pytest tests/ -m "not live" --cov=turnstone --cov-report=term-missing --cov-report=xml -v
+      - run: pytest tests/ -m "not live and not e2e_recovery" --cov=turnstone --cov-report=term-missing --cov-report=xml -v
      - uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
        if: always()
        with:
@@ -66,7 +68,7 @@ jobs:

  test-postgres:
    runs-on: ubuntu-latest
-    timeout-minutes: 20
+    timeout-minutes: 30
    services:
      postgres:
        image: postgres:18
@@ -82,23 +84,23 @@ jobs:
          --health-timeout=5s
          --health-retries=5
    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-      - uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1 # v6
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7
        with:
          python-version: "3.14"
-      - uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6
+      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
        with:
          node-version: "24"
      - run: pip install -e ".[test]"
-      - run: pytest tests/ -m "not live" --storage-backend=postgresql -v
+      - run: pytest tests/ -m "not live and not e2e_recovery" --storage-backend=postgresql -v
        env:
          TURNSTONE_TEST_PG_URL: postgresql+psycopg://postgres:postgres@localhost:5432/turnstone_test

  wheel-completeness:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-      - uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1 # v6
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7
        with:
          python-version: "3.14"
      - run: pip install build
@@ -151,8 +153,8 @@ jobs:
  lock-check:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-      - uses: astral-sh/setup-uv@11f9893b081a58869d3b5fccaea48c9e9e46f990 # v8.3.2
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: astral-sh/setup-uv@c771a70e6277c0a99b617c7a806ffedaca235ff9 # v9.0.0
        with:
          uv-version: "0.9.18"
      - run: uv lock --check
@@ -160,11 +162,11 @@ jobs:
  security:
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-      - uses: astral-sh/setup-uv@11f9893b081a58869d3b5fccaea48c9e9e46f990 # v8.3.2
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: astral-sh/setup-uv@c771a70e6277c0a99b617c7a806ffedaca235ff9 # v9.0.0
        with:
          uv-version: "0.9.18"
-      - uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1 # v6
+      - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7
        with:
          python-version: "3.14"
      - run: uv sync --frozen --all-extras
@@ -188,8 +190,8 @@ jobs:
      run:
        working-directory: sdk/typescript
    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-      - uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
        with:
          node-version: "24"
      - run: npm ci
@@ -1,46 +0,0 @@
-name: Claude Code Review
-
-on:
-  pull_request:
-    types: [opened, synchronize, ready_for_review, reopened]
-    # Optional: Only run on specific file changes
-    # paths:
-    #   - "src/**/*.ts"
-    #   - "src/**/*.tsx"
-    #   - "src/**/*.js"
-    #   - "src/**/*.jsx"
-
-jobs:
-  claude-review:
-    if: github.event.pull_request.head.repo.full_name == github.repository
-    # Optional: Filter by PR author
-    # if: |
-    #   github.event.pull_request.user.login == 'external-contributor' ||
-    #   github.event.pull_request.user.login == 'new-developer' ||
-    #   github.event.pull_request.author_association == 'FIRST_TIME_CONTRIBUTOR'
-
-    runs-on: ubuntu-latest
-    permissions:
-      contents: read
-      pull-requests: write # post the review + inline comments
-      issues: read
-      id-token: write
-
-    steps:
-      - name: Checkout repository
-        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-        with:
-          fetch-depth: 1
-
-      - name: Run Claude Code Review
-        id: claude-review
-        uses: anthropics/claude-code-action@e90deca47693f9457b72f2b53c17d7c445a87342 # v1
-        with:
-          claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
-          allowed_bots: 'renovate[bot]' # let Renovate PRs get reviewed
-          plugin_marketplaces: 'https://github.com/anthropics/claude-code.git'
-          plugins: 'code-review@claude-code-plugins'
-          prompt: '/code-review:code-review ${{ github.repository }}/pull/${{ github.event.pull_request.number }}'
-          # See https://github.com/anthropics/claude-code-action/blob/main/docs/usage.md
-          # or https://code.claude.com/docs/en/cli-reference for available options
-
@@ -1,63 +0,0 @@
-name: Claude Code
-
-on:
-  issue_comment:
-    types: [created]
-  pull_request_review_comment:
-    types: [created]
-  issues:
-    types: [opened, assigned]
-  pull_request_review:
-    types: [submitted]
-
-jobs:
-  claude:
-    if: |
-      (
-        github.event_name == 'issue_comment' &&
-        contains(github.event.comment.body, '@claude') &&
-        contains(fromJSON('["OWNER","MEMBER","COLLABORATOR"]'), github.event.comment.author_association)
-      ) || (
-        github.event_name == 'pull_request_review_comment' &&
-        contains(github.event.comment.body, '@claude') &&
-        contains(fromJSON('["OWNER","MEMBER","COLLABORATOR"]'), github.event.comment.author_association)
-      ) || (
-        github.event_name == 'pull_request_review' &&
-        contains(github.event.review.body, '@claude') &&
-        contains(fromJSON('["OWNER","MEMBER","COLLABORATOR"]'), github.event.review.author_association)
-      ) || (
-        github.event_name == 'issues' &&
-        (contains(github.event.issue.body, '@claude') || contains(github.event.issue.title, '@claude')) &&
-        contains(fromJSON('["OWNER","MEMBER","COLLABORATOR"]'), github.event.issue.author_association)
-      )
-    runs-on: ubuntu-latest
-    permissions:
-      contents: read
-      pull-requests: write # post comments/reviews when @-mentioned on a PR
-      issues: write # post comments when @-mentioned on an issue
-      id-token: write
-      actions: read # Required for Claude to read CI results on PRs
-    steps:
-      - name: Checkout repository
-        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-        with:
-          fetch-depth: 1
-
-      - name: Run Claude Code
-        id: claude
-        uses: anthropics/claude-code-action@e90deca47693f9457b72f2b53c17d7c445a87342 # v1
-        with:
-          claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
-
-          # This is an optional setting that allows Claude to read CI results on PRs
-          additional_permissions: |
-            actions: read
-
-          # Optional: Give a custom prompt to Claude. If this is not specified, Claude will perform the instructions specified in the comment that tagged it.
-          # prompt: 'Update the pull request description to include a summary of changes.'
-
-          # Optional: Add claude_args to customize behavior and configuration
-          # See https://github.com/anthropics/claude-code-action/blob/main/docs/usage.md
-          # or https://code.claude.com/docs/en/cli-reference for available options
-          # claude_args: '--allowed-tools Bash(gh pr *)'
-
@@ -33,7 +33,7 @@ jobs:
      startsWith(github.event.workflow_run.head_branch, 'v')
    runs-on: ubuntu-latest
    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
        with:
          ref: ${{ github.event.workflow_run.head_sha }}
          fetch-depth: 0
@@ -54,7 +54,7 @@ jobs:

      - name: Log in to GHCR
        if: steps.tag.outputs.skip == 'false'
-        uses: docker/login-action@af1e73f918a031802d376d3c8bbc3fe56130a9b0 # v4
+        uses: docker/login-action@dbcb813823bdd20940b903addbd779551569679f # v4
        with:
          registry: ${{ env.REGISTRY }}
          username: ${{ github.actor }}
@@ -30,7 +30,7 @@ jobs:
    runs-on: ubuntu-latest
    environment: pypi
    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
        with:
          ref: ${{ github.event.workflow_run.head_sha }}
          fetch-depth: 0
@@ -50,7 +50,7 @@ jobs:
            echo "skip=false" >> "$GITHUB_OUTPUT"
          fi

-      - uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1 # v6
+      - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7
        if: steps.tag.outputs.skip == 'false'
        with:
          python-version: "3.14"
@@ -58,12 +58,12 @@ jobs:
        if: steps.tag.outputs.skip == 'false'
      - run: python -m build
        if: steps.tag.outputs.skip == 'false'
-      - uses: pypa/gh-action-pypi-publish@cef221092ed1bacb1cc03d23a2d87d1d172e277b # release/v1
+      - uses: pypa/gh-action-pypi-publish@dc37677b2e1c63e2034f94d8a5b11f265b73ba33 # release/v1
        if: steps.tag.outputs.skip == 'false'

      - name: Create GitHub Release
        if: steps.tag.outputs.skip == 'false'
-        uses: softprops/action-gh-release@718ea10b132b3b2eba29c1007bb80653f286566b # v3
+        uses: softprops/action-gh-release@3d0d9888cb7fd7b750713d6e236d1fcb99157228 # v3
        with:
          tag_name: ${{ steps.tag.outputs.tag }}
          generate_release_notes: true
@@ -31,8 +31,8 @@ jobs:
        # Floor and ceiling of the example's requires-python (>=3.11).
        python-version: ["3.11", "3.13"]
    steps:
-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
-      - uses: actions/setup-python@ece7cb06caefa5fff74198d8649806c4678c61a1 # v6
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+      - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7
        with:
          python-version: ${{ matrix.python-version }}
      - run: pip install -e ".[test,dev]"
@@ -61,7 +61,7 @@ jobs:
          fi
          echo "head_ref=${ref}" >> "$GITHUB_OUTPUT"

-      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
        with:
          ref: ${{ steps.ref.outputs.head_ref }}

@@ -19,6 +19,8 @@ docker-compose.override.yml
 .ruff_cache/
 .pytest_cache/
 *.db
+*.db-shm
+*.db-wal
 .plan.md
 .plan-*.md
 .hypothesis/
@@ -29,3 +31,4 @@ tools/skill_audit_analysis/output/
 design_ideas/
 .claude/
 docs/design/
+/.idea
@@ -18,6 +18,51 @@ Earlier stable lines (`stable/1.6`, `stable/1.5`) are frozen.

 ### Added

+- **Large plain-text pastes become attachments.** Pasting text longer than the
+  fixed 2,000-character threshold stages `pasted-text.txt` across all five
+  attachment-capable create/send composers. Clipboard files retain priority,
+  text above the 512 KiB upload ceiling stays inline, identical synthesized
+  pastes collapse to one chip, and rejected attachment sends keep the staged
+  message and files so they can be corrected or retried.
+- **`server_parses_reasoning` model capability.** Declare it on a model
+  definition whose backend segregates reasoning into its own channel (a
+  vLLM launched with a reasoning parser, a commercial provider): the
+  inline think-tag scan turns off on every lane — interactive and
+  drained alike — so content is trusted verbatim and prose that merely
+  quotes a tag can no longer be misrouted into the reasoning lane, and
+  the utility lanes stop suppressing reasoning they'd otherwise pin off.
+  Default off for local lanes, preserving the passthrough-server
+  behavior; the built-in capability tables declare it for every real
+  commercial endpoint (known models and table-miss defaults alike),
+  which also removes the quoted-tag false positive from those lanes.
+- **Per-model Entra gateway authentication.** Model definitions can bind either
+  a caller-delegated OBO token (`entra_obo`) or a shared app-identity token
+  (`entra_app`) through the provider SDK credential surface. Mints reuse the
+  encrypted cluster token cache, refresh-rotation CAS, and advisory locking;
+  add a host-local memo, failure cooldown, long-lived mint HTTP client, audience
+  allow-list/permission boundary, identity-unlink purge, and optional
+  `model.auth_fail_closed` refusal policy. Delegated identity now propagates
+  through judge, output-guard, and principal-scoped perception lanes, and
+  unattended watch restoration reacquires the persisted workstream owner.
+  Ownerless OBO calls and dynamic aliases without a real static fallback always
+  fail closed; grant modes are never silently switched. Static authentication
+  remains the default.
+
+- **Compaction is visible now: lifecycle events, a progress bar, and a
+  persistent transcript card.** Context compaction (manual `/compact` and
+  auto) emits a first-class `compaction` SSE event
+  (`start` / `progress` / `end` — see the API reference) instead of loose
+  info lines. The web UI renders an in-transcript card with a real progress
+  bar (determinate `part k of N` during chunked summarization, indeterminate
+  for single-call compactions) that settles into a result card — token delta
+  plus the summary behind a fold — in both the interactive pane and the
+  coordinator viewer. The result survives reloads: the persisted compaction
+  marker now projects through `/history` as a `role="system"`,
+  `source="compaction"` entry (resume/export/search unchanged), stamped with
+  the end event's id so repaint and SSE replay can't double-render. The
+  marker's `meta` additionally records `before_tokens` / `after_tokens` /
+  `trigger`. Python and TypeScript SDKs gain a typed `CompactionEvent`.
+
 - **One provider transport: every model call now streams (#831).**
  The per-adapter non-streaming entry (`create_completion`) is retired;
  single-shot lanes — judges, titles, compaction, web-fetch extraction,
@@ -113,6 +158,53 @@ Earlier stable lines (`stable/1.6`, `stable/1.5`) are frozen.

 ### Changed

+- **lacme 1.2.0 and HTTPX2 now back the core mTLS path.** TLS/ACME is a typed
+  core dependency rather than optional-import-era code; renewal and admin
+  clients use lacme's public close/stop lifecycle. Consoles can set
+  `TURNSTONE_ACME_EXTERNAL_URL` to a routable responder base ending in `/acme`,
+  so nodes on another host receive usable directory and follow-up URLs during
+  enrollment. Signing routes now require a dedicated, purpose-confined rotating
+  service JWT, and HTTPX2 pins that credential to canonical resources on
+  configured responder origins. Internal and external console identities use
+  separate persistence namespaces; cluster identities are reused only after
+  key, SAN, validity, EKU, and active-root verification. Responder-side keyless
+  CSR results cannot overwrite managed identities, failed live reloads restore
+  the last usable bundle, certificate lifetime stays at 48 hours with a
+  12-hour renewal cadence, and cancellation drains renewal clients before
+  propagating. Operator-supplied IPv4 and IPv6 literals are passed to lacme as
+  typed IP identifiers and issued as IP SANs; unexpired legacy certificates
+  containing `DNS:<ip>` are reissued instead of being reused as an invalid IP
+  identity
+  ([#1011](https://github.com/turnstonelabs/turnstone/issues/1011)).
+
+- **OpenAI SDK v3 and its HTTPX2 default transport are now supported (#1009).**
+  Chat Completions and Responses streams normalize native HTTPX2 connection
+  deaths through the same retry boundary as legacy HTTPX-backed providers,
+  including failures observed after safe cross-thread client closure during a
+  model-registry reload. The OpenAI v3 runtime escape hatch for explicitly
+  injected legacy HTTPX clients remains supported. OpenAI connections now
+  follow HTTPX2's operating-system trust store by default; deployments that
+  relied on a modified `certifi` bundle must install that CA in the system
+  store or set `SSL_CERT_FILE` / `SSL_CERT_DIR`.
+
+- **Log event rename: `drain_stream.post_finish_blip` is now
+  `stream.post_finish_blip`; its `usage_captured` field is retained.** The
+  single-shot drain normalizes mid-body transport deaths through the same
+  `transport_guarded` wrapper the interactive loop uses, so its
+  post-finish-blip tolerance logs under the wrapper's event name. Update
+  any external log filters pinned to the old name; the drained result's
+  possible `usage=None` on a post-finish blip is unchanged and documented
+  on `drain_stream`.
+
+- **Breaking (1.8): compaction feedback moved from `info` events to the
+  typed `compaction` SSE event.** Pre-1.8 SSE/SDK clients that ignore
+  unknown event types no longer see compaction lines (they are
+  deliberately not dual-emitted — dual emission would double-render on
+  every current client). Consume the `compaction` lifecycle event (see
+  the API reference and the `CompactionEvent` SDK type); embedders
+  driving `ChatSession` through a duck-typed `SessionUI` are unaffected
+  (the classic `on_info` lines are restored for them — see Fixed).
+
 - **Sampling knobs (temperature, reasoning effort) now ride one assignment
  scheme: per-model alias value → operator-stored global setting → the
  model definition's declared default (effort only) → field omitted.**
@@ -160,6 +252,306 @@ Earlier stable lines (`stable/1.6`, `stable/1.5`) are frozen.

 ### Fixed

+- **MCP servers may expose resources without resource templates, or templates
+  without concrete resources (#993).** The MCP handshake has one aggregate
+  `resources` capability, but implementations are not required to support both
+  list methods. An exact JSON-RPC `Method not found` response from either
+  method is now treated as an empty half-catalog during static and per-user
+  discovery and refresh, while every other error still fails closed. A failed
+  static registration also tears down its transport and publishes none of its
+  staged tools/resources/prompts, eliminating the live callable “ghost” tools
+  that could remain after the UI reported registration failure.
+
+- **A cancelled judge, guard, or compaction call can now stop before its
+  request goes out (#972).** Previously it could not: `model_turn` refused
+  to *re-issue* an abandoned call after a mid-stream death, but nothing
+  checked before a first dispatch, so a call whose caller had already gone
+  away still sent — and the reply was discarded unread after the endpoint
+  had accepted the work. It now checks immediately before sending, so a
+  Stop observed by that point costs no request, and again on entry, so a
+  call already cancelled when it arrives also skips credential resolution.
+  Cancellation is cooperative, which bounds what that buys: a Stop only
+  saves the request if it lands before dispatch — sending is a moment, the
+  response streaming back is the rest of the call, and an abort arriving
+  then still meets a request in flight, closed in place exactly as before.
+  The window that did widen usefully is a delegated-auth alias whose token
+  mint blocks; a Stop during that mint now costs no request (though a mint
+  already under way still completes). What a stopped call saves is the
+  request, its prompt-side billing, and — on a capacity-bounded
+  self-hosted endpoint — a slot a live request wanted. Unchanged: the
+  interactive turn, which has its own pre-send cancellation check on a
+  different path, and the lanes that thread no cancellation handle
+  (attachment perception, title generation, web-fetch extraction,
+  sub-agents, optimizer, eval) — and web-fetch extraction deliberately
+  never will, since it runs on parallel tool threads where registering one
+  would clobber the main stream's.
+- **Unmarked chain-of-thought no longer leaks into titles, summaries, or
+  web-fetch tool results (#940).** Some serving setups emit reasoning
+  inline with no tags and no `reasoning_content` at all — nothing any
+  parser can segregate. The bounded-artifact lanes (title, compaction,
+  web-fetch extraction) now ask the model for no reasoning instead:
+  the model definition's declared thinking toggle is pinned off for that
+  call — the same suppression transcription already used — and the
+  reasoning-effort channels (the relayed session knob, the definition's
+  default, the graded template key) are withheld with it, since an
+  effort value beside a pinned-off toggle re-requests the reasoning the
+  pin declined. A no-op on backends that segregate reasoning
+  server-side. Title generation additionally stopped trusting line
+  position: it takes the last line that reads as a title (within the
+  word cap and ending in a word character, so explanation sentences,
+  sign-offs, and reasoning headings lose in any script) rather than the
+  first non-empty line, which unmarked reasoning turned into titles
+  like "Thinking Process:".
+- **A think tag split across a reasoning delta now reassembles.** The
+  non-streaming drain closes content runs at interleaving signals; a
+  partial-tag tail is carried across reasoning-delta boundaries (a
+  reasoning delta cannot terminate a tag) so the tag is consumed instead
+  of its halves passing through as visible content. Tool-call boundaries
+  still flush — no tag spans a tool call.
+- **Streaming consumers follow the ACTIVE model's capabilities.** The
+  interactive tag-scan posture and the drain's scan gate now read the
+  capabilities of the lane that owns the stream being consumed (fallback
+  walks included) instead of the session's primary alias.
+- **Notification bodies no longer fuse multi-block answers.** `Turn.text`
+  joins text blocks with a newline; a final assistant turn stored as
+  multiple text blocks previously concatenated the last word of one
+  block to the first word of the next in completion notifications and
+  every other flattened read.
+- **String-typed boolean capability overrides coerce instead of
+  truthiness-flipping.** A hand-edited `"false"`/`"0"` in a model
+  definition's capabilities JSON now means false; unrecognized values
+  drop the key and keep the field's default.
+- **Inline `<think>`/`<reasoning>` blocks no longer leak into drained
+  results (#965, #940).** On servers without a reasoning parser
+  (parserless vLLM/llama.cpp, LM Studio, bare gateways), reasoning
+  arrives as literal tags inside content; segregation now happens once
+  at the drain seam, so web-fetch tool results, sub-agent syntheses,
+  judge verdicts, titles, summaries, and optimizer prompts receive
+  tag-free content and the extracted reasoning rides the native lane.
+  Two behavior notes: a web-fetch extraction whose whole response was
+  reasoning now returns an explicit `Error: extraction returned no
+  answer` tool result (previously the raw reasoning text persisted as a
+  successful result and was replayed every following turn), and a
+  mismatched-vocabulary close tag (`<think>…</reasoning>`) now closes
+  the block — matching the interactive lane's long-standing rule —
+  where the old per-lane strips treated it as unterminated.
+
+- **A transport failure mid-generation no longer kills the interactive
+  turn (#937).** A wire death during body streaming (TLS record failure,
+  connection reset — `httpx.ReadError` and kin) surfaces after the
+  request has already returned its stream handle, so neither the SDK's
+  request retries nor the creation-time retry ladder ever saw it: the
+  turn died with a bare `ReadError: …`, the partial output was
+  discarded, and nothing was logged. The interactive loop now normalizes
+  mid-body transport deaths exactly like the single-shot lanes and
+  re-issues the turn (bounded, cancel-aware, exponential backoff),
+  finalizing the dead attempt across every UI surface first so retried
+  text never double-renders (web transcript, CLI markdown fences,
+  Slack/Discord streamed messages). Before re-creating the stream the
+  session re-resolves its registry binding, so a concurrent model-registry
+  reload that closed the old client cannot turn the retry into a
+  misleading closed-client error. On exhaustion the surfaced error names
+  the provider, endpoint, and model with a stream-death message instead
+  of a bare exception string, and every fatal turn now leaves a
+  `session.fatal.recorded` log line (INFO for a user Ctrl-C, ERROR
+  otherwise).
+
+- **A failed worker-thread spawn no longer wedges the workstream — at
+  either spawn site — and never masquerades as success.** If
+  `Thread.start()` itself raised (thread exhaustion, out-of-memory), the
+  dispatcher had already claimed the worker slot but the flag's only
+  clearer lived in the never-started thread — the workstream looked idle
+  forever while every subsequent message queued behind a worker that
+  didn't exist, until an operator force-cancel. The claim is now rolled
+  back under the lock and the error propagates, so the workstream is
+  dispatchable again as soon as resources recover. Affected every
+  dispatch path (sends, wakes, retries, deferred-send drain, init). The
+  same failure at the deferred-send drain's own spawn rolls back the
+  just-accepted entry and answers the retryable `queue_full` (previously
+  a 500 landed *after* the entry was registered — an invisible,
+  unretractable phantom that later dispatched as duplicate turns), and a
+  `/command` whose worker never spawned now answers **503**
+  `{"status": "error"}` instead of the generic 200 ok that told SDK
+  callers their `/clear` or `/resume` had applied.
+
+- **Manual `/compact` from the web UI: no phantom user turn, no frozen
+  server, cancellable.** A slash command typed into the web composer no
+  longer renders as a user chat bubble (it echoes as a distinct command
+  chip — commands aren't conversation turns and were never persisted as
+  such). `/compact` itself now dispatches onto the workstream's worker
+  slot instead of running inline on the server's event loop — previously a
+  long compaction froze every SSE stream on the node for its whole
+  duration, which is also why its own progress only ever arrived as one
+  burst after the fact. The manual path carries `send()`'s full generation
+  discipline (`compact_now()`): a force-abandoned compaction goes stale
+  instead of swapping history under a successor turn — and retires at its
+  next checkpoint instead of running out its remaining summary calls,
+  with its late lifecycle events fenced off (`compaction_id` on every
+  event, `superseded` on end events — both in the SDKs) so they can't
+  animate, tear down, re-title, or falsely narrate a successor's card or
+  activity pill; a cancel aimed at it is consumed on exit (previously it
+  bricked every `/compact` retry until the next message); a Stop click on
+  an idle session can't pre-abort the next compaction; a Stop that lands
+  in the completion tail — after the last cancel check, or during a retry
+  backoff (which now aborts immediately instead of sleeping it out) — is
+  honored rather than silently eaten; and Stop now aborts the in-flight
+  summary HTTP call itself (the compaction lane registers its stream in
+  the same abort seam the main loop uses), so cancelling a compaction is
+  immediate instead of waiting out a model call.
+
+- **Sends during a command window are deferred, ordered, bounded, and
+  honestly rendered — never silently truncated or lost.** Messages sent
+  while any slash command holds the worker slot are **deferred**: answered
+  `{"status": "queued", "msg_id"}` immediately and dispatched as ordinary
+  full-fidelity sends (attachments and sender identity included) when the
+  command finishes — never routed through the mid-turn interjection
+  queue, whose semantics are turn-shaped: previously a send during a
+  manual `/compact` was silently truncated to 2,000 characters, a second
+  participant in a shared workstream was locked out with a misleading
+  "another participant's turn" 409 for the whole compaction, and a
+  message queued across a `/resume`/`/new` could be answered into the
+  post-swap workstream. Because the response is immediate,
+  timeout-bounded callers — the coordinator's `send_message`, the console
+  proxy, SDKs, anything behind a stock reverse proxy — can no longer lose
+  a message to a multi-minute command window; the deferred send is
+  retractable until dispatch via the same `DELETE .../send` used for
+  queued interjections (node-local, in-memory — the API reference
+  documents the at-most-once durability contract). Deferred responses
+  carry `"deferred": true`; the pending list is the **order authority**
+  (a fresh send — or a coordinator dispatch, or a queued-nudge wake —
+  lines up behind acknowledged entries instead of overtaking them, with
+  the two-term barrier defined once on the workstream so the wake gate
+  also honors a claimed entry whose dispatch is mid-flight, and the gate
+  re-arms at the drain's exit even when everything pending was
+  retracted); acceptance is **bounded** (10 pending per workstream — the
+  interjection queue's own backpressure contract; the 11th answers the
+  retryable `queue_full` instead of pinning attachment bytes without
+  limit and then running one unattended turn per entry); a dispatch
+  crash re-queues the entry instead of eating an acknowledged message,
+  and a drain thread that fails to *start* rolls the acceptance back and
+  answers `queue_full` rather than parking a phantom the client can
+  neither see nor retract; each dispatch emits a pane-tier
+  `message_dispatched` event (`folded: true` for interjection fold-ins)
+  so queued-bubble UI keeps its retract affordance exactly until the
+  message truly leaves — including when the send was accepted by a pane
+  that believed the workstream idle, which now renders a real queued
+  chip instead of a sent-looking bubble, releases the composer (a
+  deferred send has no running worker to wait on), and cleans up fully
+  when the send is refused or the chip retracted instead of stranding
+  the pane in Stop mode. Dismissing a queued bubble — interjection or
+  deferred — is a server-confirmed `DELETE`, and retracting a deferred
+  send that carried attachments tells the user they were discarded
+  instead of silently expiring them.
+
+- **Slash commands hold the worker slot with a loud contract.**
+  A `/compact` raced against an in-flight turn is refused with an
+  explicit busy response. Every other slash command runs through the same
+  worker slot too — mutual exclusion against sends, a running compaction,
+  and each other, with a busy answer replacing the old silent interleave —
+  while the endpoint still awaits quick commands' completion off-loop
+  (without parking an executor thread per request); the post-command pane
+  refreshes (`clear_ui` after `/clear`/`/new`/`/resume`, the
+  workstream-name sync) ride the worker itself, so a command that
+  outlives the endpoint's 25s response backstop still refreshes every
+  pane on completion (the backstop sits under the console proxy's 30s
+  client timeout so the degraded `running` answer can actually traverse
+  a proxied pane, which now surfaces it instead of silence; the
+  `/command` response contract — `ok` / `running`, with busy refusals
+  answering a loud HTTP 409 rather than a silent 200 — is now documented
+  in the API reference and the OpenAPI spec).
+
+- **Compaction status stays truthful across every UI surface.** Manual
+  compaction
+  success also refreshes the status line/context pill immediately (parity
+  with auto-compaction), compaction failures keep feeding the typed
+  `error` event and the node error counter (while a CLI Ctrl-C reports as
+  cancelled, not a failure), one Stop prints one notice (a cancelled
+  auto-compaction no longer stacks "Compaction cancelled." on top of
+  send's own "[Generation cancelled]"), the workstream activity pill
+  shows "Compacting context…" for the whole summarize phase, restores
+  cleanly afterwards, and can no longer be stranded by a force-stopped
+  compaction (a new turn's generation claim breaks a stale latch). Every
+  retry backoff on the session (stream retries, task agents, notify
+  delivery, compaction) now aborts immediately on Stop via one shared
+  cancel-aware helper instead of sleeping out its exponential delay.
+
+- **Compaction failures report exactly once, to the right owner.** A
+  compaction failure reports
+  exactly once (auto-compaction errors defer to the turn's fatal handler
+  instead of doubling the red row and the error metric), failed-end
+  notice suppression is computed once by the emitter (a `notice` bool on
+  the end event — in the SDKs — replaces hand-synced client policy), and
+  a manual `/compact` failure no longer crashes the CLI REPL. `/compact`
+  on a workstream showing the `error` badge restores the badge on exit
+  instead of stamping `idle` over it (the compaction neither retried nor
+  resolved the failed turn). A force-cancelled initial send that
+  completes late still delivers its scheduled-run completion
+  notification (the only completion signal unattended workstreams have);
+  the other post-command pane refreshes and error notices remain
+  owner-guarded, so a force-cancelled wedged command that unwedges late
+  can't wipe panes or inject stray notices into a successor turn.
+
+- **Pre-1.8 embedder UIs keep their compaction lines.** Embedders
+  driving `ChatSession` with a pre-1.8 duck-typed `SessionUI`
+  (no `on_compaction` hook) get the classic `on_info` compaction lines
+  back — threshold notice, `part k/N`, retry waits, token delta +
+  summary box — instead of silent history swaps. (See the breaking
+  event-contract note under **Changed** for SSE/SDK clients.)
+
+- **Static MCP servers: a pushed catalog change no longer wedges the shared
+  session (#839).** The static-path `*/list_changed` handler awaited its
+  catalog refresh inline in the SDK's receive loop, but the refresh's own
+  request can only be answered by that (now parked) loop — the refresh never
+  completed, and every user's in-flight calls on the shared per-node session
+  stalled behind it, unbounded, until the health loop's ping timeout tore the
+  transport down (which was also the only way the changed catalog ever
+  landed). Push refreshes now run as spawned tasks — debounced, coalesced per
+  (server, kind), bounded by the connect timeout, and serialized on the
+  per-server connect lock — and the manual and post-reconnect refreshes
+  publish under that same lock, so a slower publisher can no longer land a
+  staler catalog over a fresher one. Every teardown path now also clears the
+  notification debounce stamp, so a reconnected server's first push refreshes
+  immediately. Push-refresh debouncing is now per (server, kind) on BOTH the
+  static and per-user pool paths — a tools push no longer swallows a prompts
+  push arriving in the same 5-second window. A change genuinely lost to the
+  debounce window (a same-kind push landing after the prior refresh finished,
+  which the server will never re-announce) is recovered by an automatic
+  health-tick retry rather than staying invisible until an unrelated push or
+  a reconnect. The resource-refresh fan-out on both paths no longer orphans
+  its sibling list call when one of the pair fails fast — the real error
+  surfaces immediately (not masked as a 30-second timeout) and the surviving
+  sibling is cancelled and reaped, under a bounded grace, inside the scope. A
+  push refresh that fails while the connection stays up is likewise retried on
+  the next health-loop tick until one completes — previously a single
+  transient blip left the shared catalog stale for every user on the node
+  until an operator intervened. An operator `/mcp refresh` no longer parks
+  behind a busy per-server connect lock (a slow reconnect attempt could eat
+  the whole 30-second refresh budget and fail the pass for every healthy
+  server behind it) — the busy server is skipped on both the connected and
+  disconnected branches, reported distinctly as "skipped" rather than as a
+  false "no changes", the skip arms the automatic retry, and a
+  force-reconnect drops the session up front so queued push refreshes can't
+  starve it. Static-path resource and prompt catalogs are now size-capped
+  like the pool path's (and like static tools) at discovery and on every
+  refresh, so a misbehaving server's push can't balloon the node's merged
+  catalogs. Deleting or reconfiguring a server can no longer leave it
+  half-removed: the config removal and all cleanup are serialized under the
+  connect lock (a cancelled removal completes its cleanup rather than
+  stranding a live session and published catalog with the config already
+  gone), and `reconcile_sync` retries a removal that timed out instead of
+  marking it done — previously a DB-driven delete of a busy server could be a
+  silent, permanent no-op until process restart. A refresh outcome now
+  threads consistently to every operator surface off one source of truth
+  (the per-server `last_refresh_outcome`): a busy-skip and a genuine failure
+  are each reported distinctly from a real "no changes" — `/mcp refresh`
+  prints "skipped" or "failed" rather than a false "no changes", and the
+  node-internal refresh endpoint returns `202 skipped` instead of a
+  misleading `200 ok` for a refresh that never ran. A single-kind push
+  refresh no longer paints the whole server healthy: because the
+  error/outcome state is server-scoped, a successful tools push while the
+  prompts catalog is still broken (or vice versa) no longer clears the
+  failure — only a full refresh pass declares "ok".
+
 - **OpenAI Responses streaming: truncated and refused responses no longer
  vanish.** A response that hit `max_output_tokens` terminates the stream
  with `response.incomplete`, which the stream consumer did not handle —
@@ -8,6 +8,11 @@ The following people have contributed code to the project — thank you:
 - Burhan ([@Burhan-Q](https://github.com/Burhan-Q))
 - chrismuzyn ([@chrismuzyn](https://github.com/chrismuzyn))
 - daoxley ([@daoxley](https://github.com/daoxley))
+- metaclassing ([@metaclassing](https://github.com/metaclassing))
+- posixpositive ([@bensonjohnson](https://github.com/bensonjohnson))
 - Robert DeAngelis ([@OriginalOrangeXD](https://github.com/OriginalOrangeXD))
+- Sanjay Santhanam ([@Sanjays2402](https://github.com/Sanjays2402))
+- Stefano Maffeis ([@lesbass](https://github.com/lesbass))
 - William ([@sillyWillieBilly](https://github.com/sillyWillieBilly))
+- [@BlackMyrmidon](https://github.com/BlackMyrmidon)
 - [@pizzaandcheese](https://github.com/pizzaandcheese)
@@ -8,7 +8,7 @@ FROM python:3.14-slim
 LABEL org.opencontainers.image.title="turnstone" \
      org.opencontainers.image.description="Multi-node AI orchestration platform"

-COPY --from=ghcr.io/astral-sh/uv:0.11.27 /uv /usr/local/bin/uv
+COPY --from=ghcr.io/astral-sh/uv:0.12.3 /uv /usr/local/bin/uv

 # Remove the slim image's man page exclusion so man-db has actual content
 RUN rm -f /etc/dpkg/dpkg.cfg.d/docker
@@ -55,13 +55,17 @@ COPY docker/healthcheck.py /usr/local/bin/healthcheck.py

 # Entrypoint script — runs migrations before starting
 COPY docker/entrypoint.sh /usr/local/bin/entrypoint.sh
-
+RUN chmod +x /usr/local/bin/entrypoint.sh
 # Data directory — SQLite DB is created in CWD
 WORKDIR /data
 RUN chown turnstone:turnstone /data

-# Workspace mount point — bind-mount a host directory here
+# Workspace mount point — bind-mount a host directory here.  The env var
+# surfaces the path in the model's shell/file tool descriptions
+# (config.get_workspace_dir); without it the mount is invisible to the
+# model, whose cwd is /data below.
 RUN mkdir -p /workspace && chown turnstone:turnstone /workspace
+ENV TURNSTONE_WORKSPACE=/workspace

 USER turnstone

@@ -17,11 +17,17 @@ Named after the [Ruddy Turnstone](https://en.wikipedia.org/wiki/Ruddy_turnstone)

 **What is a harness?**

+<p align="center">
+  <a href="https://media.githubusercontent.com/media/turnstonelabs/turnstone/main/docs/diagrams/harness.png">
+    <img src="https://media.githubusercontent.com/media/turnstonelabs/turnstone/main/docs/diagrams/harness.png" alt="ℋ : s_{n+1} ~ T(s_n) for n < τ_H — the whole controlled loop: π lowers state to context, M_W proposes a readout, γ authorizes it, Q_E acts on the world, ρ verifies and folds back" width="960"/>
+  </a>
+</p>
+
 ```
-ℋ :  s_{n+1} ~ T(s_n)   for n < τ*,    T = ρ ∘ (M_W ∘ π, E)
+ℋ :  s_{n+1} ~ T(s_n)   for n < τ_H
 ```

-[**the primer →**](PRIMER.md)
+[**the primer →**](PRIMER.md) · [**the formalism →**](HYPOTHESIS.md)

 ### Release Tracks

@@ -53,8 +53,9 @@
 # The node registers in Postgres, auto-enrolls its mTLS cert from the console's
 # ACME endpoint (when the cluster runs mTLS), and the console collector reaches
 # it back via host.docker.internal. To join from ANOTHER machine, set
-# TURNSTONE_HOST_IP to this host's LAN IP and use it in the URLs above (and the
-# node's TURNSTONE_ADVERTISE_URL = the NODE host's IP) — see docs/docker.md.
+# TURNSTONE_HOST_IP to this host's LAN IP and set TURNSTONE_ACME_EXTERNAL_URL to
+# http://<this-host-ip>:8090/acme. Use the same host IP in the node's URLs above
+# (and the NODE host's IP in TURNSTONE_ADVERTISE_URL) — see docs/docker.md.
 # =============================================================================

 name: turnstone
@@ -154,9 +155,11 @@ services:
    # Publishes the console's plain-HTTP listener so a bare-metal node can reach
    # the ACME endpoint, fetch the CA, and enroll its cert (the console serves
    # HTTP here even under mTLS). Bound to 127.0.0.1 by default; setting
-    # TURNSTONE_HOST_IP exposes the WHOLE console HTTP API — including the
-    # cert-issuing ACME endpoint — on that interface, so the JWT secret's
-    # strength is the only gate. Browsers use Caddy :8443, never this port.
+    # TURNSTONE_HOST_IP exposes the WHOLE console HTTP API on that interface.
+    # ACME signing routes require a dedicated short-lived service JWT, but the
+    # listener and bearer token are still plain HTTP: bind only a trusted LAN
+    # or VPN interface and restrict it to enrolling nodes. Browsers use Caddy
+    # :8443, never this port.
    ports:
      - "${TURNSTONE_HOST_IP:-127.0.0.1}:8090:8090"
    environment:
@@ -164,6 +167,9 @@ services:
      TURNSTONE_DB_BACKEND: *db-backend
      TURNSTONE_DB_URL: *db-url
      TURNSTONE_CONSOLE_URL: http://console:8090
+      # Separate from TURNSTONE_CONSOLE_URL: this is the canonical responder
+      # base embedded in ACME directory/order URLs for cross-host enrollment.
+      TURNSTONE_ACME_EXTERNAL_URL: "${TURNSTONE_ACME_EXTERNAL_URL:-}"
    extra_hosts:
      - "host.docker.internal:host-gateway"
    networks:
@@ -305,6 +311,9 @@ services:
      SKIP_PERMISSIONS: ${SKIP_PERMISSIONS:-}
      TURNSTONE_NODE_ID: node-1
      TURNSTONE_ADVERTISE_URL: http://node-1:8080
+      # Lets the authenticated ACME client follow the console's canonical LAN
+      # URLs without trusting destinations learned from the public directory.
+      TURNSTONE_ACME_EXTERNAL_URL: "${TURNSTONE_ACME_EXTERNAL_URL:-}"
    extra_hosts:
      - "host.docker.internal:host-gateway"
    networks:
@@ -38,10 +38,18 @@ services:
        condition: service_completed_successfully
    volumes:
      - tls-certs:/certs:ro
+    # The production base keeps :8090 private. The TLS overlay publishes it on
+    # localhost for same-host enrollment; use a trusted LAN/VPN address for a
+    # remote node and firewall it to that node.
+    ports:
+      - "${TURNSTONE_CONSOLE_HTTP_BIND:-127.0.0.1}:8090:8090"
    environment:
      TURNSTONE_TLS_ENABLED: "true"
      TURNSTONE_TLS_SANS: "console"
      TURNSTONE_CONSOLE_URL: "http://console:8090"
+      # Canonical ACME responder base advertised to enrolling nodes. Set this
+      # when they reach the console through a different host/address.
+      TURNSTONE_ACME_EXTERNAL_URL: "${TURNSTONE_ACME_EXTERNAL_URL:-}"
    command:
      - turnstone-console
      - --host=0.0.0.0
@@ -58,11 +66,7 @@ services:
    environment:
      TURNSTONE_TLS_ENABLED: "true"
      TURNSTONE_TLS_SANS: "server"
-    # Disable healthcheck — server serves HTTPS with mTLS which the
-    # stdlib healthcheck script can't satisfy. The base compose
-    # healthcheck uses plain HTTP which won't work on an HTTPS listener.
-    healthcheck:
-      disable: true
+      TURNSTONE_ACME_EXTERNAL_URL: "${TURNSTONE_ACME_EXTERNAL_URL:-}"

  # Channel: TLS
  channel:
@@ -2,11 +2,11 @@ apiVersion: v2
 name: turnstone
 description: Multi-node AI orchestration platform with tool use, agent routing, and cluster simulation
 type: application
-version: 0.1.0
+version: 0.2.0
 appVersion: "0.3.0"

 dependencies:
  - name: postgresql
-    version: ~18.7.0
+    version: ~18.8.0
    repository: https://charts.bitnami.com/bitnami
    condition: postgresql.enabled
@@ -110,6 +110,153 @@ Determine the PostgreSQL username.
 {{- end }}
 {{- end }}

+{{/*
+The PostgreSQL password when the chart stores it itself, empty when it
+does not. Doubles as the predicate for "does <fullname>-secrets need to
+carry POSTGRES_PASSWORD", so an inline password is never written
+anywhere but <fullname>-secrets, and an operator-supplied Secret is
+never duplicated into it.
+
+An operator-supplied existingSecret wins outright: writing the value
+into a second Secret nothing reads would only duplicate a credential.
+
+Both branches need "default" because this is reached through include,
+which captures rendered text rather than a value: a key that is unset
+rather than empty — "password:" with nothing after it — renders as the
+literal "<no value>", and a ten-character string is truthy. Without the
+default that lands base64-encoded in POSTGRES_PASSWORD and the workloads
+authenticate with it.
+*/}}
+{{- define "turnstone.db.inlinePassword" -}}
+{{- if .Values.postgresql.enabled }}
+{{- .Values.postgresql.auth.password | default "" }}
+{{- else if not .Values.database.external.existingSecret }}
+{{- .Values.database.external.password | default "" }}
+{{- end }}
+{{- end }}
+
+{{/*
+The name of the bundled subchart's own Secret.
+
+Mirrors the subchart's naming rather than calling its helpers, which
+expect a context scoped to the subchart that this chart cannot hand
+them. Release-derived, so deliberately not turnstone.fullname: a
+fullnameOverride here renames this chart's resources and leaves the
+subchart's alone, and pointing at "<fullname>-postgresql" would then
+name a Secret that does not exist.
+
+The subchart also normalises the release name through a regex before
+using it, which is a no-op for the DNS-1123 names Helm accepts, so it is
+not reproduced.
+*/}}
+{{- define "turnstone.postgresql.fullname" -}}
+{{- $global := ((.Values.global).postgresql).fullnameOverride }}
+{{- if $global }}
+{{- $global | trunc 63 | trimSuffix "-" }}
+{{- else if .Values.postgresql.fullnameOverride }}
+{{- .Values.postgresql.fullnameOverride | trunc 63 | trimSuffix "-" }}
+{{- else }}
+{{- $name := .Values.postgresql.nameOverride | default "postgresql" }}
+{{- if contains $name .Release.Name }}
+{{- .Release.Name | trunc 63 | trimSuffix "-" }}
+{{- else }}
+{{- printf "%s-%s" .Release.Name $name | trunc 63 | trimSuffix "-" }}
+{{- end }}
+{{- end }}
+{{- end }}
+
+{{- define "turnstone.postgresql.secretName" -}}
+{{- $existing := coalesce (((.Values.global).postgresql).auth).existingSecret .Values.postgresql.auth.existingSecret }}
+{{- if $existing }}
+{{- tpl $existing . }}
+{{- else }}
+{{- include "turnstone.postgresql.fullname" . }}
+{{- end }}
+{{- end }}
+
+{{/*
+The subchart stores the named user's password under "password" and the
+superuser's under "postgres-password", and lets an operator rename
+either through auth.secretKeys.
+*/}}
+{{- define "turnstone.postgresql.passwordKey" -}}
+{{- $user := .Values.postgresql.auth.username | default "" }}
+{{- $keys := .Values.postgresql.auth.secretKeys | default dict }}
+{{- if or (empty $user) (eq $user "postgres") }}
+{{- $keys.adminPasswordKey | default "postgres-password" }}
+{{- else }}
+{{- $keys.userPasswordKey | default "password" }}
+{{- end }}
+{{- end }}
+
+{{/*
+Determine the secret holding the PostgreSQL password, and the key within
+it. Three sources, and the two helpers agree by construction because
+they branch identically:
+
+  - an external database pointed at a Secret the chart does not own (a
+    CloudNativePG-generated secret, an External Secrets target, ...), in
+    which case the key is rarely "POSTGRES_PASSWORD" — hence the
+    companion existingSecretPasswordKey
+  - the bundled subchart's own Secret, when it generates the password
+  - <fullname>-secrets, when the password is supplied inline in values
+
+Note the last is deliberately not turnstone.llm.secretName: that
+resolves to llm.existingSecret when the operator supplies one, which
+holds LLM API keys and has no reason to carry a database password.
+*/}}
+{{- define "turnstone.db.secretName" -}}
+{{- if not .Values.postgresql.enabled }}
+{{- if .Values.database.external.existingSecret }}
+{{- .Values.database.external.existingSecret }}
+{{- else }}
+{{- printf "%s-secrets" (include "turnstone.fullname" .) }}
+{{- end }}
+{{- else if include "turnstone.db.inlinePassword" . }}
+{{- printf "%s-secrets" (include "turnstone.fullname" .) }}
+{{- else }}
+{{- include "turnstone.postgresql.secretName" . }}
+{{- end }}
+{{- end }}
+
+{{- define "turnstone.db.passwordKey" -}}
+{{- if not .Values.postgresql.enabled }}
+{{- if .Values.database.external.existingSecret }}
+{{- .Values.database.external.existingSecretPasswordKey | default "password" }}
+{{- else }}
+{{- printf "POSTGRES_PASSWORD" }}
+{{- end }}
+{{- else if include "turnstone.db.inlinePassword" . }}
+{{- printf "POSTGRES_PASSWORD" }}
+{{- else }}
+{{- include "turnstone.postgresql.passwordKey" . }}
+{{- end }}
+{{- end }}
+
+{{/*
+Database environment shared by the server, console and migrate Job.
+
+Every value except the password is rendered inline rather than pulled
+from the ConfigMap via envFrom, so that one definition serves all three
+workloads and the URL is assembled in exactly one place.
+
+POSTGRES_PASSWORD must still precede TURNSTONE_DB_URL: the kubelet
+expands $(VAR) only against env entries declared earlier in the list, so
+a later definition would leave a literal "$(POSTGRES_PASSWORD)" in the
+URL.
+*/}}
+{{- define "turnstone.db.env" -}}
+- name: TURNSTONE_DB_BACKEND
+  value: {{ .Values.database.backend | quote }}
+- name: POSTGRES_PASSWORD
+  valueFrom:
+    secretKeyRef:
+      name: {{ include "turnstone.db.secretName" . }}
+      key: {{ include "turnstone.db.passwordKey" . }}
+- name: TURNSTONE_DB_URL
+  value: "postgresql+psycopg://{{ include "turnstone.postgresql.username" . }}:$(POSTGRES_PASSWORD)@{{ include "turnstone.postgresql.host" . }}:{{ include "turnstone.postgresql.port" . }}/{{ include "turnstone.postgresql.database" . }}{{ if and (not .Values.postgresql.enabled) .Values.database.external.sslmode }}?sslmode={{ .Values.database.external.sslmode }}{{ end }}"
+{{- end }}
+
 {{/*
 Determine the secret name for LLM API keys.
 */}}
@@ -7,6 +7,17 @@ metadata:
    app.kubernetes.io/component: console
 spec:
  replicas: {{ .Values.console.replicas }}
+  {{- if eq (int .Values.console.replicas) 1 }}
+  # The console registers itself under the fixed service_id "console" and
+  # deregisters on shutdown. Under RollingUpdate the outgoing pod's
+  # deregister runs *after* the incoming pod registers and deletes its
+  # row -- and the heartbeat only touches last_heartbeat, so the row is
+  # never recreated and the console stays invisible in the registry until
+  # the next clean start. Recreate orders shutdown strictly before
+  # startup. Only valid at one replica; see console.replicas.
+  strategy:
+    type: Recreate
+  {{- end }}
  selector:
    matchLabels:
      {{- include "turnstone.selectorLabels" . | nindent 6 }}
@@ -18,6 +29,18 @@ spec:
        app.kubernetes.io/component: console
    spec:
      serviceAccountName: {{ include "turnstone.serviceAccountName" . }}
+      {{- with .Values.console.nodeSelector }}
+      nodeSelector:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.console.affinity }}
+      affinity:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.console.tolerations }}
+      tolerations:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
      containers:
        - name: console
          image: {{ include "turnstone.image" . }}
@@ -36,8 +59,18 @@ spec:
            - secretRef:
                name: {{ include "turnstone.llm.secretName" . }}
                optional: true
-          {{- if or .Values.auth.existingSecret .Values.auth.jwtSecret }}
          env:
+            {{- include "turnstone.db.env" . | nindent 12 }}
+            # Self-registration URL for the service registry. Unlike a
+            # server node the console is one logical endpoint behind its
+            # Service, so the Service DNS name is correct here. Without
+            # it the console registers gethostname() (its pod name),
+            # which no server node can resolve. Stops at ".svc" rather
+            # than assuming a "cluster.local" DNS domain, which is
+            # configurable per cluster.
+            - name: TURNSTONE_CONSOLE_URL
+              value: "http://{{ include "turnstone.fullname" . }}-console.{{ .Release.Namespace }}.svc:{{ .Values.console.service.port }}"
+          {{- if or .Values.auth.existingSecret .Values.auth.jwtSecret }}
            - name: TURNSTONE_JWT_SECRET
              valueFrom:
                secretKeyRef:
@@ -18,6 +18,18 @@ spec:
        app.kubernetes.io/component: server
    spec:
      serviceAccountName: {{ include "turnstone.serviceAccountName" . }}
+      {{- with .Values.server.nodeSelector }}
+      nodeSelector:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.server.affinity }}
+      affinity:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.server.tolerations }}
+      tolerations:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
      containers:
        - name: server
          image: {{ include "turnstone.image" . }}
@@ -39,8 +51,20 @@ spec:
                name: {{ include "turnstone.llm.secretName" . }}
                optional: true
          env:
-            - name: TURNSTONE_DB_URL
-              value: "postgresql+psycopg://$(TURNSTONE_DB_USER):$(POSTGRES_PASSWORD)@$(TURNSTONE_DB_HOST):$(TURNSTONE_DB_PORT)/$(TURNSTONE_DB_NAME)"
+            {{- include "turnstone.db.env" . | nindent 12 }}
+            # Each replica is a distinct node in the rendezvous ring, so it
+            # must advertise an address that reaches *itself*. The Service
+            # DNS name would load-balance across every replica, sending
+            # console traffic routed for node A to an arbitrary pod; the
+            # default (gethostname(), i.e. the pod name) is not resolvable
+            # at all. The pod IP is unique, routable in-cluster, and
+            # re-registered on every start, so churn is self-healing.
+            - name: POD_IP
+              valueFrom:
+                fieldRef:
+                  fieldPath: status.podIP
+            - name: TURNSTONE_ADVERTISE_URL
+              value: "http://$(POD_IP):{{ .Values.server.service.port }}"
          {{- if or .Values.auth.existingSecret .Values.auth.jwtSecret }}
            - name: TURNSTONE_JWT_SECRET
              valueFrom:
@@ -6,11 +6,23 @@ metadata:
    {{- include "turnstone.labels" . | nindent 4 }}
    app.kubernetes.io/component: migrate
  annotations:
-    "helm.sh/hook": pre-install,pre-upgrade
+    # post-install, not pre-install: on a first install nothing the
+    # migration needs exists yet — not the ConfigMap, not the Secret, and
+    # with the bundled subchart not the database either, since Helm
+    # creates ordinary resources only once hooks have finished. On an
+    # upgrade all of it is already running, so pre-upgrade is both safe
+    # and preferable: migrations land before the new code rolls out
+    # rather than after.
+    "helm.sh/hook": post-install,pre-upgrade
    "helm.sh/hook-weight": "-1"
    "helm.sh/hook-delete-policy": before-hook-creation,hook-succeeded
 spec:
-  backoffLimit: 3
+  # Helm does not wait for the database to be ready before running
+  # post-install hooks, so on a first install this Job is what waits: it
+  # exits non-zero until PostgreSQL accepts connections, and the retry
+  # budget has to cover a cold StatefulSet pulling its image and
+  # initialising.
+  backoffLimit: 10
  template:
    metadata:
      labels:
@@ -19,6 +31,18 @@ spec:
    spec:
      serviceAccountName: {{ include "turnstone.serviceAccountName" . }}
      restartPolicy: OnFailure
+      {{- with .Values.migrate.nodeSelector }}
+      nodeSelector:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.migrate.affinity }}
+      affinity:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      {{- with .Values.migrate.tolerations }}
+      tolerations:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
      containers:
        - name: migrate
          image: {{ include "turnstone.image" . }}
@@ -27,12 +51,5 @@ spec:
            - python
            - -m
            - turnstone.core.storage._migrate
-          envFrom:
-            - configMapRef:
-                name: {{ include "turnstone.fullname" . }}-config
-            - secretRef:
-                name: {{ include "turnstone.llm.secretName" . }}
-                optional: true
          env:
-            - name: TURNSTONE_DB_URL
-              value: "postgresql+psycopg://$(TURNSTONE_DB_USER):$(POSTGRES_PASSWORD)@$(TURNSTONE_DB_HOST):$(TURNSTONE_DB_PORT)/$(TURNSTONE_DB_NAME)"
+            {{- include "turnstone.db.env" . | nindent 12 }}
@@ -1,4 +1,19 @@
-{{- if not .Values.llm.existingSecret }}
+{{/*
+This Secret backs every credential supplied inline in values, so it is
+rendered whenever any one of them is set — not, as it once was, only
+when llm.existingSecret is empty. Under that older gate an operator who
+supplied an LLM Secret lost the unrelated inline values with it: both
+POSTGRES_PASSWORD and TURNSTONE_JWT_SECRET silently went unrendered
+while the workloads went on referencing them, so every pod stalled in
+CreateContainerConfigError.
+
+Each key keeps its own condition, so an operator-supplied Secret still
+suppresses the value it replaces and nothing else.
+*/}}
+{{- $apiKey := and .Values.llm.apiKey (not .Values.llm.existingSecret) }}
+{{- $dbPassword := include "turnstone.db.inlinePassword" . }}
+{{- $jwtSecret := and .Values.auth.jwtSecret (not .Values.auth.existingSecret) }}
+{{- if or $apiKey $dbPassword $jwtSecret }}
 apiVersion: v1
 kind: Secret
 metadata:
@@ -7,15 +22,13 @@ metadata:
    {{- include "turnstone.labels" . | nindent 4 }}
 type: Opaque
 data:
-  {{- if .Values.llm.apiKey }}
+  {{- if $apiKey }}
  OPENAI_API_KEY: {{ .Values.llm.apiKey | b64enc | quote }}
  {{- end }}
-  {{- if and .Values.postgresql.enabled .Values.postgresql.auth.password }}
-  POSTGRES_PASSWORD: {{ .Values.postgresql.auth.password | b64enc | quote }}
-  {{- else if and (not .Values.postgresql.enabled) .Values.database.external.password }}
-  POSTGRES_PASSWORD: {{ .Values.database.external.password | b64enc | quote }}
+  {{- if $dbPassword }}
+  POSTGRES_PASSWORD: {{ $dbPassword | b64enc | quote }}
  {{- end }}
-  {{- if and .Values.auth.jwtSecret (not .Values.auth.existingSecret) }}
+  {{- if $jwtSecret }}
  TURNSTONE_JWT_SECRET: {{ .Values.auth.jwtSecret | b64enc | quote }}
  {{- end }}
 {{- end }}
@@ -14,7 +14,13 @@ database:
    port: 5432
    database: turnstone
    username: turnstone
+    # Secret holding the password for `username`. Leave empty to supply
+    # `password` inline below instead.
    existingSecret: ""
+    # Key within existingSecret holding the password. CloudNativePG
+    # generates "password"; other operators differ.
+    existingSecretPasswordKey: password
+    password: ""
    sslmode: prefer

 # -- Bitnami PostgreSQL subchart
@@ -37,6 +43,10 @@ server:
  service:
    type: ClusterIP
    port: 8080
+  # -- Node scheduling constraints
+  nodeSelector: {}
+  affinity: {}
+  tolerations: []

 # -- Turnstone console (cluster dashboard)
 console:
@@ -51,6 +61,17 @@ console:
  service:
    type: ClusterIP
    port: 8090
+  # -- Node scheduling constraints
+  nodeSelector: {}
+  affinity: {}
+  tolerations: []
+
+# -- Database migration Job (post-install/pre-upgrade hook)
+migrate:
+  # -- Node scheduling constraints
+  nodeSelector: {}
+  affinity: {}
+  tolerations: []

 # -- LLM provider configuration
 llm:
@@ -15,11 +15,14 @@ counterpart to the quick `turnstone-server …` invocation in
 ## Cluster-side prerequisite

 The compose stack must publish Postgres, the console's ACME endpoint, and SearxNG
-on an address the bare-metal host can reach. Start it with `TURNSTONE_HOST_IP`
-set to the compose host's LAN IP (default `127.0.0.1` keeps everything host-local):
+on an address the bare-metal host can reach. Use a trusted LAN or VPN interface,
+firewall it to the joining node, and advertise the same reachable ACME endpoint
+(default `127.0.0.1` keeps everything host-local):

 ```bash
-TURNSTONE_HOST_IP=<compose-host-ip> docker compose up -d
+TURNSTONE_HOST_IP=<compose-host-ip> \
+TURNSTONE_ACME_EXTERNAL_URL=http://<compose-host-ip>:8090/acme \
+  docker compose up -d
 ```

 ## Install (run as root on the bare-metal host)
@@ -65,8 +68,20 @@ journalctl -u turnstone-server -f          # watch it register + (if the cluster
 shared settings (the database). If the cluster runs mTLS, the node auto-enrolls a
 cert from the console's ACME endpoint and re-advertises itself over `https://`.

-> **mTLS + cross-host caveat:** a node on a *different* host than the console
-> currently can't complete ACME enrollment — the console advertises an
-> unroutable in-container address in its ACME directory
-> ([turnstonelabs/lacme#22](https://github.com/turnstonelabs/lacme/issues/22)).
-> Same-host bare-metal nodes, and any node in a non-mTLS cluster, are unaffected.
+For a node on a different host, `TURNSTONE_ACME_EXTERNAL_URL` is required on the
+console and should also be set in the node drop-in. It is the full, externally
+reachable responder base
+(including `/acme`) that the console embeds in the ACME protocol's follow-up
+URLs and that the node trusts as an enrollment-credential destination. The
+node's `TURNSTONE_CONSOLE_URL` should point at the same host and port, without
+the `/acme` suffix.
+
+For mTLS, `TURNSTONE_ADVERTISE_URL` may use a resolvable DNS hostname or a
+literal IP address. Turnstone enrolls literals as IP SANs. Bracket IPv6 literals
+inside URLs, for example `http://[2001:db8::10]:8080`; do not use wildcard,
+unspecified, or scoped addresses as certificate identities. Restart the node
+after changing its advertised identity so it enrolls a matching certificate.
+
+The dedicated service JWT authenticates enrollment but the direct `:8090`
+bootstrap is still plain HTTP/TOFU. Use HTTPS through an independently trusted
+proxy when the network itself is not trusted.
@@ -4,16 +4,20 @@
 # they live here; the JWT secret + DB URL live in /etc/turnstone/config.toml.
 #
 # Addresses below use RFC 5737 documentation IPs — replace them:
-#   <this-host>     = the bare-metal host's own LAN IP (what the console dials back)
+#   <this-host>     = the bare-metal host's own reachable DNS name or IP address
+#                     (its mTLS identity and the address peers dial)
 #   <compose-host>  = the host running the cluster / docker-compose stack, started
-#                     with TURNSTONE_HOST_IP=<compose-host> so :8090 and :8081 are
-#                     published on its LAN interface (see docs/docker.md).
+#                     with TURNSTONE_HOST_IP=<compose-host> and
+#                     TURNSTONE_ACME_EXTERNAL_URL=http://<compose-host>:8090/acme
+#                     so enrollment links and published ports are reachable
+#                     (see docs/docker.md).
 [Service]
 # Unique node id (defaults to the hostname if unset).
 Environment=TURNSTONE_NODE_ID=host-1

 # The address peers + the console collector dial back. Auto-upgrades to https://
-# once the node enrolls its mTLS cert.
+# once the node enrolls its mTLS cert. IPv6 literals require URL brackets, for
+# example http://[2001:db8::10]:8080.
 Environment=TURNSTONE_ADVERTISE_URL=http://192.0.2.10:8080

 # The cluster console's reachable plain-HTTP ACME/API endpoint. A bare-metal node
@@ -21,5 +25,9 @@ Environment=TURNSTONE_ADVERTISE_URL=http://192.0.2.10:8080
 # port; turnstone-server honors this for cert enrollment.
 Environment=TURNSTONE_CONSOLE_URL=http://192.0.2.1:8090

+# Trusted canonical responder base. This pins where the node may send its
+# enrollment JWT; it must match the console-side value (a literal IP is fine).
+Environment=TURNSTONE_ACME_EXTERNAL_URL=http://192.0.2.1:8090/acme
+
 # The cluster's published SearxNG, for the web_search tool.
 Environment=TURNSTONE_SEARXNG_URL=http://192.0.2.1:8081
@@ -195,13 +195,17 @@ both and the gateway hosts both adapters in one process.
 - All subsequent messages in the thread are routed to the same workstream.
 - The bot streams responses via message edits, updated approximately every
  1.5 seconds.
- If the workstream is evicted for capacity, the next message in the
-  thread auto-creates a new workstream and atomically resumes the
-  previous workstream via the `resume_ws` field on
-  `CreateWorkstreamMessage`. The server resumes the workstream during
-  creation (same HTTP request), and the server emits a
-  `WorkstreamResumedEvent` back to the channel. The thread receives a
-  *"Resumed: {name} ({count} messages restored)"* confirmation.
+- If a persisted channel route is no longer active on its owning node, the
+  router asks the create endpoint to fork the old workstream into a new ID via
+  `resume_ws`. The saved source can still resolve normally; its
+  checkpoint-bounded history, configuration, persona, effective project, and
+  attachment references are cloned before the channel route is repointed. The
+  old route remains durable until the replacement (and any initial message)
+  succeeds. If the create endpoint returns the ordinary
+  source-not-found response *and* a fresh authoritative storage lookup confirms
+  that the source is gone, the router retries once without `resume_ws` and
+  starts a fresh conversation. Other access, conflict, routing, and storage
+  failures remain visible rather than silently discarding history.

 ### Slash Commands

@@ -284,15 +288,17 @@ See [Security: Database Schema](security.md#database-schema) for the
   `channel_routes` table.
 2. **Active** — messages are routed bidirectionally. The bot streams
   responses via message edits (updated every ~1.5 seconds).
-3. **Eviction** — the server evicts an idle workstream for capacity. The
-   route is preserved and the thread stays open.
-4. **Reactivation** — the next message in the thread detects the stale
-   route and creates a new workstream with the old `ws_id`
-   as `resume_ws` on the creation request. The server resumes
-   the workstream during creation (no separate command or reverse lookup
-   needed). The channel receives a `WorkstreamResumedEvent`, and
-   the thread displays *"Resumed: {name} ({count} messages restored)"*.
-   If the old workstream was pruned, a fresh one starts with no error.
+3. **Eviction** — the server evicts an idle workstream for capacity. Its saved
+   source row and channel route remain durable, and the thread stays open.
+4. **Reactivation** — the next message resolves the saved route and probes
+   whether that workstream is live on its owning node. If it is not, the router
+   creates a distinct workstream with the old `ws_id` as `resume_ws`. The
+   create response confirms the fork and message count; there is no separate
+   resume command or channel-specific resumed event. Only after the replacement
+   succeeds does the router swap the persisted route. If the source was deleted
+   or pruned, an exact source-not-found response plus a second authoritative
+   storage miss triggers one fresh-create retry; other fork failures leave the
+   old route intact and are surfaced normally.
 5. **Close** — `/close` command closes the workstream via HTTP, deletes the
   route, unsubscribes from events, and archives the Discord thread.

@@ -174,17 +174,32 @@ Request:
 {
  "node_id": "db-west-04",
  "name": "perf-analysis",
-  "model": "gpt-5"
+  "model": "gpt-5",
+  "project_id": "proj_analytics",
+  "initial_message": "Profile the slow query"
 }
 ```

 All fields are optional:
 - `node_id` — targeting mode:
  - **omitted or `"auto"`** — console picks the reachable node with the most available capacity (max_ws - ws_total) and proxies the request to it.
-  - **`"pool"`** — console picks a reachable node with available capacity using round-robin selection.
+  - **`"pool"`** — compatibility alias for automatic placement on the reachable node with the most headroom.
  - **specific node ID** — proxies the request to that node directly.
 - `name` — workstream display name. Auto-generated if omitted.
 - `model` — model alias from the target node's registry. Uses the node's default model if omitted.
+- `judge_model` — optional judge-model alias for this workstream.
+- `initial_message` — first message dispatched after the workstream is published.
+- `skill` — enabled profile/skill to snapshot onto a fresh workstream.
+- `persona` — enabled persona slug; empty uses the interactive default.
+- `project_id` — project to attach, subject to the target node's membership gate.
+- `resume_ws` — source ID to **fork** atomically into a new workstream. The
+  source remains unchanged; its checkpoint-bounded history, configuration,
+  persona, project, and attachment references are copied transactionally.
+
+The endpoint also accepts the same multipart create shape as a node: one
+JSON-encoded `meta` field plus up to ten `file` parts. Files require an
+`initial_message` in the dashboard launcher. Files cannot be combined with
+`resume_ws`; fork first and upload on the new workstream.

 Response:

@@ -196,7 +211,19 @@ Response:
 }
 ```

-The response confirms the workstream creation request was proxied to the target node. A `ws_created` event on the cluster SSE stream confirms the workstream was actually created.
+The response is returned only after the target node has durably published the
+workstream. Its hidden `creating` reservation has already crossed to `idle`,
+and the node emitted `ws_created` before any initial-message state event. The
+cluster SSE event may therefore arrive before or after the HTTP response;
+clients should reconcile both by the returned `correlation_id`/workstream ID
+rather than treating them as two creates.
+
+For safety, the console masks most target-node failures as the opaque `502`
+shape `{"error":"Dispatch to node <node_id> failed"}` instead of reflecting
+arbitrary node text or retry-triggering 401/429 responses. The coded
+`server.require_project` refusal is the exception and remains a `400` with
+actionable wording. Consult the target node's logs for the underlying create
+correlation when a reachable node returns a masked 502.

 ### `GET /v1/api/cluster/events`

@@ -310,8 +337,8 @@ The auth system uses three scopes instead of the earlier read/full role model:
 | Scope | Grants |
 |-------|--------|
 | `read` | Read-only access: dashboards, workstream lists, SSE streams, health |
-| `write` | Send messages, create/close workstreams, approve tool calls |
-| `approve` | Admin operations: manage users and API tokens |
+| `write` | Non-approval mutations: send, create/open/close/delete, cancel, attachments, rewind, and retry |
+| `approve` | Tool-approval and admin HTTP surfaces (with their additional RBAC permission checks) |

 Scopes are cumulative — a user with `approve` scope can also perform `write` and `read` operations.

@@ -348,52 +375,87 @@ SSE streams (`/v1/api/workstreams/{ws_id}/events`, `/v1/api/events/global`) are

 ### Authentication

-The proxy mints a short-lived (5-minute) JWT per request carrying the real user's `user_id`, `scopes`, and `permissions` with `aud: turnstone-server`.  The user's console JWT (`aud: turnstone-console`) cannot be forwarded directly — it would be rejected by the server's audience validation — so the console re-signs a new server-audience JWT from the validated `AuthResult`.  This preserves audit attribution (the upstream server sees the real user, not a service identity) and enforces scope narrowing as defense in depth (a read-only console user's proxied request carries only `read` scope).  The JWT `src` claim is set to `"console-proxy"` for audit traceability.  When no user context is available (auth disabled), the proxy falls back to a `ServiceTokenManager` with service identity `console-proxy`.  The static `--auth-token` / `proxy_auth_token` is used as a final fallback.
+The proxy mints a short-lived (5-minute) JWT per request carrying the real user's `user_id`, `scopes`, and `permissions` with `aud: turnstone-server`.  The user's console JWT (`aud: turnstone-console`) cannot be forwarded directly — it would be rejected by the server's audience validation — so the console re-signs a new server-audience JWT from the validated `AuthResult`.  This preserves audit attribution (the upstream server sees the real user, not a service identity) and enforces scope narrowing as defense in depth (a read-only console user's proxied request carries only `read` scope). Ordinary users are re-minted with `src="console-proxy"`; coordinator tokens retain `src="coordinator"` plus `coord_ws_id`, and only the validated console service identity with `service` scope retains `src="console"` for trusted owner forwarding. When no user context is available, the proxy falls back to a `ServiceTokenManager` identity `console-proxy` carrying `src="console"` and `{read, write, approve, service}` scopes.  The static `--auth-token` / `proxy_auth_token` is used as a final fallback.

 ---

 ## Browser Dashboard

-The web UI has five views, toggled client-side:
+The console uses an L-shaped application shell: a collapsible navigation rail,
+a tab bar, and a pane host. On mobile the rail becomes an off-canvas drawer.
+The rail is fed by the cluster SSE snapshot and shows:

-### 1. Cluster Overview (landing)
+- state/count filters and the live compute-node list, including version drift;
+- active coordinator and interactive workstreams, nested under their
+  coordinator parent and grouped by project when project metadata is visible;
+- permission-filtered Manage groups that open the singleton Admin pane.

- **State cards** — 5 clickable cards (running, thinking, attention, idle, error) with count and colored top border. Clicking filters to that state.
- **Aggregate bar** — total tokens and tool calls across the cluster.
- **Node table** — columns: NODE, WS, RUN, ATTN, TOKENS, VER, LOAD. Sorted by activity. Clickable rows drill down to node detail. Version column shows per-node version; hidden on mobile.
- **Version drift indicator** — when nodes report different versions, the status bar shows a yellow "DRIFT" warning with a tooltip listing all versions. Node groups show "mixed" with a yellow badge when their members disagree.
- **"+ new" button** — opens the workstream creation modal (see below).
+Coordinator and interactive conversations open as tabs inside the same shell.
+Interactive panes use the owning node's console proxy, so users do not need
+direct network access to compute-node ports. Split-right and split-down actions
+can display several panes at once. Closing a pane removes only that tab; use the
+pane menu's explicit close or delete action to change the workstream lifecycle.

-### 2. Node Drill-down
+### Dashboard pane

-Breadcrumb: `Cluster > db-west-04`. Shows the node's workstreams in a table matching the per-node dashboard layout (STATE, NAME, MODEL, NODE, TASK, TOKENS, CTX) with activity sub-lines. Includes a link to the node's proxied server UI.
+The home view is coordinator-first. It contains the persistent workstream
+launcher plus the saved-sessions list. Selecting a state count opens the
+filtered workstream table inside the same Dashboard pane; selecting a compute
+node opens its proxied node surface. Cluster SSE updates keep rail state,
+workstream rows, and tab state glyphs synchronized.

-**Proxy deep-linking:** Clicking a workstream row opens the node's server UI in a new tab via the proxy at `/node/{node_id}/?ws_id=<id>`, which auto-selects that workstream. Users do not need direct network access to the server node.
+### Workstream launcher

-### 3. Filtered Workstreams
+The landing-page composer starts a workstream with an optional initial task and
+attachments. When the caller can create both kinds, a Coordinator / Interactive
+toggle selects the target kind. Its options include:

-Breadcrumb: `Cluster > Running` or `Cluster > db-west-04`. Server-side paginated workstream table. NODE column values are clickable to filter further. Pagination controls at bottom. Workstream rows use proxy deep-links.
-
-### 4. Workstream Creation Modal
-
-Triggered by the "+ new" header button. A modal dialog with:
-
- **Node selector** — dropdown with three targeting modes: "Auto (best available)" picks the node with the most headroom, "General pool (any node)" picks a node with available capacity using round-robin, or a specific node from the list (showing capacity).
+- **Node placement** — "Least loaded" picks the reachable node with the most
+  headroom, or "Specific node" pins the create to a node from the live list.
 - **Persona** — optional dropdown listing the enabled personas for the workstream kind. Sets the system-message composition and capability envelope at creation, snapshotted server-side; empty uses the kind's default. Picking one requires no `persona.*` permission.
- **Profile** — optional dropdown listing enabled skills. Applies the skill's model, auto-approve policy, token budget, and other behavioral settings at creation time.
+- **Skill** — optional dropdown listing enabled skills. Applies the skill's model, auto-approve policy, token budget, and other behavioral settings at creation time.
+- **Project** — optional project filing. Private projects require owner/member access. A coordinator child inherits its parent's project unless explicitly routed to another attachable project.
 - **Name** — optional text input. Auto-generated if left empty.
- **Model** — optional text input for a model alias from the target node's registry.
- **Judge Model** — optional text input for the judge model alias (overrides the default judge model for this workstream).
+- **Model** — optional selector populated from the target model registry.
+- **Judge Model** — optional selector for the judge alias (overrides the default
+  judge model for this workstream).

-Keyboard shortcuts: Ctrl+Shift+R (refresh title), Ctrl+Shift+E (edit title), Ctrl+Shift+F (fork), Ctrl+Shift+X (delete). Press ? for full shortcut help.
+Interactive launches additionally expose node strategy / node selection.
+Submitting uses `POST /v1/api/cluster/workstreams/new`; coordinator launches use
+the console's coordinator create surface. A toast confirms the committed
+create, while SSE updates the dashboard and opens the resulting pane.

-On submit, `POST /v1/api/cluster/workstreams/new` dispatches the creation request. A toast confirms success; the SSE stream delivers the `ws_created` event to update the dashboard.
+Files require a non-empty initial task so the first turn consumes the staged
+attachments. The console shell does not currently expose a fork action; use the
+node's standalone workstream UI or the create API's `resume_ws` field.

-All five views receive live updates via SSE — state cards update counts, node rows update metrics, workstream rows update state indicators.
+### Large pasted text

-The browser maintains a local `clusterState` object that mirrors the cluster snapshot. It is initialized from the SSE `snapshot` event on connect (or via `GET /v1/api/cluster/snapshot` on initial page load) and updated incrementally by SSE events. View navigation reads from local state — no API round-trips needed after the initial snapshot.
+Browser composers turn plain text longer than 2,000 Unicode code points into a
+`text/plain` attachment named `pasted-text.txt`. A paste exactly at the
+threshold stays inline. This applies to the interactive and coordinator send
+boxes, the console home launcher, and the node dashboard and new-workstream
+composers.

-### 5. Admin Panel
+Clipboard files take priority over clipboard text. Text larger than the 512 KiB
+attachment ceiling also stays inline, so the browser does not discard it before
+a rejected upload. Attachments require a companion message and cannot be sent
+as live-turn interjections; a busy composer preserves its message and chips for
+an idle retry.
+
+### Saved and filtered sessions
+
+Saved coordinator and interactive sessions share one list with kind and persona
+labels, filtering, pagination, and multi-select deletion. Opening a saved
+coordinator rehydrates it in the console; opening a saved interactive session
+resolves its node, calls `open`, and then connects the node-proxied pane.
+
+The filtered live table carries STATE, NAME, MODEL, NODE, TASK, TOKENS, and CTX
+columns. The browser maintains a local `clusterState` initialized from the
+cluster snapshot and updated incrementally by SSE; the filtered view normally
+renders from that state without another API round trip.
+
+### Admin pane

 Accessed via the "admin" button in the header (visible when authenticated
 with `approve` scope). Provides user, API token, channel link, MCP server,
@@ -405,8 +467,14 @@ Audit tabs, and [Settings](settings.md) for the database-backed
 configuration editor.

 The **Channels** tab links users to either a Discord or Slack account
-via a per-row channel-type selector.  The **Models** tab is a CRUD
-editor for `model_definitions`, the **Nodes** tab edits per-node
+via a per-row channel-type selector. The **Models** tab is a CRUD
+editor for `model_definitions`, including static and dynamic backend-auth
+modes and a per-process **Max concurrent generations** limit for each alias
+(`0` means unlimited). The limit is shared by every model-backed role using
+that alias and a streaming generation holds its slot through the full decode.
+Model edits rebind existing workstreams at their next send while
+in-flight requests keep their original definition snapshot; see
+[Settings](settings.md#model-definition-reloads) for the full contract. The **Nodes** tab edits per-node
 metadata, and the **TLS** tab manages CA and leaf certificates for the
 internal mTLS fabric.  The **Settings** tab edits ConfigStore values
 live; edits apply without restart.
@@ -504,7 +572,7 @@ Run history is automatically pruned (runs older than 90 days) approximately once
 | Mode | Behavior |
 |------|----------|
 | `auto` | Picks the reachable node with the most available capacity |
-| `pool` | Picks a reachable node with available capacity using round-robin |
+| `pool` | Compatibility alias for the reachable node with the most headroom |
 | `all` | Fan-out to all reachable nodes (capped at `max_fan_out`, default 20) |
 | `<node_id>` | Targets a specific node by ID |

@@ -664,4 +732,7 @@ turnstone-server --port 8080
 turnstone-console --port 8090
 ```

-Open `http://localhost:8090` for the cluster dashboard. Create workstreams via the "+ new" button. Click any workstream to open the proxied server UI — no direct access to server ports required.
+Open `http://localhost:8090` for the cluster dashboard. Create workstreams from
+the persistent Dashboard launcher. Selecting a workstream opens a coordinator
+or node-proxied interactive pane in the console shell — no direct access to
+server ports is required.
@@ -37,7 +37,7 @@ schema changes.
 | # | Action                       | Operation                                                   |
 |---|------------------------------|-------------------------------------------------------------|
 | 1 | Create                       | `POST /v1/api/workstreams/new`                              |
-| 2 | Subscribe to events          | `GET /v1/api/workstreams/{ws_id}/events` (SSE)              |
+| 2 | Bootstrap history + subscribe | `GET .../history`, then `GET .../events` (SSE)             |
 | 3 | Send a user message          | `POST /v1/api/workstreams/{ws_id}/send`                     |
 | 4 | Inspect children             | `GET /v1/api/workstreams/{ws_id}/children`                  |
 | 5 | Inspect one workstream       | `GET /v1/api/cluster/ws/{ws_id}/detail`                     |
@@ -91,14 +91,34 @@ subscribers (step 2) see the session warm up as token traffic starts.

 ---

-## 2. Subscribe to the per-coordinator event stream
+## 2. Bootstrap history, then subscribe to the event stream
+
+Read and render history before opening the initial stream:

 ```http
-GET /v1/api/workstreams/{ws_id}/events HTTP/1.1
+GET /v1/api/workstreams/{ws_id}/history?limit=100 HTTP/1.1
+Authorization: Bearer <token>
+```
+
+For a loaded coordinator, `messages` is the requested tail of one total
+accepted conversation-row prefix: user, assistant, tool, and system rows,
+including projected compaction checkpoints and cancellation-generated markers.
+The response's optional `cursor` and `handoff_token` belong to that exact
+render. Pass both once on the initial stream URL:
+
+```http
+GET /v1/api/workstreams/{ws_id}/events?last_event_id={cursor}&history_token={handoff_token}&user_turn=1&tool_turn=1 HTTP/1.1
 Accept: text/event-stream
 Authorization: Bearer <token>
 ```

+Omit either query parameter when its history field is `null`. A handoff token
+is opaque and process-local: do not parse, persist, or reuse it. Admission of a
+later conversation row changes the token; durable acknowledgement does not. If
+history returns `503 {"error":"History temporarily unavailable"}`, the response
+is not authoritative: retain the current transcript, do not open a tokenless
+replacement stream, and retry the read.
+
 One persistent SSE connection per browser tab / SDK caller — the
 console fans each event out to every listener queue (cap 500 events
 per queue, put_nowait drop on overflow).  Events come in flat JSON
@@ -110,10 +130,10 @@ with a `type` field.  The recurring shapes a UI has to handle:
 | `reasoning`         | Reasoning-token stream chunk (when the model exposes it)                                   | `text` |
 | `content`           | Assistant-content stream chunk                                                             | `text` |
 | `stream_end`        | End of a single provider stream                                                            | — |
-| `tool_result`       | A tool call completed (success or error)                                                   | `call_id`, `name`, `output`, `is_error?` |
+| `tool_result`       | A tool call completed; capable panes also receive the accepted-history replacement          | `call_id`, `name`, `output`, `is_error?`, `accepted?`, `_event_id?`, `preview?`, `effect_status?` |
 | `tool_output_chunk` | Streaming tool output (e.g. long bash command)                                             | `call_id`, `chunk` |
-| `approve_request`   | One or more tool calls need operator approval                                              | `items: [{call_id, header, preview, func_name, approval_label, needs_approval}]` |
-| `approval_resolved` | Operator answered the approval prompt                                                      | `approved`, `feedback` |
+| `approve_request`   | One approval cycle needs operator action; several cycles may coexist                       | `cycle_id`, `items: [{call_id, header, preview, func_name, approval_label, needs_approval}]` |
+| `approval_resolved` | One identified approval cycle was answered                                                 | `cycle_id`, `call_ids`, `approved`, `feedback`, `always` |
 | `state_change`      | Worker-thread state transition (also re-emitted with the current state on every fresh subscribe so refresh-mid-stream restores composer mode) | `state` ∈ `running`, `thinking`, `attention`, `idle`, `error` |
 | `in_progress_snapshot` | One-shot replay of the in-progress turn's content + reasoning when this client connects mid-stream                              | `content`, `reasoning` |
 | `status`            | Token usage + context-window snapshot (fires on every streaming tick)                      | `prompt_tokens`, `completion_tokens`, `total_tokens`, `context_window`, `pct`, `effort`, `cache_creation_tokens`, `cache_read_tokens` |
@@ -127,10 +147,11 @@ with a `type` field.  The recurring shapes a UI has to handle:
 | `wait_started` / `wait_progress` / `wait_ended` | `wait_for_workstream` tool lifecycle (see §6)                  | `call_id`, `ws_ids`, `elapsed`, `results`, `complete` |
 | `batch_started` / `batch_ended` | `spawn_batch` / `close_all_children` tool lifecycle                            | `call_id`, `op`, `total`/`succeeded`/`denied`/`closed`/`failed`/`skipped` |
 | `info` / `error`    | Operational messages                                                                       | `message` |
+| `history_resync`    | The rendered history token no longer names the accepted row prefix                         | `ws_id`, `reason` |

 **Reconnection contract:** a freshly-opened SSE connection receives
-the current snapshot of any pending tool approval (`approve_request`
-is re-sent if unresolved), any in-flight `wait_*` / `batch_*`
+one `approve_request` snapshot for every unresolved approval cycle, keyed by
+the same stable `cycle_id`, plus any in-flight `wait_*` / `batch_*`
 indicator, the worker's current `state_change`, and an
 `in_progress_snapshot` carrying any partial content / reasoning the
 model has produced for the in-progress turn — so a tab refresh
@@ -138,6 +159,11 @@ mid-approval, mid-tool-execution, or mid-stream restores both the
 correct composer mode and the partial assistant text without waiting
 for the response to complete.

+`history_resync` is stronger than a numeric replay gap. The server closes that
+stream; fetch and render `/history` again, then open a new stream with its new
+cursor/token pair. The API and SDK expose these primitives but deliberately do
+not choose a reconnect policy for callers.
+
 ---

 ## 3. Send the first user message
@@ -324,24 +350,35 @@ uses the cascade-mutation shape and how it differs from the

 The `approve` endpoint is what resolves an `approve_request` SSE
 event.  The coordinator's worker thread is blocked inside
-`ui.approve_tools` waiting for this POST.
+`ui.approve_tools` waiting for this POST. Parallel task agents can leave
+several approval cycles live at once, so current clients echo the event's
+`cycle_id` (or a member `call_id`). A selector-less request resolves the oldest
+cycle for compatibility.

 ```http
 POST /v1/api/workstreams/{ws_id}/approve
-{"approved": true, "feedback": null, "always": false}
+{"approved": true, "feedback": null, "always": false, "cycle_id": "cycle_789"}
 {"approved": false, "feedback": "spawn count looks too high — try 3 not 10"}
-{"approved": true, "feedback": null, "always": true}    // always-approve this tool name
+{"approved": true, "feedback": null, "always": true}    // remember this cycle's tool names
 ```

-`cancel` drops the coordinator's in-flight generation and, for a
-coordinator, auto-cascades the cancel to its direct children:
+Success returns `{"status": "ok", "cycle_id": "cycle_789"}`. A stale selector
+returns `409` with the currently oldest cycle/call IDs. `always` remembers only
+the tool names in the cycle that actually resolved; it does not enable blanket
+approval.
+
+`cancel` requests cooperative cancellation of the coordinator's in-flight
+generation and auto-cascades to its direct children:
 `cancel_workstream` is dispatched through the routing proxy for
-every direct child in the registry.  The coordinator itself is left
-idle and open for a fresh `send`:
+every direct child in the registry. The HTTP acknowledgement is immediate;
+the worker becomes idle after unwinding. Pass `{"force": true}` only to release
+a wedged worker slot immediately. The coordinator itself remains open for a
+fresh `send`:

 ```http
 POST /v1/api/workstreams/{ws_id}/cancel
 {}
+{"status": "ok", "dropped": {}}
 ```

 ---
@@ -361,6 +398,11 @@ disconnect.  The row is reopenable via
 `POST /v1/api/workstreams/{ws_id}/open` so long as it hasn't been
 deleted.

+If any accepted live conversation row still requires persistence
+reconciliation, close returns `409 {"error":"workstream has unresolved
+persistence"}`. The coordinator remains loaded, its journal is retained, and
+no history is discarded; retry after storage recovers.
+
 ---

 ## Further reading
@@ -77,7 +77,7 @@ or MCP config can do adds to it.  Current members:
 | `delete_workstream`       | wind-down       | Hard-delete one child.  Requires approval.                          |
 | `list_nodes`              | discover        | Enumerate live cluster nodes + capabilities.                        |
 | `skills` (action=find)    | discover        | Browse the skill catalog; opt-in `kind` filter narrows by audience. |
-| `memory`                  | persist         | Durable orchestration memory (`coordinator` scope, per-user — survives across coordinator sessions). |
+| `memory`                  | persist         | Durable acting-user orchestration memory (`coordinator`), plus shared memory when attached to a project. |
 | `notify`                  | broadcast       | Post a status update to a human channel at a narrative beat.        |
 | `tasks`                   | plan            | Orchestrator-only scratchpad.  Children don't see it.               |

@@ -126,10 +126,11 @@ the skill should end on.

 `tasks` is the coordinator's scratchpad — a persisted, ordered
 list of rows with fields `{id, title, status, child_ws_id, created,
-updated}` that only this coordinator sees.  Children don't see it;
-the user does via the sidebar.  Five actions: `add`, `update`,
-`remove`, `reorder`, `list` (only `list` is auto-approved; the
-mutators go through the approval flow).
+updated}`, plus `note` on rows where one has been set (the key is
+absent otherwise), that only this coordinator sees.  Children don't
+see it; the user does via the sidebar.  Five actions: `add`,
+`update`, `remove`, `reorder`, `list` (only `list` is auto-approved;
+the mutators go through the approval flow).

 The input schema refers to rows by `task_id`; the persisted row
 object exposes the same id as `id`.  The `child_ws_id` field is a
@@ -142,11 +143,20 @@ A skill's initial prompt can seed the task list by calling
 `tasks(action="add", title=...)` as its very first tool calls —
 the user gets a visible plan before any child is spawned, and the
 coordinator's future self has something concrete to iterate on.
-Status transitions (`pending` → `in_progress` → `done` / `blocked`)
-are the skill's main feedback loop: mutate the task when the child
-covering it finishes, not when the child starts.  Use
-`tasks(action="update", task_id=..., child_ws_id=<ws_id>)` to
-link a task to the child that owns it once spawn returns.
+Status transitions (`pending` → `in_progress` → `done` / `blocked` /
+`needs_user`) are the skill's main feedback loop: mutate the task
+when the child covering it finishes, not when the child starts.
+`blocked` and `needs_user` are not interchangeable — `blocked` is a
+dependency the coordinator may be able to clear itself, while
+`needs_user` marks a task that cannot move without a decision,
+approval, or grant only the user can give.  The distinction is
+load-bearing: a coordinator that goes idle holding open tasks gets
+nudged to pick them back up — even when children are still running, so
+keep the matrix honest rather than expecting the reminder to wait for
+an all-clear — and `needs_user` is what tells that nudge the stop was
+deliberate.  Pair it with `note` to record what is being asked for.
+Use `tasks(action="update", task_id=..., child_ws_id=<ws_id>)` to link
+a task to the child that owns it once spawn returns.

 A final gotcha: parallel tool dispatch does NOT serialise reads
 after writes in the same batch.  If a skill issues an `update` and
@@ -297,11 +307,11 @@ and the coordinator's planning step is itself valuable.
 tasks(action='add', title='...') × N   # the plan, visible in the sidebar
 for task in tasks:
    spawn_workstream(skill=..., initial_message=task.brief)
-    tasks(action='update', task_id=task.id, notes='ws=<child_ws_id>')
+    tasks(action='update', task_id=task.id, note='ws=<child_ws_id>')
 wait_for_workstream(ws_ids=[...], mode='all', timeout=...)
 for child in children:
    inspect_workstream(ws_id=child)
-    tasks(action='update', task_id=..., status='done', notes='result summary')
+    tasks(action='update', task_id=..., status='done', note='result summary')
 → synthesise
 ```

@@ -13,7 +13,7 @@ cloud "LLM Providers" as llm {
    component [OpenAI-compatible API\n(OpenAI, vLLM, llama.cpp)] as llm_openai
    component [Anthropic Messages API] as llm_anthropic
 }
-database "SQLite\n(.turnstone.db)" as sqlite
+database "SQLite / PostgreSQL\n(durable state)" as storage

 ' Turnstone System Boundary
 package "Turnstone Platform" {
@@ -33,24 +33,26 @@ eval_user --> eval : Python API

 ' Internal connections
 cli --> llm : LLM Provider API\n(via provider adapters)
-cli --> sqlite : SQLite
+cli --> storage : persistence

 server --> llm : LLM Provider API\n(via provider adapters)
-server --> sqlite : SQLite
+server --> storage : persistence

 eval --> llm : LLM Provider API\n(non-streaming)
-eval --> sqlite : SQLite
+eval --> storage : persistence

-console --> server : HTTP proxy\n(hash-ring bucket lookup,\nproxy /node/{id}/* traffic)
+console --> server : HTTP routing/UI proxy + cluster SSE\n(FNV-1a rendezvous placement,\nproxy /node/{id}/* traffic)

-channel --> server : HTTP + SSE\n(POST /v1/api/workstreams/{ws_id}/send,\nGET /v1/api/workstreams/{ws_id}/events)
+channel --> console : multi-node route/create/live/send/approve
+channel --> server : direct mode + owning-node SSE\n(POST /v1/api/workstreams/{ws_id}/send,\nGET /v1/api/workstreams/{ws_id}/events)

 ' Notes
 note right of console
  Multi-node router:
-  - Hash-ring bucket lookup
+  - FNV-1a rendezvous placement
  - Proxies create/send/approve
-  - Direct SSE from client to node
-  - HTTP polling for dashboard
+  - Collector aggregates node SSE
+  - Browser dashboard receives console SSE fanout
+  - /node/{id} proxies pane HTTP + SSE
 end note
@enduml
@@ -18,6 +18,7 @@ skinparam component {
 package "Entry Points" <<Rectangle>> {
    component [cli.py\nturnstone] as cli <<entry>>
    component [server.py\nturnstone-server] as server <<entry>>
+    component [console/server.py\nturnstone-console] as consoleentry <<entry>>
    component [eval.py\nturnstone-eval] as eval <<entry>>
    component [admin.py\nturnstone-admin] as admin <<entry>>
    component [bootstrap.py\nturnstone-bootstrap] as bootstrap <<entry>>
@@ -25,9 +26,16 @@ package "Entry Points" <<Rectangle>> {

 ' Core engine
 package "turnstone/core/" <<Rectangle>> {
-    component [session.py\nChatSession, SessionUI] as session <<core>>
+    component [session.py\nChatSession, SessionUI\ngeneration-fenced turn loop] as session <<core>>
+    component [session_manager.py\nSessionManager\nshared lifecycle invariants] as sessionmanager <<core>>
+    component [adapters/\ninteractive + coordinator\nconstruction/event policies] as adapters <<core>>
+    component [model_turn.py\nModelLane, model_turn()\nlower / sample / re-ingest] as modelturn <<core>>
+    component [trajectory.py\ncanonical Turn IR] as trajectory <<core>>
+    component [lowering.py\nprovider-wire lowering] as lowering <<core>>
+    component [state_writer.py\nordered durable state tail] as statewriter <<core>>
+    component [model_backend_auth.py\nper-call backend credentials] as modelauth <<core>>
    component [providers/\nLLMProvider, OpenAI, Anthropic, Google] as providers <<core>>
-    component [workstream.py\nWorkstreamManager] as workstream <<core>>
+    component [workstream.py\nWorkstream types + state] as workstream <<core>>
    component [tools.py\nTool loader] as tools <<core>>
    component [memory.py\nPersistence facade] as memory <<core>>
    component [storage/\nStorageBackend protocol\nSQLite + PostgreSQL] as storage <<core>>
@@ -79,18 +87,18 @@ package "turnstone/api/" <<Rectangle>> {
 package "turnstone/sdk/" <<Rectangle>> {
    component [server.py\nTurnstoneServer (sync+async)] as sdkserver <<sdk>>
    component [console.py\nTurnstoneConsole (sync+async)] as sdkconsole <<sdk>>
-    component [events.py\n27 SSE event types] as sdkevents <<sdk>>
+    component [events.py\nTyped SSE event stream] as sdkevents <<sdk>>
    component [_base.py\nhttpx client base] as sdkbase <<sdk>>
 }

 ' Tool schemas
 package "turnstone/tools/" <<Rectangle>> {
-    component [*.json\n19 tool schemas] as schemas <<artifact>>
+    component [*.json\nBuilt-in tool schemas] as schemas <<artifact>>
 }

 ' Entry point dependencies
 cli --> session
-cli --> workstream
+cli --> sessionmanager
 cli --> config
 cli --> memory
 cli --> colors
@@ -99,7 +107,8 @@ cli --> spinner
 cli --> tools

 server --> session
-server --> workstream
+server --> sessionmanager
+server --> adapters
 server --> config
 server --> memory
 server --> metrics
@@ -113,11 +122,26 @@ eval --> memory
 eval --> config
 eval --> tools

+consoleentry --> sessionmanager
+consoleentry --> adapters
+consoleentry --> consoleserver
+
 admin --> auth
 bootstrap --> providers

 ' Core internal deps
-session --> providers
+sessionmanager --> workstream
+sessionmanager --> adapters
+sessionmanager --> storage
+adapters --> session : constructs
+session --> modelturn
+session --> trajectory
+session --> lowering
+session --> statewriter
+session --> modelauth
+modelturn --> providers
+modelturn --> trajectory
+modelturn --> lowering
 session --> tools
 session --> memory
 memory --> storage
@@ -129,6 +153,7 @@ session --> mcp : optional
 session --> toolsearch : optional
 session --> registry : optional
 registry --> providers
+modelturn --> registry : coherent snapshot
 healthcheck --> metrics
 mcp --> config
 registry --> config
@@ -138,15 +163,17 @@ tools --> schemas
 gateway --> discordbot
 gateway --> slackbot
 gateway --> router
-discordbot --> sdkserver : HTTP + SSE
-slackbot --> sdkserver : HTTP + SSE
+discordbot --> sdkserver : direct HTTP + node SSE
+slackbot --> sdkserver : direct HTTP + node SSE
+router --> sdkserver : single-node/direct mode
+router --> sdkconsole : multi-node route/create/live
 router --> storage : channel_routes

 ' Console dependencies
 consoleserver --> collector
 consoleserver --> config
 consoleserver --> auth
-collector --> server : HTTP polling
+collector --> server : discovery HTTP + cluster SSE aggregation

 ' API dependencies
 serverspec --> openapi
@@ -32,7 +32,7 @@ class "TerminalUI" as TerminalUI {
 class "WorkstreamTerminalUI" as WsTermUI {
    - _output_buffer: list[tuple]
    - ws_id: str
-    - manager: WorkstreamManager
+    - manager: SessionManager
    + flush_buffer()
    --
    Buffers output when workstream
@@ -41,14 +41,14 @@ class "WorkstreamTerminalUI" as WsTermUI {

 class "WebUI" as WebUI {
    - _listeners: list[Queue]
-    - _approval_event: Event
+    - _approval_cycles: dict[str, ApprovalCycle]
    - _ws_prompt_tokens: int
    - _ws_tool_calls: dict
-    + resolve_approval(approved, feedback)
+    + resolve_approval(approved, feedback, cycle_id?, call_id?)
    --
    Enqueues JSON events for SSE.
-    Blocks on threading.Event for
-    approval.
+    Concurrent approval cycles each own
+    a threading.Event and result slot.
    SSE handlers bridge Queue to
    async via run_in_executor().
    --
@@ -126,31 +126,82 @@ class "ModelCapabilities" as ModelCaps <<frozen>> {
    + supports_reasoning_replay: bool
 }

+class "ModelLane" as ModelLane <<frozen>> {
+    + provider: LLMProvider
+    + client: Any
+    + model: str
+    + alias: str
+    + capabilities: ModelCapabilities
+    + extra_params: dict | None
+    + registry: ModelRegistry | None
+    + admission: ModelAdmission | None
+    + backend_auth_config: ModelConfig | None
+    + backend_auth_resolver: Callable | None
+}
+
+class "ResolvedModelBinding" as ResolvedBinding <<frozen>> {
+    + lane: ModelLane
+    + config: ModelConfig | None
+    + registry_generation: int
+}
+
+class "ModelTurnResult" as ModelTurnResult <<frozen>> {
+    + turn: Turn
+    + tool_calls: list[dict]
+    + finish_reason: str
+    + usage: UsageInfo | None
+    + wire_msgs: list[dict] | None
+    + producer: str
+    + serving_model: str
+}
+
+class "model_turn()" as ModelTurnFn {
+    Turn IR → lower → provider stream
+    → drain → canonical assistant Turn
+    --
+    core/model_turn.py
+}
+
+class "Backend auth resolver" as BackendAuth {
+    + resolve_model_backend_auth_token(...)
+    --
+    Resolves static / Entra OBO /
+    Entra app / RFC 8693 per call.
+    Dynamic failure can fail closed.
+    --
+    core/model_backend_auth.py
+}
+
 ' ChatSession
 class "ChatSession" as ChatSession {
-    - client: Any
-    - provider: LLMProvider
-    - model: str
+    - _model_binding: ResolvedModelBinding
+    - _model_binding_lock: Lock
    - ui: SessionUI
-    - messages: list[dict]
+    - messages: list[Turn]
    - _msg_tokens: list[int]
    - _ws_id: str
    - _mcp_client: MCPClientManager | None
    - _tool_search: ToolSearchManager | None
    - _registry: ModelRegistry | None
+    - _generation: int
+    - _cancel_event: Event
+    - _durability_next_ticket: int
    + model_alias: str | None {property}
    - _tools: list[dict]
    - _task_tools: list[dict]
    - _read_files: set[str]
    - system_messages: list[dict]
    --
-    + send(user_input: str)
+    + send(user_input: str, ..., acting_user_id: str | None)
+    + cancel()
+    + compact_now() → bool
+    + fork_from_storage(source_ws_id, principal_id, ...)
    + handle_command(command: str)
    + resume(ws_id: str)
    - _save_config()
-    - _stream_response(stream) → dict
-    - _create_stream_with_retry(msgs) → Stream (+ fallback)
-    - _try_stream(client, model, msgs) → Stream
+    - _stream_response(my_generation) → ModelTurnResult
+    - _model_turn_with_fallback(consumer, prepare_wire) → ModelTurnResult
+    - _model_turn_with_retry(lane, tracker, ...) → ModelTurnResult
    - _execute_tools(tool_calls) → (results, feedback)
    - _prepare_tool(tc) → item dict
    - _prepare_mcp_tool(call_id, name, args) → item dict
@@ -162,8 +213,10 @@ class "ChatSession" as ChatSession {
    - _rebuild_tool_search()
    + close()
    - _run_agent(messages, tools, ...) → str
-    - _compact_messages(auto: bool)
-    - _full_messages() → list[dict]
+    - _compact_messages(auto: bool, my_generation: int)
+    - _commit_for_generation(generation, commit)
+    - _publish_for_generation(generation, publish)
+    - _full_messages() → list[Turn]
    - _update_token_table(msg)
    - _emit_state(state: str)
    - _generate_title()
@@ -180,15 +233,36 @@ class "HeadlessSession" as HeadlessSession {
    records all tool calls
 }

-' WorkstreamManager
-class "WorkstreamManager" as WsMgr {
-    - _session_factory: Callable[[SessionUI], ChatSession]
+' SessionManager
+interface "SessionKindAdapter" as KindAdapter <<Protocol>> {
+    + kind: WorkstreamKind
+    + build_ui(ws) → SessionUI
+    + build_session(ws, ...) → ChatSession
+    + cleanup_ui(ws)
+}
+
+interface "SessionEventEmitter" as EventEmitter <<Protocol>> {
+    + emit_created(ws)
+    + emit_rehydrated(ws)
+    + emit_state(ws, state)
+    + emit_closed(ws_id, reason, name)
+}
+
+class "SessionManager" as SessionMgr {
+    - _adapter: SessionKindAdapter
+    - _storage: StorageBackend
    - _workstreams: dict[str, Workstream]
+    - _pending_creates: dict[str, Workstream]
+    - _retiring_ids: set[str]
+    - _state_writer: StateWriter | None
    - _order: list[str]
    - _active_id: str
-    - _on_state_change: Callable
    --
-    + create(name, ui_factory) → Workstream
+    + create(user_id, name, ..., defer_emit_created) → Workstream
+    + commit_create(ws) → bool
+    + discard(ws, ...) → bool
+    + open(ws_id) → Workstream | None
+    + delete(ws_id) → bool
    + close(ws_id)
    + get(ws_id) → Workstream
    + get_active() → Workstream
@@ -203,11 +277,17 @@ class "Workstream" as Ws <<dataclass>> {
    + id: str
    + name: str
    + state: WorkstreamState
-    + session: ChatSession
-    + ui: SessionUI
-    + worker_thread: Thread
+    + session: ChatSession | None
+    + ui: SessionUI | None
+    + worker_thread: Thread | None
    + error_message: str
    + last_active: float
+    + kind: WorkstreamKind
+    + user_id: str
+    + parent_ws_id: str | None
+    + project_id: str | None
+    - _fork_reservation_token: str
+    - _closed: bool
    - _lock: Lock
 }

@@ -278,12 +358,13 @@ class "ModelRegistry" as ModelReg {
    - _models: dict[str, ModelConfig]
    - _clients: dict[str, Any]
    - _providers: dict[str, LLMProvider]
+    - _admissions: dict[str, ModelAdmission]
    - _client_lock: Lock
    + default: str
    + fallback: list[str]
    + agent_model: str | None
    --
-    + resolve(alias) → (client, model, config)
+    + resolve_binding(alias) → (client, model, config, provider, admission, generation)
    + get_client(alias) → Any
    + get_provider(alias) → LLMProvider
    + has_alias(alias) → bool
@@ -297,6 +378,22 @@ class "ModelRegistry" as ModelReg {
    core/model_registry.py
 }

+class "ModelAdmission" as ModelAdmission {
+    - alias: str
+    - _limit: int
+    - _in_flight: int
+    - _waiters: deque
+    + acquire(cancel_ref) → AdmissionLease
+    + set_limit(limit)
+    + snapshot() → AdmissionSnapshot
+    --
+    Per-process FIFO generation gate.
+    Stable across alias hot reloads;
+    queue time is deadline credit.
+    --
+    core/admission.py
+}
+
 class "ModelConfig" as ModelCfg <<frozen>> {
    + alias: str
    + provider: str
@@ -306,6 +403,10 @@ class "ModelConfig" as ModelCfg <<frozen>> {
    + temperature: float | None
    + max_tokens: int | None
    + reasoning_effort: str | None
+    + max_concurrency: int
+    + auth_mode: str
+    + obo_audience: str
+    + obo_scopes: str
 }

 ' Circuit breaker state
@@ -375,22 +476,35 @@ LLMProvider <|.. AnthropicProv
 OpenAIProv <|-- GoogleProv

 ChatSession --> SessionUI : uses
-ChatSession --> LLMProvider : delegates LLM calls
+ChatSession --> ResolvedBinding : owns coherent snapshot
+ChatSession --> ModelTurnFn : every model-backed role
 ChatSession --> MCPMgr : optional
 ChatSession --o ToolSearchMgr : _tool_search
 ChatSession --> ModelReg : optional
 ChatSession <|-- HeadlessSession

-WsMgr --> "*" Ws : manages
+SessionMgr --> "*" Ws : manages
+SessionMgr --> KindAdapter : delegates construction
+SessionMgr --> EventEmitter : lifecycle fan-out
 Ws --> "1" ChatSession : wraps
 Ws --> "1" SessionUI : wraps
 Ws --> "1" WsState : has

-WsMgr ..> ChatSession : creates via\nsession_factory(ui, model_alias)
+KindAdapter ..> ChatSession : constructs

 ModelReg --> "*" ModelCfg : holds
 ModelReg --> "*" LLMProvider : caches
+ModelReg --> "*" ModelAdmission : owns per alias
 LLMProvider --> ModelCaps : returns
+ModelReg --> ResolvedBinding : resolves atomically
+ResolvedBinding --> ModelLane
+ModelLane --> LLMProvider
+ModelLane --> ModelCaps
+ModelLane --> ModelCfg : auth/config snapshot
+ModelLane --> ModelAdmission : admission lease
+ModelTurnFn --> ModelLane
+ModelTurnFn --> ModelTurnResult
+ModelTurnFn ..> BackendAuth : per-call resolver

 ChatSession --> HealthMon : checks circuit
 HealthMon --> "1" CircuitState : has
@@ -403,7 +517,9 @@ note bottom of ChatSession
  Provider-agnostic — delegates all LLM
  communication to LLMProvider adapters.

-  core/session.py (~2700 lines)
+  Every live/durable publication is fenced by
+  its generation. Model calls use immutable lanes;
+  provider-wire mutation stays at lowering.
 end note

@enduml
@@ -1,183 +1,171 @@
@startuml
 !theme plain
-title Turnstone — Conversation Turn Lifecycle
+title Turnstone — Generation-Fenced Conversation Turn

 skinparam sequenceArrowThickness 1.5
 skinparam sequenceLifeLineBackgroundColor #F5F5F5

-participant "User /\nHTTP Client" as User
-participant "ChatSession" as CS
-participant "SessionUI" as UI
-participant "LLMProvider\n(OpenAI / Anthropic)" as LLM
-participant "Tool Executor\n(ThreadPool)" as TP
-database "SQLite" as DB
+participant "HTTP / CLI\ncaller" as User
+participant "SessionManager" as Manager
+participant "ChatSession" as Session
+participant "SessionUIBase" as UI
+participant "Accepted-row handoff\n(total live prefix)" as Handoff
+participant "model_turn()\n+ lowering" as Plant
+participant "ModelAdmission\n(per alias)" as Admission
+participant "LLM provider" as Provider
+participant "Tool workers" as Tools
+database "StorageBackend\n(SQLite / PostgreSQL)" as Storage

-== User Input ==
+== Admission and generation claim ==

-User -> CS : send(user_input)
-activate CS
-
-CS -> CS : messages.append({role: "user", content: input})
-CS -> DB : save_message(ws_id, "user", input)
-
-== LLM Call Loop ==
-
-group loop [while tool_calls present]
-
-    CS -> UI : on_turn_start()
-    note right of UI
-      SessionUIBase resets the per-turn inflight
-      buffers (_ws_inflight_content / reasoning /
-      seq) that fuel the SSE in_progress_snapshot
-      event for mid-stream refresh resume.
-    end note
-
-    CS -> UI : on_state_change("thinking")
-    CS -> UI : on_thinking_start()
-
-    CS -> LLM : provider.create_streaming(\n  client, model, messages, tools, ...)\n  (normalized StreamChunk iterator)
-    activate LLM
-
-    note right of CS
-      Retry up to 3× on transient errors:
-      RateLimitError, APITimeoutError,
-      APIConnectionError, InternalServerError,
-      ServiceUnavailableError, APIError
-      Backoff: 1s, 2s, 4s
-    end note
-
-    == Streaming Response ==
-
-    loop for each chunk in stream
-        LLM --> CS : delta
-        note right of CS
-          on_thinking_stop() called on first
-          delta token via _stop_spinner_once()
-        end note
-        alt reasoning_content present
-            CS -> UI : on_reasoning_token(text)
-        else content present
-            CS -> UI : on_content_token(text)
-        else tool_call delta
-            CS -> CS : accumulate in tool_calls_acc
-        else info_delta present
-            CS -> UI : on_info(text)\n(e.g. server-side web search status)
-        end
-    end
-
-        note right of CS
-          **Cancellation checkpoint:**
-          _check_cancelled() runs per chunk.
-          If cancel_event is set, raises
-          GenerationCancelled — preserves
-          partial content, emits idle state.
-        end note
-
-    LLM --> CS : stream complete (usage stats)
-    deactivate LLM
-
-    CS -> UI : on_thinking_stop()  (no-op guard: already called by _stop_spinner_once)
-    CS -> UI : on_stream_end()
-
-    CS -> CS : _update_token_table()\ncalibrate chars_per_token ratio
-    CS -> CS : messages.append(assistant_msg)
-    CS -> UI : on_turn_committed()
-    note right of UI
-      Drops the per-turn inflight buffers — the
-      assistant message is now in the history
-      list, so the in_progress_snapshot must
-      not re-render it during the next tool-
-      execution window or the next streaming turn.
-    end note
-    CS -> DB : save_message(ws_id, "assistant", content)
-    CS -> DB : save_message(ws_id, "tool_call", ...) ×N
-
-    == Tool Dispatch (if tool_calls) ==
-
-    alt no tool_calls
-        CS -> UI : on_status(usage, context_window, effort)
-
-        opt prompt_tokens > context_window × auto_compact_pct
-            CS -> CS : _compact_messages(auto=True)
-            CS -> LLM : Non-streaming summarization call
-            CS -> CS : Replace messages with [summary]
-        end
-
-        opt first exchange & no title
-            CS -> CS : Background thread: _generate_title()
-        end
-
-        CS -> UI : on_state_change("idle")
-        CS --> User : return
-
-    else has tool_calls
-        CS -> UI : on_state_change("running")
-
-        == Phase 1: Prepare ==
-        CS -> CS : [_prepare_tool(tc) for tc in tool_calls]\nParse JSON args, validate,\nbuild preview + header
-
-        == Phase 2: Approve ==
-        CS -> UI : on_state_change("attention")
-        CS -> UI : approve_tools(items)
-        activate UI
-        note right of UI
-          TerminalUI: input() prompt
-          WebUI: _approval_event.wait()
-          NullUI: returns (True, None)
-        end note
-        UI --> CS : (approved: bool, feedback: str?)
-        deactivate UI
-        CS -> UI : on_state_change("running")
-
-        == Phase 3: Execute ==
-        CS -> TP : ThreadPoolExecutor(max_workers=4)\nrun_one(item) for each tool
-        activate TP
-
-        note right of TP
-          Parallel execution:
-          bash → Popen + line-by-line streaming
-          read_file → open().read() or base64 image
-          search → grep subprocess
-          edit_file → string replace
-          task → _run_agent() sub-loop
-          web_fetch → httpx + LLM summarize
-          web_search → provider-native or SearxNG fallback
-          memory/recall → SQLite
-        end note
-
-        note right of TP
-          bash: on_tool_output_chunk(call_id, line)
-          called per stdout line,
-          then on_tool_result(call_id, name, output, is_error).
-          is_error=True when execution failed.
-          call_id routes chunks/results to correct
-          tool div during parallel execution.
-          Other tools: on_tool_result() only.
-        end note
-
-        TP --> CS : [(call_id, output), ...]
-        deactivate TP
-
-        loop for each result
-            CS -> CS : messages.append({role: "tool", ...})
-            CS -> DB : save_message(ws_id, "tool_result", ...)
-        end
-
-        opt user_feedback from approval
-            CS -> CS : messages.append({role: "user", content: feedback})
-        end
-
-        note right of CS : Loop back for next LLM call
-
-    else GenerationCancelled
-        CS -> CS : Preserve partial content\nor roll back incomplete tools
-        CS -> UI : on_info("[Generation cancelled]")
-        CS -> UI : on_state_change("idle")
-        CS --> User : return (no re-raise)
-    end
+User -> Manager : dispatch send on one Workstream
+Manager -> Session : bind_acting_user(principal)\nsend(text, attachments, send_id)
+activate Session
+Session -> Session : refresh immutable ResolvedModelBinding

+opt token budget exhausted
+  Session -> UI : approve_tools(__budget_override__)
+  note right of UI
+    This gate precedes a generation claim but carries
+    a monotonic cancellation witness. Stop cannot be
+    mistaken for a budget-policy denial.
+  end note
 end

-deactivate CS
+Session -> Session : _claim_generation() → generation N\ninstall fresh cancel event
+Session -> Session : plan memory / participant context
+Session -> Handoff : admit USER row\ncommit_key + prefix revision
+Session -> Storage : ordered durable batch:\nappend canonical user Turn + metadata

+note over Session, Handoff
+  Every accepted conversation row enters this lane before durability:
+  USER, ASSISTANT, TOOL, SYSTEM, compaction checkpoints, and cancellation
+  markers. Admission shares the handoff lock with its live UI transition
+  or history_resync repair event.
+end note
+
+note over Handoff, Storage
+  _commit_for_generation(N) admits bounded live mutations under the
+  generation lock, then executes immutable persistence closures in FIFO
+  ticket order. A force successor either follows the whole commit or
+  prevents it. /history projects durable prefix + pending journal suffix;
+  durable ACK removes the pending copy without changing the prefix revision.
+end note
+
+opt already over the hard context ceiling
+  Session -> Session : compact before first model call\n(preserve the new user turn)
+end
+
+== Model / tool loop ==
+
+loop until final answer and no queued input
+  Session -> UI : on_turn_start()\nreset per-stream replay buffers
+  Session -> UI : state = thinking\non_thinking_start()
+  Session -> Session : _stream_response(N)\nretry + fallback policy
+  Session -> Plant : model_turn(active ModelLane, Turns,\n tools, cancel_ref, on_chunk)
+  activate Plant
+  Plant -> Plant : canonical Turns → provider wire\nrestore ids + repair + lane-specific fold
+  Plant -> Plant : materialize attachment refs\n(nested perception before outer slot)
+  Plant -> Admission : acquire(cancel_ref)
+  activate Admission
+  Plant -> Plant : resolve per-call backend credential\nfrom lane's pinned ModelConfig
+  Plant -> Provider : create_streaming(...)
+  activate Provider
+
+  loop normalized stream chunks
+    Provider --> Plant : StreamChunk
+    Plant --> Session : on_chunk(StreamChunk)
+    Session -> Session : check cancel event + generation N
+    Session -> UI : reasoning / content / info token
+  end
+
+  Provider --> Plant : finish + usage + native blocks
+  deactivate Provider
+  Plant -> Plant : drain + re-ingest assistant Turn\nwith serving-lane provenance
+  Plant -> Admission : release before retry backoff
+  deactivate Admission
+  Plant --> Session : ModelTurnResult
+  deactivate Plant
+
+  Session -> UI : on_stream_end()
+  Session -> Session : generation-fenced result commit:\nappend assistant Turn + token accounting
+  Session -> UI : on_turn_committed()
+  Session -> Handoff : admit ASSISTANT row\ncommit_key + prefix revision
+  Session -> Storage : ordered durable assistant row\n(content + tool mirror + native lane)
+
+  alt no tool calls
+    opt over soft threshold
+      Session -> Session : cooperative / end-of-turn compaction
+      Session -> Handoff : admit SYSTEM/source=compaction\ncheckpoint projection
+      Session -> Storage : append checkpoint summary marker\nwith source watermark
+      note right of Storage
+        Full history remains durable. Resume loads
+        [summary] + rows after the checkpoint.
+      end note
+      opt model stopped for compaction
+        Session -> Handoff : admit USER/source=compaction_resume row
+        Session -> Storage : append synthetic compaction_resume Turn
+      end
+    end
+    alt queued messages drained
+      Session -> Handoff : admit combined queued USER row
+      Session -> Storage : append combined queued user Turn
+    else truly complete
+      Session -> UI : state = idle
+    end
+  else tool calls present
+    Session -> UI : state = running
+    Session -> Session : prepare items + previews\nattach cancellation witnesses
+
+    opt one or more items require a human
+      Session -> UI : approve_tools(items)\nregister independent ApprovalCycle
+      note right of UI
+        Parallel agents may own concurrent cycles.
+        cycle_id / call_id routes exactly one decision;
+        Smart Approvals may clear qualifying items.
+      end note
+      User -> UI : approve / deny selected cycle
+      UI --> Session : decision + optional feedback
+    end
+
+    Session -> Tools : execute admitted items in parallel
+    activate Tools
+    Tools --> UI : chunks + result card\nwith effect disposition
+    Tools --> Session : outputs / errors / effect statuses
+    deactivate Tools
+    Session -> Session : output-guard evaluation\nthen generation N re-check
+
+    opt compaction owed before result sizing
+      Session -> Session : compact, preserving assistant tool-call Turn
+      Session -> Handoff : admit SYSTEM/source=compaction checkpoint
+      Session -> Storage : append checkpoint marker
+    end
+
+    Session -> Session : one generation-fenced batch:\nappend all Tool Turns, advisories, feedback
+    Session -> Handoff : admit FIFO TOOL rows\ncommit keys + prefix revisions
+    Session -> Storage : FIFO durable tool rows + metadata
+  end
+end
+
+== Stop / force-successor boundary ==
+
+User -> Session : cancel()
+Session -> Session : atomically set generation event; snapshot\nmain stream, child scopes, judges, subprocesses
+Session -> Provider : close live stream handle
+Session -> Tools : abort child scopes + kill subprocess groups
+Session -> UI : resolve only cancelled operation's\napproval cycles
+opt cancellation produced accepted conversation rows
+  Session -> Handoff : admit partial ASSISTANT and/or\nsynthesized TOOL cancellation markers
+  Session -> Storage : idempotent keyed cancellation rows
+end
+
+note over Session, Storage
+  Every later publish/commit checks generation ownership. An abandoned
+  worker may unwind, but cannot append Turns, overwrite state, resolve a
+  successor approval, or repaint the successor UI. Observed tool effects
+  are preserved as controller-authored cancellation receipts; unreviewed
+  tool bytes are not laundered into model context.
+end note
+
+deactivate Session
@enduml
@@ -1,134 +1,117 @@
@startuml
 !theme plain
-title Turnstone — Tool Execution Pipeline (Three Phases)
+title Turnstone — Tool Pipeline: Prepare, Approve, Execute, Fold

 start

-partition "Phase 1: Prepare" #E8F5E9 {
-    :Receive tool_calls list from LLM response;
+partition "Phase 1 — Prepare and assess" #E8F5E9 {
+    :Receive tool calls from one assistant Turn;
+    :Capture the generation's cancel event\nand acting principal;

-    while (more tool_calls?) is (yes)
-        :Extract call_id, func_name, raw_args;
-
-        if (json.loads(raw_args) succeeds?) then (yes)
-            :parsed_args = JSON dict;
+    while (more tool calls?) is (yes)
+        :Parse arguments and dispatch to\nthe tool-specific preparer;
+        if (preparation succeeds?) then (yes)
+            :Build item: call_id, name, header, preview,\nneeds_approval, execute closure;
        else (no)
-            :Fallback 1: regex extraction;
-            if (regex found keys?) then (yes)
-                :parsed_args = extracted dict;
-            else (no)
-                :Fallback 2: bare string →\nPRIMARY_KEY_MAP[func_name];
-            endif
+            :Build an error item for this call only;\nkeep sibling calls valid;
        endif
-
-        :Dispatch to _prepare_{func_name}();
-
-        note right
-            **Dispatch table (16 built-in + tool_search):**
-            ┌───────────────┬──────────────────┐
-            │ Tool          │ Needs Approval?  │
-            ├───────────────┼──────────────────┤
-            │ bash          │ ✓ Yes            │
-            │ read_file     │ ✗ Auto-approve   │
-            │ write_file    │ ✓ Yes            │
-            │ edit_file     │ ✓ Yes            │
-            │ search        │ ✗ Auto-approve   │
-            │ diff_file     │ ✗ Auto-approve   │
-            │ web_fetch     │ ✗ Auto-approve   │
-            │ web_search    │ ✗ Auto-approve   │
-            │ tool_search   │ ✗ Auto-approve   │
-            │ task_agent    │ ✓ Yes            │
-            │ memory        │ ✗ Auto-approve   │
-            │ recall        │ ✗ Auto-approve   │
-            │ notify        │ ✗ Auto-approve   │
-            │ watch         │ ✓ create only    │
-            │ skill         │ ✓ load only      │
-            │ read_resource │ ✓ Yes            │
-            │ use_prompt    │ ✓ Yes            │
-            ├───────────────┼──────────────────┤
-            │ mcp__*        │ ✓ Yes (external) │
-            └───────────────┴──────────────────┘
-        end note
-
-        :Build item dict:
-        {call_id, func_name, header,
-         preview, needs_approval,
-         approval_label, execute: Callable};
+        :Attach operation-local cancellation witness\nand pinned principal;
    endwhile (no)
+
+    :Reject only unsafe ordering shapes\n(for example tasks read + write in one batch);
+    :Run heuristic intent assessment immediately;
+    :Start generation-pinned LLM judge in background;
+    :Stamp one immutable Smart Approval\nsettings snapshot on the batch;
+
+    note right
+      Preparation is per-call isolated: one bad preparer
+      becomes one error Tool Turn rather than orphaning the
+      assistant's entire tool-call set.
+    end note
 }

-partition "Phase 2: Approve" #FFF3E0 {
-    if (any items need approval?) then (yes)
-        :_emit_state("attention");
-        :ui.approve_tools(items);
+partition "Phase 2 — Approval cycle" #FFF3E0 {
+    :Apply explicit bypasses:\nskill / always / policy / blanket;
+
+    if (Smart Approvals enabled?) then (yes)
+        :Wait within the batch's bounded judge deadline;
+        :Auto-approve only LLM approve verdicts\nat or above the captured threshold;
+    endif
+
+    if (human-gated items remain?) then (yes)
+        :Acquire approval-publication lease;
+        :Register independent ApprovalCycle\n(cycle_id, call_ids, event, result);
+        :Publish approve_request + heuristic verdicts;

        note right
-            **auto_approve check is handled
-            internally by ui.approve_tools()**
-
-            **TerminalUI**: Print headers/previews,
-            prompt [y/n/a, optional message]
-            If user chose "always":
-              Add pending tool names to auto_approve_tools
-              (auto-approve these tool types going forward)
-            **WebUI**: Enqueue approve_request,
-            block on _approval_event.wait()
-            **NullUI**: Return (True, None)
+          Parallel task agents can hold several cycles at once.
+          A decision selects one cycle_id / call_id (or the oldest
+          cycle for a legacy selector-less client). Double resolve
+          is a guarded no-op; one cycle cannot wake a sibling.
        end note

-        if (user approved?) then (yes)
-            :_emit_state("running");
-        else (denied)
-            :Mark all pending items as denied;
-            :denial_msg = "Denied by user";
-            :_emit_state("running");
+        if (operator approves?) then (yes)
+            :Record decision and optional feedback;
+        else (denies / policy blocks)
+            :Mark only pending items denied;\nEffectStatus = none;
        endif
-    else (all auto-approved)
-        :ui enqueues tool_info event\n(no blocking);
+        :Publish approval_resolved;\nunregister this cycle;
+    else (all bypassed / auto-approved)
+        :Publish tool_info with the exact\nauto-approve reason per item;
+    endif
+
+    if (owning operation cancelled?) then (yes)
+        :Cancel only cycles carrying that witness;
+        :Stage every unstarted call as\nEffectStatus = none;
+        stop
    endif
 }

-partition "Phase 3: Execute" #E3F2FD {
-    :_check_cancelled();
-    note right: Cancellation checkpoint:\nraises GenerationCancelled if\ncancel event is set
-    if (single tool call?) then (yes)
-        :Execute sequentially:\nrun_one(items[0]);
-    else (multiple)
-        :Execute in parallel:\nThreadPoolExecutor(max_workers=4)\npool.map(run_one, items);
+partition "Phase 3 — Execute" #E3F2FD {
+    :Generation + cancellation checkpoint;
+
+    if (batch requires serial ordering?) then (yes)
+        :Execute in provider order;
+    else (no)
+        :Execute via bounded ThreadPoolExecutor;
    endif

    note right
-        **run_one(item):**
-        if item.error → return error string
-        if item.denied → return denial message
-        else → item["execute"](item)
-          ├─ _exec_bash: subprocess.run(["bash", script.sh])
-          ├─ _exec_read_file: open().readlines() or _exec_read_image (base64)
-          ├─ _exec_write_file: makedirs + write
-          ├─ _exec_edit_file: find_occurrences + replace
-          ├─ _exec_search: grep subprocess
-          ├─ _exec_web_fetch: httpx.get + LLM summary
-          ├─ _exec_web_search: SearxNG JSON GET (fallback for local models)
-          ├─ _exec_tool_search: BM25 search + expand_visible()
-          ├─ _exec_task: _run_agent(TASK_AGENT_TOOLS)
-          ├─ _exec_notify: HTTP POST to channel gateway
-          ├─ _exec_memory: structured memory save/search/delete/list
-          ├─ _exec_recall: conversation history FTS5 search
-          ├─ _exec_read_resource: MCPClientManager.read_resource_sync()
-          ├─ _exec_use_prompt: MCPClientManager.get_prompt_sync()
-          └─ _exec_mcp_tool: MCPClientManager.call_tool_sync()
+      Each worker marks its call started only after the final
+      generation/cancel check. A missing result after that edge is
+      conservatively unknown; an unstarted call is definitively none.
    end note

-    :Collect results: [(call_id, output), ...];
+    :Stream tool chunks to the matching call card;
+    :Capture result / error / preview and effect disposition;

-    :_truncate_output() on each result\n(max context_window × chars_per_token × 0.5 chars\ndefault: ~context_window × 2 chars);
+    if (Stop interrupts execution?) then (yes)
+        :Abort child model scopes and subprocess groups;
+        :Synthesize cancellation receipts;
+        note right
+          EffectStatus vocabulary:
+          committed / none / unknown /
+          partial / rolled_back.

-    :bash: ui.on_tool_output_chunk(call_id, line) per stdout line;
-    :ui.on_tool_result(call_id, name, output, is_error) for each;
+          Observed but unreviewed bytes are omitted from the
+          model-facing receipt; effect truth is retained.
+        end note
+    endif
 }

-:Return (results, user_feedback);
+partition "Phase 4 — Guard and atomic fold" #F3E5F5 {
+    if (compaction already owed?) then (yes)
+        :Compact before sizing/folding results;\npreserve the assistant tool-call Turn;
+    endif
+
+    :Truncate each result against the remaining shared budget;
+    :Run heuristic + optional LLM output guard;
+    :Re-check generation after guard work;
+
+    :Under one generation commit, append the complete\nTool Turn block + advisories + feedback;
+    :Persist rows and effect/preview metadata\non the ordered durability lane;
+    :Return results to the next model turn;
+}

 stop
-
@enduml
@@ -3,6 +3,7 @@
 title Turnstone — Workstream State Machine

 skinparam state {
+    BackgroundColor<<lifecycle>> #ECEFF1
    BackgroundColor<<idle>> #E8F5E9
    BackgroundColor<<thinking>> #E3F2FD
    BackgroundColor<<running>> #FFF3E0
@@ -10,13 +11,18 @@ skinparam state {
    BackgroundColor<<error>> #FFCDD2
 }

+state "CREATING (persisted only)" as creating <<lifecycle>> : Hidden durable reservation.\nNot returned by ordinary list/open/history.
 state "IDLE" as idle <<idle>> : Waiting for user input.\nNo active LLM call or tool execution.
 state "THINKING" as thinking <<thinking>> : LLM streaming response.\nTokens flowing (reasoning + content).
 state "RUNNING" as running <<running>> : Tools executing.\nThreadPoolExecutor active.
 state "ATTENTION" as attention <<attention>> : Blocked on user action.\nTool approval needed.
 state "ERROR" as error <<error>> : Exception occurred.\nRecoverable on next send().
+state "CLOSED (persisted only)" as closed <<lifecycle>> : Unloaded, explicitly reopenable row.\nNot a live WorkstreamState member.

-[*] --> idle : Session created
+[*] --> creating : register exact incarnation\nstate="creating"
+creating --> idle : finalize + publish create\nemit ws_created
+creating --> [*] : immediate exact-token rollback\n(no lifecycle birth emitted)
+creating --> [*] : stale >2h recovery\natomic hard delete; no close event

 idle --> thinking : send() called\n_emit_state("thinking")

@@ -38,6 +44,22 @@ running --> error : Exception during\ntool execution

 error --> thinking : New send() call\n_emit_state("thinking")

+idle --> closed : close / eviction\n[journal reconciled]
+error --> closed : close\n[journal reconciled]
+thinking --> closed : close\n[journal reconciled]
+running --> closed : close\n[journal reconciled]
+attention --> closed : close\n[journal reconciled]
+closed --> [*] : hard delete
+closed --> idle : open / rehydrate
+
+note right of closed
+  Before every soft-close / eviction transition,
+  the total accepted conversation-row journal must
+  be durably reconciled. An unresolved row makes an
+  explicit close return HTTP 409 (eviction refuses),
+  and the workstream remains loaded in its live state.
+end note
+
 thinking --> idle : cancel() called\nstream aborted\n_emit_state("idle")

 running --> idle : cancel() called\n_emit_state("idle")
@@ -45,33 +67,75 @@ running --> idle : cancel() called\n_emit_state("idle")
 attention --> idle : cancel() unblocks\napproval wait\n_emit_state("idle")

 note left of idle
-  **Cancel escalation:**
-  1. **Cooperative**: cancel() sets event + closes
-     SDK stream → worker exits at next checkpoint
-  2. **Force**: force=true abandons the worker
-     thread, emits stream_end immediately.
-     Orphaned thread still kills subprocesses
-     but skips message mutations (generation
-     counter prevents stale writes).
+  **Generation-scoped Stop:**
+  • Sets the active generation event.
+  • Closes its SDK stream; aborts child model
+    scopes and judges; kills subprocess groups.
+  • Sweeps every approval cycle owned by the
+    cancelled workstream operation.
+  • Every later send/model live or durable commit
+    re-checks generation ownership.
+
+  **force=true:** also abandons the stuck worker
+  slot and emits stream_end + IDLE immediately.
+  An orphaned send/model generation may unwind
+  but cannot publish into a successor generation.
+  Quick slash-command workers are a best-effort
+  escape hatch: without generation checkpoints,
+  one may finish an in-place mutation concurrently.
+
+  **Capacity eviction:** an IDLE candidate is only
+  a hint. Per-ID + object lifecycle lanes and the
+  workstream lock revalidate it as worker- and
+  send-barrier-free,
+  then install a terminal claim before slot swap.
 end note

 note right of thinking
  **Emitted via:**
  session._emit_state(state)
    → ui.on_state_change(state)
+    → SessionManager state tail

  **Propagation:**
  • WebUI → global SSE queue (ws_state)
-  • Console → HTTP polling picks up state
-  • CLI → WorkstreamManager.set_state()
+  • Console → cluster event / HTTP state
+  • CLI → SessionManager.set_state()
+
+  Non-terminal persistence may use StateWriter;
+  a per-id tail orders storage + subscribers and
+  prevents a late state from overwriting CLOSED.
 end note

 note left of attention
  **Blocking mechanisms:**
  • TerminalUI: input() prompt
-  • WebUI: threading.Event.wait()
+  • WebUI: one Event per ApprovalCycle
  • ChannelBot: SSE event + Discord button
  • NullUI: auto-approve (never reaches)
 end note

+note right of creating
+  CREATING and CLOSED are storage lifecycle
+  values, not members of WorkstreamState. The
+  live enum remains IDLE / THINKING / RUNNING /
+  ATTENTION / ERROR.
+
+  **Crash-abandoned CREATING recovery:**
+  • Boot pass, then every 5 min even when idle
+    eviction is disabled.
+  • Only rows >2h old; manager loaded/pending
+    IDs and live remote owners are protected.
+  • The current stable node ID is not a live-owner
+    exemption, allowing restart recovery.
+  • Unknown liveness/storage fails closed. Deletion
+    is atomic across dependents and attachment refs.
+  • Tokenless legacy/corrupt rows are locked,
+    reaped, and logged with a warning.
+
+  A loaded hard delete closes publication, drains
+  admitted session durability + state tails, then
+  conditionally removes the exact durable token.
+end note
+
@enduml
@@ -31,7 +31,7 @@ node "Docker Host" as host {
                Command: turnstone-console
                  --port 8090
                Depends: server
-                Hash-ring router for
+                FNV-1a rendezvous router for
                  multi-node clusters
            end note
        }
@@ -69,7 +69,7 @@ apiclient --> server : HTTP + SSE\nport 8080
 ' Internal connections
 server --> llm_api : OpenAI API\n(HTTPS/HTTP)

-console --> server : HTTP proxy\n(hash-ring lookup,\nproxy /node/{id}/*)
+console --> server : HTTP proxy\n(FNV-1a rendezvous placement,\nproxy /node/{id}/*)

 ' Database connections (production/cluster profiles)
 server ..> pgbouncer : PostgreSQL\n(pool_size=2)
@@ -32,7 +32,8 @@ package "turnstone/sdk/ (Python)" {
        + approve()
        + command()
        + cancel(ws_id)
-        + stream_events(ws_id)
+        + get_history(ws_id, limit) → WorkstreamHistoryResponse
+        + stream_events(ws_id, last_event_id?, history_token?)
        + stream_global_events()
        + send_and_wait()
        + list_saved_workstreams()
@@ -87,6 +88,13 @@ package "turnstone/sdk/ (Python)" {
        + ok: bool
    }

+    class WorkstreamHistoryResponse <<type>> {
+        + ws_id: str
+        + messages: list[dict]
+        + cursor: int | None
+        + handoff_token: str | None
+    }
+
    class ServerEvent <<event>> {
        + type: str
        + ws_id: str
@@ -105,6 +113,7 @@ package "turnstone/sdk/ (Python)" {
    TurnstoneConsole --> AsyncTurnstoneConsole : wraps
    TurnstoneConsole --> _SyncRunner : uses
    AsyncTurnstoneServer ..> TurnResult : returns
+    AsyncTurnstoneServer ..> WorkstreamHistoryResponse : renders before SSE
    AsyncTurnstoneServer ..> ServerEvent : yields
    AsyncTurnstoneConsole ..> ClusterEvent : yields
 }
@@ -122,7 +131,8 @@ package "sdk/typescript/ (TypeScript)" {
    class "TurnstoneServer" as TSServer <<ts>> {
        + listWorkstreams()
        + send()
-        + streamEvents()
+        + getHistory() → WorkstreamHistoryResponse
+        + streamEvents(cursor?, token?)
        + sendAndWait()
        ...
    }
@@ -154,4 +164,10 @@ note right of AsyncTurnstoneServer
  (no type duplication)
 end note

+note bottom of ServerEvent
+  history_resync is a typed repair signal.
+  SDKs expose the REST cursor/token handshake but
+  never refetch, render, or reconnect automatically.
+end note
+
@enduml
@@ -1,170 +1,191 @@
@startuml
 !theme plain
-title Turnstone — Storage Architecture
+title Turnstone — Storage, Deferred Create, Fork, and Checkpoint Architecture

 skinparam class {
    BackgroundColor<<protocol>> #E8EAF6
    BackgroundColor<<sqlite>> #C8E6C9
    BackgroundColor<<postgres>> #B3E5FC
-    BackgroundColor<<facade>> #FFF9C4
-    BackgroundColor<<migration>> #FFE0B2
+    BackgroundColor<<lifecycle>> #FFF9C4
    BackgroundColor<<schema>> #F3E5F5
+    BackgroundColor<<helper>> #FFE0B2
 }

-' -- Protocol --
-interface "StorageBackend" as SB <<protocol>> {
-    +save_message(ws_id, role, content, ...)
-    +load_messages(ws_id) → list[dict]
-    +register_workstream(ws_id, node_id, name, state)
-    +update_workstream_state(ws_id, state)
-    +update_workstream_name(ws_id, name)
-    +set_workstream_alias(ws_id, alias) → bool
-    +update_workstream_title(ws_id, title)
-    +resolve_workstream(alias_or_id) → str | None
-    +delete_workstream(ws_id) → bool
-    +prune_workstreams(retention_days) → (int, int)
-    +list_workstreams(node_id, limit, *, parent_ws_id, kind, user_id) → list
-    +save_workstream_config(ws_id, config)
-    +load_workstream_config(ws_id) → dict
-    +kv_get(key) → str | None
-    +kv_set(key, value) → str | None
-    +kv_delete(key) → bool
-    +kv_list() → list[(str, str)]
-    +kv_search(query) → list[(str, str)]
-    +search_history(query, limit) → list
-    +search_history_recent(limit) → list
-    +create_user(user_id, username, display_name, pw_hash)
-    +get_user(user_id) / get_user_by_username(username)
-    +list_users() / delete_user(user_id)
-    +create_api_token(...) / get_api_token_by_hash(hash)
-    +list_api_tokens(user_id) / delete_api_token(id)
-    +close()
-}
-
-' -- Backends --
-class "SQLiteBackend" as SQLite <<sqlite>> {
-    -_engine: sa.Engine
-    -_fts5_available: bool
-    +__init__(path: str)
+interface "StorageBackend" as Storage <<protocol>> {
+    + load_message_turns(ws_id, checkpointed=True) → list[Turn]
+    + save_message(ws_id, role, content, metadata...)
+    + clone_workstream(source, destination, principal, expected_session) → ForkCloneSnapshot
    --
-    FTS5 full-text search
-    Default pool, check_same_thread=False
+    + register_workstream(..., state, reservation_token) → bool
+    + ensure_workstream_incarnation_snapshot(ws_id) → row + token
+    + finalize_deferred_create(ws_id, token, config...) → bool
+    + publish_deferred_create(ws_id, token) → bool
+    + delete_workstream_if_fork_reserved(ws_id, token) → bool
+    + delete_stale_creating_reservations(...) → list[ws_id]
+    + update_workstream_state(ws_id, state)
+    + delete_workstream(ws_id) → bool
+    --
+    + attachment / project / memory / auth / governance APIs
+}
+
+class "SQLiteBackend" as SQLite <<sqlite>> {
+    - _engine: sa.Engine
+    - _fts5_available: bool
+    --
+    Fork clone: BEGIN IMMEDIATE
+    FTS5 refresh in same transaction
 }

 class "PostgreSQLBackend" as PG <<postgres>> {
-    -_engine: sa.Engine
-    +__init__(url: str, pool_size: int = 2,\n    max_overflow: int = 3)
+    - _engine: sa.Engine
    --
-    tsvector + ILIKE search
-    Connection pooling (5 max per process)
+    Fork clone: SERIALIZABLE + row locks
+    Retry SQLSTATE 40001 / 40P01
+    DML success uses RETURNING rows
 }

-' -- Schema --
-class "_schema.py" as Schema <<schema>> {
-    +metadata: MetaData
-    +memories: Table
-    +conversations: Table
-    +workstreams: Table (node_id, alias, title,\n    state, skill_id)
-    +workstream_config: Table
-    +users: Table (username, password_hash)
-    +api_tokens: Table (token_hash, scopes)
-    +channel_users: Table (channel_type)
-    +scheduled_tasks: Table (..., skill)
+class "_utils.py" as Utils <<helper>> {
+    + reconstruct_turns(rows) → list[Turn]
+    + recover_trajectory(turns) → list[Turn]
+    + reconstruct_turns_checkpointed(...)
+    + retain_attachment_refs(conn, ids)
+    + release_attachment_refs(conn, ids)
+    + clone_workstream_transaction(...) → ForkCloneSnapshot
+}
+
+class "ForkCloneExpectation" as Expectation <<lifecycle>> {
+    + persona_config
+    + project_id / name / writable
+    + source_reservation_token
+    + destination_reservation_token
+}
+
+class "ForkCloneSnapshot" as Snapshot <<lifecycle>> {
+    + turns: tuple[Turn, ...]
+    + config: dict[str, str]
+    + project_id: str | None
+}
+
+class "workstreams" as Workstreams <<schema>> {
+    ws_id PK
+    state: creating | live state | closed
+    user_id, node_id, kind, parent_ws_id
+    project_id, persona, alias, title
+}
+
+class "conversations" as Conversations <<schema>> {
+    canonical persisted Turn rows
+    provider_data + tool_calls mirror
+    event_id, source, is_error, meta
+    attachment-id ref list
    --
-    SQLAlchemy Core
-    Single source of truth
+    compaction marker:
+    source="compaction"
+    meta.watermark=<folded row id>
 }

-' -- Migration --
-class "_migrate.py" as Migrate <<migration>> {
-    +run_migrations(storage, backend)
-    -_bootstrap_existing_sqlite()
-    --
-    Programmatic Alembic
-    Auto-bootstrap existing DBs
+class "workstream_config" as WorkstreamConfig <<schema>> {
+    PK (ws_id, key)
+    stamped persona/session config
+    private durable incarnation fence:
+    __fork_destination_reservation
 }

-class "migrations/" as Versions <<migration>> {
-    001_initial_schema.py
-    002_user_identity.py
+class "workstream_attachments" as Attachments <<schema>> {
+    content-addressed blob
+    attachment_id, bytes, kind
+    refcount
 }

-' -- Registry --
-class "_registry.py" as Registry {
-    -_storage: StorageBackend | None
-    +init_storage(backend, path, url) → StorageBackend
-    +get_storage() → StorageBackend
-    +reset_storage()
-    --
-    Auto-initializes SQLite
-    if not configured
+class "projects + project_members" as Projects <<schema>> {
+    visibility / owner / membership
+    active project-memory envelope
 }

-' -- Facade --
-class "memory.py" as Facade <<facade>> {
-    +save_message()
-    +load_messages()
-    +register_workstream()
-    +update_workstream_state()
-    +save_workstream_config()
-    +save_memory() / delete_memory()
-    +search_memories()
-    +... (all delegated functions)
-    --
-    Thin delegation to
-    get_storage()
-    Silent failure behavior
+class "SessionManager" as Manager <<lifecycle>> {
+    + create(..., defer_emit_created)
+    + commit_create(ws)
+    + discard(ws)
+    + reap_stale_creating_reservations(max_age=2h)
+    + open / close / delete
 }

-' -- Consumers --
-class "session.py\nChatSession" as Session {
+class "ChatSession" as Session <<lifecycle>> {
+    + append canonical Turns
+    + compact / resume checkpoint
+    + fork_from_storage(...)
 }

-class "server.py\nWeb UI" as Server {
-}
+SQLite ..|> Storage
+PG ..|> Storage
+SQLite --> Utils
+PG --> Utils

-class "cli.py\nTerminal" as CLI {
-}
+Storage --> Workstreams
+Storage --> Conversations
+Storage --> WorkstreamConfig
+Storage --> Attachments
+Storage --> Projects

-' -- Relationships --
-SQLite ..|> SB
-PG ..|> SB
+Manager --> Storage : lifecycle reservation + state
+Session --> Storage : turn durability + resume
+Session --> Expectation : construction witness
+Storage --> Snapshot : atomic clone result
+Expectation --> Utils : checked inside transaction
+Utils --> Snapshot : builds

-SQLite --> Schema : uses
-PG --> Schema : uses
+note right of Manager
+  **Deferred create publication**
+  1. INSERT workstream as state="creating" and store a fresh
+     private token in the same transaction.
+  2. Construct UI/session and run attachment/fork gates while
+     ordinary list/open/history reads exclude the row.
+  3. finalize_deferred_create atomically applies config/alias.
+  4. publish_deferred_create compare-and-swaps creating → idle.
+  5. Only then emit ws_created.

-Registry --> SB : creates
-Registry --> Migrate : calls
-
-Migrate --> Versions : applies
-Migrate --> Schema : references
-
-Facade --> Registry : get_storage()
-
-Session --> Facade : imports
-Server --> Facade : imports
-CLI --> Facade : imports
-
-' -- Config --
-note right of Registry
-  [database]
-  backend = "sqlite" | "postgresql"
-  url = "postgresql+psycopg://..."
-  path = ".turnstone.db"
-  pool_size = 2 (+ 3 overflow)
+  Any normal prepublication failure immediately calls exact token-checked
+  deletion. The token survives publication as the row's incarnation fence:
+  rollback or later hard delete can never ABA-delete a replacement row.
+  A legacy row acquires the same private token atomically when rehydrate,
+  delete, or fork preflight takes its authoritative snapshot. Loaded hard
+  delete drains admitted session durability before its token-checked delete.
 end note

-note bottom of SQLite
-  Default backend.
-  Zero-config for
-  single-node / dev.
+note left of Manager
+  **Crash-abandoned hidden-create recovery**
+  • Boot pass; long-lived processes repeat every 5 min,
+    even when ordinary idle eviction is disabled.
+  • Candidates remain state="creating", are >2h old,
+    and are absent from the manager loaded/pending set.
+  • Live remote owners are protected. The current stable
+    node ID does not self-protect, enabling restart recovery.
+  • Unknown liveness or storage failure deletes nothing.
+  • One transaction rechecks state, age, and token, then
+    hard-deletes dependents and releases attachment refs.
+  • Tokenless legacy/corrupt rows use their locked durable
+    row as the incarnation fence and log a warning.
+  • Retention pruning excludes creating rows. Recovery never
+    closes or publishes them as live WorkstreamState values.
 end note

-note bottom of PG
-  Production backend.
-  Multi-node / Docker default.
-  Use PgBouncer (transaction mode)
-  for clusters > 50 nodes.
+note bottom of Utils
+  **Atomic fork clone**
+  • Reject a provisional source; compare the source incarnation captured
+    by canonical preflight; re-authorize project visibility and compare the
+    live session envelope inside the transaction.
+  • Require a same-owner, empty destination still in creating state
+    with the exact reservation token.
+  • Copy the checkpoint-bounded canonical trajectory and config;
+    retain every referenced attachment or roll everything back.
+  • Preserve/rebase a valid compaction checkpoint watermark and
+    return the exact snapshot installed into the live destination.
+end note
+
+note bottom of Conversations
+  Full transcript rows are never deleted by compaction. Normal resume
+  loads the latest valid [summary] + rows after its watermark; audit and
+  export can request the full marker-free history.
 end note

@enduml
@@ -1,190 +1,153 @@
@startuml
 !theme plain
-title Turnstone — Authentication Architecture
+title Turnstone — User Authentication and Model-Backend Credentials

 skinparam class {
    BackgroundColor<<core>> #E8EAF6
-    BackgroundColor<<jwt>> #C8E6C9
+    BackgroundColor<<token>> #C8E6C9
    BackgroundColor<<storage>> #B3E5FC
-    BackgroundColor<<endpoint>> #FFE0B2
-    BackgroundColor<<scope>> #F3E5F5
+    BackgroundColor<<runtime>> #FFE0B2
+    BackgroundColor<<model>> #F3E5F5
 }

-' -- Core Auth --
-class "AuthConfig" as AC <<core>> {
-    +enabled: bool
-    +tokens: dict[str, str]
-    +check(token) → role | None
-    --
-    Static config-file tokens
-    hmac.compare_digest
+package "Request identity" {
+  class "AuthMiddleware / check_request()" as RequestAuth <<core>> {
+      Extract bearer or HttpOnly cookie
+      Validate audience + expiry
+      Check scope / permission
+      Publish AuthResult in request state
+  }
+
+  class "AuthResult" as AuthResult <<core>> {
+      + user_id: str
+      + scopes: frozenset[str]
+      + permissions: frozenset[str]
+      + token_source: str
+  }
+
+  class "JWT" as JWT <<token>> {
+      HS256, sub, aud, iat, exp
+      console proxy mints short-lived
+      server-audience identity
+  }
+
+  class "API / config token" as ApiToken <<token>> {
+      ts_* token: SHA-256 DB lookup
+      config token: constant-time compare
+  }
+
+  class "users / roles / api_tokens" as UserTables <<storage>> {
+      password hash + token hash
+      role-derived permissions
+  }
 }

-class "AuthResult" as AR <<core>> {
-    +user_id: str
-    +scopes: frozenset[str]
-    +token_source: str
-    +has_scope(scope) → bool
+package "Immutable model binding" {
+  class "ModelRegistry" as Registry <<model>> {
+      + resolve_binding(alias)
+      + generation: int
+      --
+      Atomically resolves client, provider,
+      model, ModelConfig, generation.
+  }
+
+  class "ModelConfig snapshot" as ModelConfig <<model>> {
+      + alias / provider / endpoint / static key
+      + auth_mode
+      + obo_audience
+      + obo_scopes
+      --
+      static | entra_obo | entra_app | rfc8693_obo
+  }
+
+  class "ModelLane" as Lane <<model>> {
+      + client / provider / model / capabilities
+      + backend_auth_config: ModelConfig
+      + backend_auth_resolver: Callable
+  }
+
+  class "Model definitions" as ModelTable <<storage>> {
+      DB + config-file definitions
+      encrypted protected fields
+  }
 }

-class "check_request()" as CR <<core>> {
-    auth_config, method, path,
-    auth_header, cookie_header,
-    jwt_secret, storage
-    → (allowed, status, msg, AuthResult)
-    --
-    1. Auth disabled → allow
-    2. Public path → allow
-    3. Extract Bearer / cookie
-    4. Detect token type
-    5. Validate → AuthResult
-    6. Check scope vs path
+package "Per-call credential resolution" {
+  class "resolve_model_backend_auth_token()" as Resolver <<runtime>> {
+      + alias + pinned ModelConfig
+      + initiating principal_id
+      + ConfigStore + mint client
+      → dynamic token | None | fail closed
+  }
+
+  class "Model mint client" as Mint <<runtime>> {
+      + mint_model_obo_token_sync(...)
+      + mint_app_token_sync(...)
+      --
+      Cached by alias / principal / grant leg;
+      retains refusal cause for diagnostics.
+  }
+
+  class "OIDC / OBO protected state" as OBOState <<storage>> {
+      encrypted user refresh credential
+      deployment Fernet key
+      configured grant profile
+  }
+
+  class "lane_call_client()" as CallClient <<runtime>> {
+      cancel check before mint
+      resolve once per plant call
+      cancel check after mint
+      client.with_options(api_key=token)
+  }
+
+  class "Provider SDK request" as ProviderCall <<runtime>> {
+      Anthropic: x-api-key
+      OpenAI-style: Authorization Bearer
+  }
 }

-' -- Token Types --
-class "JWT (HS256)" as JWT <<jwt>> {
-    sub: user_id
-    scopes: "read,write,approve"
-    src: "password" | "database"
-    iat, exp (24h default)
-    --
-    Detected by: contains "."
-    Validated locally
-    No DB call
-}
+RequestAuth --> JWT : validates
+RequestAuth --> ApiToken : validates
+RequestAuth --> UserTables : lookup + permissions
+RequestAuth --> AuthResult : returns

-class "API Token" as AT <<jwt>> {
-    Format: ts_ + 64 hex
-    Stored: SHA-256 hash
-    --
-    Detected by: starts with "ts_"
-    Lookup by hash in DB
-    Expiry check
-}
+ModelTable --> Registry : load / hot reload
+Registry --> ModelConfig : immutable snapshot
+Registry --> Lane : coherent binding

-class "Config Token" as CT <<core>> {
-    Raw value in memory
-    Role: "read" | "full"
-    --
-    Detected by: fallback
-    hmac.compare_digest
-    No DB needed
-}
+AuthResult --> Resolver : initiating principal
+Lane --> Resolver : callable + pinned config
+Resolver --> Mint : dynamic modes only
+Mint --> OBOState : decrypt / grant policy
+CallClient --> Lane
+CallClient --> Resolver
+CallClient --> ProviderCall : cloned SDK client

-' -- Scopes --
-class "Scope Hierarchy" as SH <<scope>> {
-    read: {read}
-    write: {read, write}
-    approve: {read, write, approve}
-    --
-    GET → read
-    POST write paths → write
-    POST /api/workstreams/{ws_id}/approve → approve
-    /api/admin/* → approve
-}
-
-' -- Storage --
-class "users" as UT <<storage>> {
-    user_id (PK)
-    username (unique)
-    display_name
-    password_hash (bcrypt)
-    created
-}
-
-class "api_tokens" as TT <<storage>> {
-    token_id (PK)
-    token_hash (SHA-256, unique)
-    token_prefix
-    user_id → users
-    name, scopes
-    created, expires
-}
-
-' -- Endpoints --
-class "POST /api/auth/login" as Login <<endpoint>> {
-    {username, password}
-    OR {token: "ts_xxx"}
-    → {jwt, role, scopes, user_id}
-    --
-    Sets HttpOnly cookie
-}
-
-class "GET /api/auth/status" as Status <<endpoint>> {
-    → {auth_enabled, has_users,
-       setup_required}
-    --
-    Public (no auth)
-    Drives UI setup wizard
-}
-
-class "POST /api/auth/setup" as Setup <<endpoint>> {
-    {username, display_name, password}
-    → {jwt, user_id, scopes}
-    --
-    Public (no auth)
-    Only when zero users exist
-    Returns 409 if already set up
-}
-
-class "Admin API (Console)" as Admin <<endpoint>> {
-    POST/GET/DELETE users
-    POST/GET tokens
-    DELETE tokens/{id}
-    --
-    Requires approve scope
-}
-
-' -- Relationships --
-CR --> AC : config tokens
-CR --> JWT : validate
-CR --> AT : hash lookup
-CR --> CT : hmac check
-CR --> AR : returns
-CR --> SH : checks
-
-Login --> JWT : issues
-Login --> UT : verify password
-Login --> TT : verify API token
-
-Setup --> UT : create first user
-Setup --> JWT : issues
-
-AT --> TT : lookup by hash
-Admin --> UT : CRUD
-Admin --> TT : CRUD
-
-AR --> SH : scopes from
-
-JWT ..> AR : produces
-AT ..> AR : produces
-CT ..> AR : produces
-
-note right of CR
-  **Middleware Flow**
-  AuthMiddleware on every request:
-  1. Extract token from header/cookie
-  2. Detect type (JWT / ts_ / config)
-  3. Validate → AuthResult
-  4. Set ctx_user_id for logging
-  5. Store auth_result in scope state
+note right of Resolver
+  **Mode policy**
+  • static: return None; registry client's explicit key remains.
+  • entra_obo / rfc8693_obo: require an effective principal. HTTP
+    turns pin the authenticated initiator; single-user internal lanes
+    may use their session owner. Never borrow another generation's identity.
+  • entra_app: use deployment app identity, no user required.
+  • rfc8693_obo alone sends obo_scopes; each dynamic mode is paired
+    with its required Entra or RFC 8693 grant profile.
 end note

-note bottom of SH
-  **Console** owns admin endpoints
-  **Server** validates JWT + config only
-  Both share JWT signing secret
+note bottom of CallClient
+  Dynamic credentials are minted at dispatch, not cached in the registry
+  snapshot. Endpoint, audience, scopes, auth mode, and static-key presence stay
+  pinned to the same ModelConfig generation as the SDK client. The global
+  model.auth_fail_closed policy is read live on every mint. A Stop that wins
+  before or during mint prevents model bytes from being sent afterward.
 end note

-note left of JWT
-  **Console Proxy Token Minting**
-  When proxying requests to server nodes:
-  1. Console AuthMiddleware validates user JWT (aud: turnstone-console)
-  2. Proxy mints new JWT (aud: turnstone-server)
-     with real user_id, scopes, permissions
-  3. src: "console-proxy" for audit traceability
-  4. 5-minute expiry (fresh per request)
-  5. Fallback: ServiceTokenManager if no user context
+note bottom of ProviderCall
+  If minting fails, a configured fail-closed deployment or a keyless alias
+  raises BackendAuthUnavailableError. A dynamic alias with an explicit static
+  key may fall back only when policy allows. Authentication refusal is not a
+  backend-health failure and does not walk to a static fallback model.
 end note

@enduml
@@ -82,10 +82,12 @@ class "DiscordBot" as Bot <<service>> {
 }

 class "ChannelRouter" as Router <<service>> {
-    +resolve_route(platform, channel_id)
-        -> ws_id | None
-    +register_route(channel_id, ws_id)
-    +resolve_identity(platform, platform_user_id)
+    +get_or_create_workstream(channel_type, channel_id)
+    +_is_ws_live(ws_id)
+    +send_message(ws_id, message)
+    +send_approval(ws_id, ...)
+    +lookup_ws_id(channel_type, channel_id)
+    +resolve_user(channel_type, channel_user_id)
        -> user_id | None
    --
    Maps channels -> workstreams
@@ -93,6 +95,16 @@ class "ChannelRouter" as Router <<service>> {
    Caches routes in memory
 }

+class "turnstone-console router" as ConsoleRouter <<server>> {
+    POST /v1/api/route/workstreams/new
+    GET  /v1/api/route/workstreams/{ws_id}/live
+    POST /v1/api/route/workstreams/{ws_id}/send
+    POST /v1/api/route/workstreams/{ws_id}/approve
+    GET  /v1/api/route?ws_id=...
+    --
+    Multi-node rendezvous + durable overrides
+}
+
 ' -- Server --
 class "turnstone-server" as Server <<server>> {
    POST /v1/api/workstreams/{ws_id}/send
@@ -148,7 +160,9 @@ Bot --> Router : on_message\non_interaction
 Router --> CU : resolve identity
 Router --> CR : resolve / register route

-Router --> Server : POST /v1/api/workstreams/{ws_id}/send\nPOST /v1/api/workstreams/{ws_id}/approve\nPOST /v1/api/workstreams/new
+Router --> Server : single-node/direct mode\ncreate + send + approve
+Router --> ConsoleRouter : multi-node mode\nroute create/live/send/approve/lookup
+ConsoleRouter --> Server : routed HTTP to owning node
 Bot --> Server : GET /v1/api/workstreams/{ws_id}/events\n(SSE via httpx-sse)
 Server --> Bot : SSE event stream

@@ -156,7 +170,7 @@ Bot --> Discord : reply / embed\nbutton callback

 Slack --> SlackBot : socket-mode\nevents
 SlackBot --> Router : on_message / on_action
-SlackBot --> Server : POST /v1/api/workstreams/{ws_id}/send\nGET /v1/api/workstreams/{ws_id}/events
+SlackBot --> Server : owning-node SSE after route lookup
 SlackBot --> Slack : post / update\nBlock Kit button callbacks

 Teams .[hidden]. Slack
@@ -175,19 +189,21 @@ note right of Bot
  **Inbound Flow**
  1. Discord message arrives via gateway
  2. Bot.on_message() fires
-  3. ChannelRouter resolves channel -> ws_id
-     (or creates new workstream)
+  3. ChannelRouter gets or creates channel -> ws_id
+     (direct server or multi-node console router)
  4. ChannelRouter resolves platform user -> user_id
     via channel_users table
-  5. Router sends POST /v1/api/workstreams/{ws_id}/send to server
+  5. Router sends through the configured server/console SDK

-  **Workstream Resume (evicted workstreams)**
-  1. Stale route detected (no active SSE listener)
-  2. Existing ws_id reused directly from route
+  **Stale-route recovery (evicted workstreams)**
+  1. Route health check reports the old ws unavailable
+  2. Existing ws_id becomes the fork source
  3. POST /v1/api/workstreams/new with
     resume_ws=<ws_id>
-  4. Server resumes atomically during creation
-  5. SSE emits WorkstreamResumedEvent -> thread
+  4. Server atomically clones source history/config/
+     persona/project/attachment refs into a new ws_id
+  5. Router stores the new destination route; source is unchanged
+  6. If the source was pruned, retry one fresh create
 end note

 note right of Server
@@ -11,7 +11,7 @@ skinparam participant {

 participant "ChatSession\n(session.py)" as Session <<session>>
 participant "WatchRunner\n(watch.py)" as Runner <<server>>
-participant "StorageBackend\n(SQLite)" as Storage <<storage>>
+participant "StorageBackend\n(SQLite / PostgreSQL)" as Storage <<storage>>
 participant "WebUI / SSE\n(server.py)" as UI <<ui>>

 == Create Phase ==
@@ -130,7 +130,7 @@ note right : action="cancel" (auto-approve)
 note over Runner, Storage
  **Startup:**
  1. WatchRunner created in main() with storage + node_id
-  2. restore_fn closure captures WorkstreamManager
+  2. restore_fn closure captures SessionManager
  3. Initial workstream: session.set_watch_runner(runner)
  4. _lifespan(): runner.start() — daemon thread begins

@@ -1,218 +1,126 @@
@startuml
 !theme plain
-title Turnstone — Intent Validation (Judge) Architecture
+title Turnstone — Intent Judge, Concurrent Approval Cycles, and Output Guard

-skinparam participant {
-    BackgroundColor<<session>> #C8E6C9
-    BackgroundColor<<judge>> #FFE0B2
-    BackgroundColor<<storage>> #B3E5FC
-    BackgroundColor<<ui>> #E8EAF6
-    BackgroundColor<<fs>> #F5F5F5
-}
+skinparam sequenceArrowThickness 1.5
+skinparam sequenceLifeLineBackgroundColor #F5F5F5

-participant "ChatSession\n(session.py)" as Session <<session>>
-participant "IntentJudge\n(judge.py)" as Judge <<judge>>
-participant "LLM Provider\n(provider)" as LLM <<judge>>
-participant "StorageBackend\n(SQLite)" as Storage <<storage>>
-participant "WebUI / SSE\n(server.py)" as UI <<ui>>
-participant "Filesystem" as FS <<fs>>
+participant "ChatSession\ngeneration N" as Session
+participant "SessionUIBase" as UI
+participant "IntentJudge" as Judge
+participant "model_turn()\n(pinned ModelLane)" as Model
+participant "Operator / client" as Operator
+participant "OutputGuardJudge" as Guard
+database "StorageBackend" as Storage

-== Tool Call Requires Approval ==
+== Intent assessment begins during preparation ==

-Session -> Session : _prepare_tool_calls()
-note right
-  Tool calls parsed from
-  LLM response. Auto-approved
-  tools dispatched immediately.
-  Remaining items need approval.
+Session -> Session : prepare each tool item independently\nattach principal + cancel witness
+Session -> Judge : evaluate(items, callback, cancel_ref)
+activate Judge
+Judge -> Judge : synchronous heuristic verdict\nfor each call (first matching rule)
+Judge --> Session : heuristic verdicts + daemon cancel event
+Session -> UI : cache / publish heuristic assessments
+Session -> Storage : persist heuristic intent verdicts
+
+note over Judge, Model
+  The judge owns an immutable resolved binding. Registry/config generations
+  are freshness watermarks: an effective lane change replaces the judge for
+  the next batch, while in-flight work keeps the lane it started with.
+  Dynamic backend auth is resolved for this batch's initiating principal.
+  parallel_evaluations (1-16) sets per-batch worker width; the model alias's
+  admission gate remains the process-wide generation ceiling.
 end note

-Session -> Session : _evaluate_intent(pending_items)
-
-== Tier 1: Heuristic (synchronous, sub-ms) ==
-
-Session -> Judge : evaluate(items, messages, callback)
-
-Judge -> Judge : evaluate_heuristic()\nfor each item
-note right
-  **36 rules (first match wins):**
-  Critical (0.90, deny): rm /, mkfs,
-    dd, pipe-to-shell, chmod 777 /,
-    write/edit /etc/ .ssh/,
-    download-then-execute chains
-  High (0.80, review): sudo, kill -9,
-    destructive git, DROP TABLE,
-    secrets, HTTP mutations, ssh/scp,
-    browser+data-export, transitive
-    install, control-plane mutation
-  Medium (0.70, review): content
-    ingestion, interpreter exec,
-    cloud CLI mutations, pkg install,
-    write_file, MCP tools, docker ops
-  Low (0.85, approve): read_file,
-    list_directory, search, recall,
-    tool_search, read_resource,
-    web_search, read-only bash
-  Default: medium, 0.50, review
-end note
-
-Judge --> Session : heuristic_verdicts[]
-
-Session -> Session : attach _heuristic_verdict\nto each pending item
-
-Session -> UI : SSE: approve_request\n{items: [{verdict: ...}],\n judge_pending: true}
-note right
-  Heuristic verdict displayed
-  immediately as risk badge.
-  Spinner shown while LLM
-  judge evaluates.
-end note
-
-Session -> Storage : create_intent_verdict()\nfor each heuristic verdict
-
-== Tier 2: LLM Judge (daemon thread, async) ==
-
-Judge -> Judge : spawn daemon thread\n"intent-judge"
-
-note over Judge, LLM
-  **Context preparation:**
-  1. FIFO-truncate conversation history
-     to max_context_ratio of context window
-  2. Append tool call details as user message
-  3. System prompt defines judge role + JSON schema
-end note
-
-loop up to 3 turns (timeout budget)
-
-    Judge -> LLM : model_turn(lane, judge_turns,\ntools=[read_file, list_directory])\nvia drained create_streaming
-    LLM --> Judge : ModelTurnResult
-
-    alt tool_calls present (turn < 3)
-        Judge -> Judge : _exec_read_only_tool()
-        note right
-          **Security hardening:**
-          Blocked: /etc/, /root/,
-          /proc/, /sys/, /dev/,
-          .ssh, .gnupg, .aws,
-          *.pem, *.key, *.p12
-          File cap: 32KB
-          Dir cap: 200 entries
-        end note
-        Judge -> FS : read_file / list_directory
-        FS --> Judge : file contents
-        Judge -> Judge : append tool result\nto judge_messages
-    else text response (final verdict)
-        Judge -> Judge : _parse_verdict()
-        note right
-          **4-stage JSON parsing:**
-          1. Direct JSON.loads
-          2. Markdown code block
-          3. Brace-counting
-          4. Regex field extraction
-        end note
+par LLM judge daemon coordinator
+  Judge -> Judge : start min(batch size, parallel_evaluations,\npositive alias capacity) workers
+  loop each worker claims one independent call
+    Judge -> Model : model_turn(judge lane, canonical Turns,\nread-only evidence tools, cancel_ref)
+    Model --> Judge : ModelTurnResult
+    alt evidence tool requested
+      Judge -> Judge : execute bounded read_file / list_directory
+    else verdict text
+      Judge -> Judge : parse + arbitrate against heuristic
    end
-
+    Judge --> UI : on_intent_verdict(verdict, judge generation)
+    UI -> Storage : persist LLM verdict / audit update
+  end
+else approval path continues
+  Session -> UI : approve_tools(items) with one\nSmart Approval config snapshot
 end

-== Tier 3: Arbitration ==
+== Policy, Smart Approval, and human gate ==

-Judge -> Judge : compare confidence:\nLLM vs heuristic
-note right
-  Only deliver LLM verdict
-  if confidence > heuristic.
-  Otherwise heuristic stands.
-end note
-
-alt LLM confidence > heuristic confidence
-    Judge -> Session : callback(llm_verdict)
-    Session -> UI : SSE: intent_verdict\n{tier: "llm", ...}
-    note right
-      UI replaces heuristic badge
-      with LLM verdict. Spinner
-      resolves to final assessment.
-    end note
-    Session -> Storage : create_intent_verdict()\nfor LLM verdict
+UI -> UI : apply explicit policy / skill / always / blanket bypasses
+opt Smart Approvals enabled
+  UI -> UI : wait within captured deadline for this batch's verdicts
+  UI -> UI : auto-approve only recommendation=approve\nand confidence >= captured threshold
+  UI -> Storage : persist auto-approval reason and decision
 end

-== User Decision ==
-
-UI -> Session : resolve_approval(\napproved, feedback)
-
-Session -> Storage : update_intent_verdict(\nverdict_id, user_decision)
-note right
-  All tracked verdicts
-  (heuristic + LLM) updated
-  with "approved" or "denied".
-  Swap-and-clear avoids racing
-  with daemon judge thread.
-end note
-
-== Tool Execution ==
-
-Session -> Session : _execute_tools()
-note right
-  Tools execute with
-  user approval.
-end note
-
-== Output Guard (synchronous, time-budgeted) ==
-
-Session -> Session : _evaluate_output()\nfor each tool result
-note right
-  **Priority-ordered checks (5s budget):**
-  P1: Prompt injection (role injection,
-    override phrases, instruction tags)
-  P2: Credential leakage (API keys,
-    PEM blocks, connection strings)
-  P3: Encoded payloads (data URIs,
-    hex shellcode)
-  P4: Adversarial URLs (cloud metadata,
-    credential query params)
-  P5: System info disclosure (private
-    IPs, sensitive paths)
-
-  Annotates + optionally redacts.
-  Does NOT gate.
-end note
-
-alt output_warning flags detected
-    Session -> UI : SSE: output_warning\n{call_id, risk_level, flags,\nfunc_name, redacted}
-    note right
-      Credential values replaced
-      with [REDACTED:<type>] before
-      output enters conversation.
-      sanitized text excluded from
-      SSE payload (defense in depth).
-    end note
-    UI -> Storage : record_output_assessment()\nfire-and-forget persistence
-    note right
-      Stored: flags, risk_level,
-      annotations, output_length,
-      redacted (bool). Raw tool
-      output is never stored.
-    end note
+alt human-gated items remain
+  UI -> UI : acquire publication lease; register ApprovalCycle\n(cycle_id, call_ids, event, result, witnesses)
+  UI -> Operator : approve_request with cycle_id + item verdicts
+  Operator -> UI : approve / deny by cycle_id or call_id
+  UI -> UI : atomically claim exactly one unresolved cycle
+  UI -> Operator : approval_resolved
+  UI --> Session : decision + optional feedback
+  UI -> Storage : stamp tracked verdicts with operator decision
+else every item bypassed / auto-approved
+  UI -> Operator : tool_info with exact auto_approve_reason
+  UI --> Session : approved
 end

-== Lifecycle ==
+note right of UI
+  Parallel task agents may register several ApprovalCycles. Each cycle owns
+  its own Event and result slot. A legacy selector-less decision targets the
+  oldest cycle; double resolution is a no-op. Cached LLM verdicts carry their
+  judge generation, so reused provider call ids cannot satisfy a new cycle.
+end note

-note over Session, Judge
-  **Lazy initialization:**
-  IntentJudge created on first approval if judge_config.enabled.
-  Re-uses session's provider/client by default (self-consistency).
-  Cross-model: separate provider/client from [judge] config.
+== Cancellation boundary ==

-  **Sub-agent exemption:**
-  Task sub-agents skip intent validation entirely.
+opt Stop / close / force-successor
+  Session -> Judge : abort all judge events owned by the cancelled operation
+  Session -> UI : resolve_all_approvals(False, "cancelled")
+  UI -> UI : block new admission leases; wait for admitted bundles;\nclaim only cycles whose cancellation witness is aborted
+  UI -> Operator : one cancelled resolution per claimed cycle
+  note over Session, UI
+    A Stop can win before cycle registration, during publication, or while a
+    click resolves. The witness + admission sweep makes exactly one terminal
+    outcome visible; a successor generation's new cycle is not swept.
+  end note
+end

-  **Output guard:**
-  Runs when judge_config.output_guard is true (default).
-  Credential redaction when judge_config.redact_secrets is true.
+note over Judge
+  Normal operator resolution does not necessarily cancel judge inference.
+  With cancel_on_approval=false, the daemon finishes and late verdicts remain
+  auditable. With it enabled, the batch event stops remaining judge work.
+end note

-  **Storage:**
-  intent_verdicts table (migration 012), output_assessments table
-  (migration 022). Both queryable via admin API endpoints
-  (requires admin.judge permission). Skills store risk_level,
-  scan_report, scan_version for install-time risk assessment.
+deactivate Judge
+
+== Tool output guard ==
+
+Session -> Session : execute admitted tools; truncate each result
+Session -> Guard : evaluate(result, tool context, cancel event)
+activate Guard
+Guard -> Guard : heuristic checks first
+opt LLM guard enabled and time remains
+  Guard -> Model : model_turn(output-guard lane, bounded prompt, cancel_ref)
+  Model --> Guard : structured verdict
+end
+Guard --> Session : assessment / redaction / warning
+deactivate Guard
+Session -> Session : re-check generation N before folding result
+Session -> UI : output warning (no raw secret payload)
+Session -> Storage : persist assessment + guarded Tool Turn metadata
+
+note over Guard, Storage
+  Output-guard objects also pin model/config lanes. Replacement retires the
+  old object but lets admitted evaluations drain before its private client is
+  closed. A cancelled or superseded evaluation cannot fold into the successor
+  trajectory. Raw pre-redaction secrets are never stored in assessment rows.
 end note

@enduml
@@ -13,65 +13,68 @@ skinparam participant {
 participant "ChatSession\n(session.py)" as Session <<session>>
 participant "MemoryFacade\n(memory.py)" as Facade <<facade>>
 participant "MemoryRelevance\n(memory_relevance.py)" as Relevance <<facade>>
-participant "StorageBackend\n(SQLite)" as Storage <<storage>>
+participant "StorageBackend\n(SQLite / PostgreSQL)" as Storage <<storage>>
 participant "Server API\n(server.py)" as API <<api>>
 participant "Console Admin\n(console/server.py)" as Admin <<api>>
 participant "SDK Client\n(sdk/)" as SDK <<sdk>>

 == Phase 1: Tool Path (session.send) ==

-Session -> Session : _prepare_tool_calls()\nparse memory(action=...)
+Session -> Session : pin acting principal\nparse memory(action=...)
 note right
-  Tool schema: 4 actions
-  save, search, delete, list
+  Tool schema: 5 actions
+  save, get, search, delete, list
  Auto-approved (no approval needed)
 end note

+Session -> Session : resolve live project access\nselect exact/inherited scope
+
 Session -> Session : _exec_memory(item)

 alt action = save
-    Session -> Facade : save_structured_memory(\nname, content, description,\nmem_type, scope, scope_id)
+    Session -> Session : require non-empty description
+    Session -> Facade : save_structured_memory_strict(\n..., require_active_project)
    Facade -> Facade : normalize_key(name)
-    Facade -> Storage : create_structured_memory()
-    alt unique constraint violation
-        Storage --> Facade : IntegrityError
-        Facade -> Storage : get_structured_memory_by_name()
-        Storage --> Facade : existing row
-        Facade -> Storage : update_structured_memory()
-    end
-    Storage --> Facade : memory_id
-    Facade --> Session : (memory_id, old_content)
-    Session -> Session : _init_system_messages()\nrefresh BM25 context
+    Facade -> Storage : guarded atomic upsert\nON CONFLICT ... RETURNING
+    Storage --> Facade : (saved row, was_update)
+    Facade --> Session : saved row
+    Session -> Session : invalidate prefix/cache\naudit acting principal
+end
+
+alt action = get
+    Session -> Facade : get_structured_memory_by_name_strict()
+    Facade -> Storage : exact scoped-name lookup
+    Storage --> Session : full row / not found
 end

 alt action = search
-    Session -> Facade : search_structured_memories(\nquery, mem_type, scope,\nscope_id, limit)
-    Facade -> Storage : search_structured_memories()
+    Session -> Storage : search exact scope or\nactor-visible scope union
    Storage --> Session : matched rows
 end

 alt action = delete
-    Session -> Facade : delete_structured_memory(\nname, scope, scope_id)
-    Facade -> Storage : delete_structured_memory()
-    Storage --> Session : bool (existed)
-    Session -> Session : _init_system_messages()\nrefresh BM25 context
+    Session -> Facade : delete_structured_memory_returning_strict()
+    Facade -> Storage : DELETE ... RETURNING
+    Storage --> Session : deleted row / not found
+    Session -> Session : invalidate + audit\nmark prefix dirty
 end

 == Phase 2: BM25 Relevance Injection ==

 Session -> Session : _init_system_messages()\nevery conversation turn

+Session -> Session : resolve acting principal\nand live project ACL
 Session -> Session : _list_visible_memories(\nlimit=fetch_limit)
 note right
  **Scope resolution:**
-  1. global scope (always)
-  2. workstream scope (ws_id)
-  3. user scope (user_id, if auth)
-  Combined and deduplicated.
+  Interactive: global + workstream
+  + acting user + readable project
+  Coordinator: acting user's coordinator
+  + readable project
 end note

-Session -> Facade : list_structured_memories()\nper scope
-Facade -> Storage : list_structured_memories()
+Session -> Facade : list_visible_structured_memories()
+Facade -> Storage : one visibility-union query
 Storage --> Session : up to fetch_limit rows

 Session -> Relevance : extract_recent_context(\nmessages, max_messages=3)
@@ -103,28 +106,31 @@ Session -> Session : inject into\nsystem message

 == Phase 3: Server API Path ==

-SDK -> API : GET /v1/api/memories\n?type=project&limit=20
-API -> Facade : list_structured_memories()
-Facade -> Storage : list_structured_memories()
+SDK -> API : GET /v1/api/memories\n?type=general&limit=20
+API -> API : bind scope to caller\ndefault global + caller user
+API -> Storage : list visible rows
 Storage --> API : rows
 API --> SDK : {"memories": [...], "total": N}

-SDK -> API : POST /v1/api/memories\n{name, content, ...}
-API -> API : validate type, scope,\nname length, content length
-API -> Facade : save_structured_memory()
-Facade -> Storage : create / update
+SDK -> API : POST /v1/api/memories\n{name, content, description, ...}
+API -> API : validate type, scope,\nname/content/description
+API -> API : reject internal scopes\nowner-bind workstream scope
+API -> Facade : save_structured_memory_strict()
+Facade -> Storage : atomic upsert
 Storage --> API : memory row
+API -> API : record_audit(actor)
 API --> SDK : 201 (created) / 200 (updated)

 SDK -> API : POST /v1/api/memories/search\n{query, type, ...}
-API -> Facade : search_structured_memories()
-Facade -> Storage : search_structured_memories()
+API -> API : bind scope to caller
+API -> Storage : search visible rows
 Storage --> API : matched rows
 API --> SDK : {"memories": [...], "total": N}

 SDK -> API : DELETE /v1/api/memories/{name}\n?scope=global
-API -> Facade : delete_structured_memory()
-Facade -> Storage : delete row
+API -> Facade : delete_structured_memory_returning_strict()
+Facade -> Storage : DELETE ... RETURNING
+API -> API : record_audit(actor)
 API --> SDK : {"status": "ok"}

 == Phase 4: Console Admin Path ==
@@ -141,7 +147,7 @@ Storage --> Admin : memory row
 Admin --> SDK : memory JSON

 SDK -> Admin : DELETE /v1/api/admin/memories/{id}
-Admin -> Storage : delete_structured_memory_by_id()
+Admin -> Storage : delete_structured_memory_by_id_returning()
 Admin -> Admin : record_audit(\n"memory.delete")
 Admin --> SDK : {"status": "ok"}

@@ -13,7 +13,7 @@ skinparam participant {
 participant "Server\n(main)" as Server <<session>>
 participant "ConfigStore\n(config_store.py)" as Store <<config>>
 participant "SettingsRegistry\n(settings_registry.py)" as Registry <<config>>
-participant "StorageBackend\n(SQLite)" as Storage <<storage>>
+participant "StorageBackend\n(SQLite / PostgreSQL)" as Storage <<storage>>
 participant "Console Admin\n(console/server.py)" as Admin <<api>>
 participant "SDK Client\n(sdk/)" as SDK <<sdk>>
 participant "ChatSession\n(session.py)" as Session <<session>>
@@ -57,9 +57,10 @@ else key not in cache
    Store --> Session : default value
 end
 note right of Session
-  Settings are captured once
-  at workstream creation.
-  Not re-read on every turn.
+  Most session settings are captured once
+  at workstream creation. Documented live readers
+  (including model.auth_fail_closed per mint)
+  apply immediately.
 end note

 == Phase 3: Admin API — List / Schema ==
@@ -128,8 +129,9 @@ Store -> Storage : get_system_settings_bulk(node_id)
 Storage --> Store : all settings
 Store -> Store : rebuild cache,\nswap atomically,\nincrement _version
 note right
-  Existing sessions: unchanged
-  (frozen at creation time).
+  Most existing-session settings are unchanged
+  (frozen at creation time); documented
+  live readers apply immediately.
  New sessions: pick up
  updated values immediately.
 end note
@@ -88,7 +88,7 @@
    <rect x="300" y="151" width="160" height="2" fill="#161b22"/>
    <text x="380" y="178" text-anchor="middle" fill="#e6edf3" font-size="11" font-weight="600">Console</text>
    <line x1="318" y1="190" x2="442" y2="190" stroke="#30363d" stroke-width="1"/>
-    <text x="380" y="208" text-anchor="middle" fill="#8b949e" font-size="9">hash-ring router</text>
+    <text x="380" y="208" text-anchor="middle" fill="#8b949e" font-size="9">FNV-1a rendezvous router</text>
    <text x="380" y="223" text-anchor="middle" fill="#8b949e" font-size="9">cluster dashboard</text>
    <text x="380" y="238" text-anchor="middle" fill="#8b949e" font-size="9">reverse proxy</text>
    <line x1="318" y1="250" x2="442" y2="250" stroke="#30363d" stroke-width="1"/>
@@ -115,7 +115,7 @@
    <rect x="608" y="162" width="184" height="28" rx="3" fill="#1c2128" stroke="#30363d" stroke-width="1"/>
    <text x="700" y="180" text-anchor="middle" fill="#8b949e" font-size="9">turnstone-server :8080</text>
    <!-- Tools label -->
-    <text x="700" y="206" text-anchor="middle" fill="#484f58" font-size="8">19 tools + MCP</text>
+    <text x="700" y="206" text-anchor="middle" fill="#484f58" font-size="8">model lanes + tools / MCP</text>
  </g>

  <!-- Node B -->
@@ -130,7 +130,7 @@
    <rect x="608" y="282" width="184" height="28" rx="3" fill="#1c2128" stroke="#30363d" stroke-width="1"/>
    <text x="700" y="300" text-anchor="middle" fill="#8b949e" font-size="9">turnstone-server :8080</text>
    <!-- Tools label -->
-    <text x="700" y="326" text-anchor="middle" fill="#484f58" font-size="8">19 tools + MCP</text>
+    <text x="700" y="326" text-anchor="middle" fill="#484f58" font-size="8">model lanes + tools / MCP</text>
  </g>

  <!-- ==================== LLM PROVIDERS ==================== -->
@@ -171,7 +171,7 @@
    <rect x="590" y="450" width="220" height="5" fill="#bc8cff"/>
    <rect x="590" y="453" width="220" height="2" fill="#161b22"/>
    <text x="700" y="476" text-anchor="middle" fill="#e6edf3" font-size="11" font-weight="600">PostgreSQL / SQLite</text>
-    <text x="700" y="492" text-anchor="middle" fill="#8b949e" font-size="9">conversations, memory, auth</text>
+    <text x="700" y="492" text-anchor="middle" fill="#8b949e" font-size="9">workstreams, turns, config, auth</text>
  </g>
  <text x="700" y="444" text-anchor="middle" fill="#bc8cff" font-size="9" font-weight="600" letter-spacing="2">STORAGE</text>

@@ -239,7 +239,7 @@
  <!-- Routing rules at bottom, left-aligned -->
  <text x="44" y="456" fill="#30363d" font-size="9" font-weight="600" letter-spacing="1">ROUTING</text>
  <circle cx="44" cy="474" r="3" fill="#3fb950" opacity="0.6"/>
-  <text x="54" y="477" fill="#484f58" font-size="9">control plane: client &#x2192; console &#x2192; server node (hash-ring bucket lookup)</text>
+  <text x="54" y="477" fill="#484f58" font-size="9">control plane: client &#x2192; console &#x2192; server node (FNV-1a rendezvous placement)</text>
  <circle cx="44" cy="494" r="3" fill="#58a6ff" opacity="0.6"/>
  <text x="54" y="497" fill="#484f58" font-size="9">data plane: client &#x2192; server node (direct SSE, node_url from create response)</text>
  <circle cx="44" cy="514" r="3" fill="#f47067" opacity="0.6"/>
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e2bfdf968e96f3720ed58674103288e2f57e9c056f5c479a57f37a849f3e69c
+size 821878
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:881a8b9bce67b5af9a52d5e50deaa72351cd99c76f18aad5caeb2b61131ca1af
-size 119798
+oid sha256:b8c1460440784f07e30afea32d4ee17687627df46a24003d761ad79c2676a361
+size 169499
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:95dd5ebc899a1261d516686a5aa3319a7f45015d411302825fa28afbfc82e1ce
-size 326766
+oid sha256:66847ccdf10ef2bd04e93bc0d3924a56ce28462ec9e76a383b53aee4500755e8
+size 631799
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9857db23fe3c4316d492073aac69c7e7558b1abe3b95ad7756d4a5933bd0ece7
-size 620214
+oid sha256:c74e99c530c3a8af9ab35b1e4d8c4fef0ea35c0c04cc35da7cf3588e71382057
+size 661175
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d9c7769a600c38e6387390e6c42db8152e0f80c31d17b2218f7f636b71c7b868
-size 355459
+oid sha256:6261604cc8b75878a8704308929ea64d121cf26547019543fbe1b21cbe700415
+size 189791
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:23ca090b5656baaf70820cbe4ab6c27f0a3a02e18b4db0695614cf9489c23980
-size 281440
+oid sha256:1b3b7b745f6006ee73d4b31fa598faa0d71ffb74ce349ab08eb3ce09ded506c3
+size 266294
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:04d2069a9b5155ad1e7d842147fd78535ad9106d6856520439c33a9868a47499
-size 156694
+oid sha256:33dddd8cd8b53fa464cc8a4c899896fee32035d63e0669fe327969e3356349c7
+size 329815
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a872556d111185f4531d1b68ee892b4ce5042d7ccf277e2cad08beb6932c9803
-size 191144
+oid sha256:16e3f3bfa0a6af637f7a9fb6765d594eb598428679c88a429c096c3dbae931e4
+size 181185
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e7c3e40c10425d721f833390ae3531c09af501157fd3142531ba4eba86ff719d
-size 197112
+oid sha256:59f14b835665244f3d32981b6c1ac4c4380393a83cc519e271831622aa3f261a
+size 197433
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b047cdc318c505f0f0895a65e14c5cc7552716053055cca57fa0a77db150e618
-size 255458
+oid sha256:1a510eeaabf4ed8dab3b268c8f6bb5b7fef629a664361f7fb5614a6b498db36e
+size 294415
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:af5ab3126bf685afe68e24bc4b0ed97371d0ebdb77bf4d76c0331ab120580cc0
-size 248809
+oid sha256:ea86c6b6c68ed96f6fd18543d2e7a873f4715332cc3a7d4df167392f668a2de7
+size 232403
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ae4f79fb22600106f8cb0af4ba5586bb26ea5d57e27ef382fdc59b6549fdbd21
-size 415473
+oid sha256:edf02b97e1e1287ebba9e74b9858474dcda42e5542656e505ea133a5b2416f47
+size 402992
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:96176a09e65e90dadc32d5e9ed778423842be89204d2cf382225f53a90cfaf01
-size 258547
+oid sha256:aa9ca9a367c79159a26d1ec544b20fc7118a49082d72f7ee0edbaa85608d49fc
+size 238991
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:79a690c466a5d6f6d4292d78a27b9474e9e9c1373fa80e17dfe37700238c8af8
-size 382508
+oid sha256:636a6b2fc1075e4863421e68b99efe7f6f6f62cedbcff36ef0934c055f39fd46
+size 281161
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c89628ed917dfd576c1af75c68fe5fed9beadaaee9dcea7aa7a1643867c4f1b9
-size 344323
+oid sha256:137d6c91a34695c820d8b0a33fd753e79165604aa92bf2ac8480d3744b2ef844
+size 305199
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:06fe076f0835a891e00afc804fd1805196ebde9fc0d34998c7873e87287f982b
-size 346887
+oid sha256:0455e0dec36ebb8bcfdadcf327a1dd24ddbdcdd8df211c08918180842497727e
+size 318681
@@ -102,13 +102,26 @@ the cluster runs mTLS; harmless otherwise), and `TURNSTONE_SEARXNG_URL` points
 are the dev-stack defaults — match whatever you set in `.env` if you changed them.

 To let a server on a **different** machine join, start the stack with
-`TURNSTONE_HOST_IP=<this host's LAN IP>` — that binds PostgreSQL, the console
-ACME endpoint, and SearxNG to that interface. Then on the remote box set the
-three URLs above to that IP, and set `TURNSTONE_ADVERTISE_URL` to the **remote**
-box's own IP (the address the console dials back). **Set a strong
-`POSTGRES_PASSWORD` first** — `TURNSTONE_HOST_IP` exposes the database (and every
-user account + API-token hash in it), the console API, and the unauthenticated
-SearxNG to your network.
+both `TURNSTONE_HOST_IP=<this host's LAN IP>` and
+`TURNSTONE_ACME_EXTERNAL_URL=http://<this host's LAN IP>:8090/acme`. The first
+binds PostgreSQL, the console ACME endpoint, and SearxNG to that interface; the
+second makes every URL in the ACME directory routable from the remote node (the
+full value must include the `/acme` mount). Set the same
+`TURNSTONE_ACME_EXTERNAL_URL` on the remote node so its authenticated ACME
+client can pin that credential destination. Set `TURNSTONE_CONSOLE_URL` and
+`TURNSTONE_SEARXNG_URL` to the compose host's IP, but set
+`TURNSTONE_ADVERTISE_URL=http://192.0.2.10:8080` to the **remote** box's own
+address. A resolvable DNS name works too. IPv6 literals must be bracketed in
+URLs, for example `http://[2001:db8::10]:8080`; Turnstone enrolls literal
+addresses as IP SANs rather than numeric DNS SANs.
+
+Use a trusted LAN or VPN address and firewall `:8090` to enrolling nodes. ACME
+signing routes require a dedicated short-lived service JWT, but direct bootstrap
+is still plain HTTP/TOFU: a bearer token provides authentication, not transport
+confidentiality or protection from an active on-path attacker. **Set a strong
+`POSTGRES_PASSWORD` first** — `TURNSTONE_HOST_IP` also exposes the database (and
+every user account + API-token hash in it), the console API, and the
+unauthenticated SearxNG to your network.

 To run the bare-metal node as a hardened, persistent service instead of by hand,
 use the systemd units in [`deploy/systemd/`](../deploy/systemd/).
@@ -148,6 +161,11 @@ certs via the console's ACME endpoint:
 docker compose -f turnstone/deploy/compose.yaml -f deploy/docker-compose.tls.yml up
 ```

+The overlay publishes the console's plain-HTTP bootstrap/API port on
+`TURNSTONE_CONSOLE_HTTP_BIND` (default `127.0.0.1`). For a cross-host node, set
+that to a trusted LAN/VPN address, set `TURNSTONE_ACME_EXTERNAL_URL` to the same
+address plus `/acme`, and firewall the port to enrolling nodes.
+
 See [tls.md](tls.md) for details.

 ## Configuration
@@ -186,6 +204,12 @@ overrides.
 > put [PgBouncer](pgbouncer.md) (transaction pooling) between turnstone and
 > PostgreSQL.

+> **Lifecycle upgrade:** the release that introduces hidden deferred-create
+> reservations must be deployed as a coordinated cohort across every server
+> sharing PostgreSQL; older processes do not understand `state='creating'`.
+> Drain create traffic until the cohort is upgraded. See
+> [PgBouncer: deferred workstream creation](pgbouncer.md#upgrade-note-deferred-workstream-creation).
+
 ### Ports

 Both stacks publish Caddy (dashboard) and PostgreSQL; the dev stack additionally
@@ -200,6 +224,8 @@ Caddy or proxied by the console:
 | `POSTGRES_PORT` | `5432` | Host port for PostgreSQL (for bare-metal joins) |
 | `SEARXNG_API_PORT` | `8081` | Host port for the SearxNG API a bare-metal node's `web_search` dials (dev stack) |
 | `TURNSTONE_HOST_IP` | `127.0.0.1` | Interface PostgreSQL, the console ACME endpoint, and SearxNG bind on (dev stack). Set to this host's LAN IP so a bare-metal node on **another machine** can reach them — set a strong `POSTGRES_PASSWORD` first (it also exposes the DB and the unauthenticated SearxNG to your network). |
+| `TURNSTONE_CONSOLE_HTTP_BIND` | `127.0.0.1` | Production TLS-overlay interface for the console's plain-HTTP bootstrap/API listener. Use only a trusted LAN/VPN address and firewall it to enrolling nodes. |
+| `TURNSTONE_ACME_EXTERNAL_URL` | request-derived | Canonical externally reachable ACME responder base, including the final `/acme` mount (for example `http://192.0.2.1:8090/acme`). Set it on the console and clients for cross-host mTLS: the console advertises it, while clients pin it as an allowed enrollment-JWT destination. A reverse-proxy prefix is supported only when the proxy maps it to Turnstone's internal `/acme` mount. |
 | `POSTGRES_BIND` | `127.0.0.1` | Production stack (`turnstone/deploy/compose.yaml`) only: interface PostgreSQL binds on; set to the host's LAN IP for remote joins. |

 ### Channel gateway
@@ -252,6 +278,7 @@ interface, or anyone who can reach it can search through your instance.
 | Variable | Default | Description |
 |----------|---------|-------------|
 | `WORKSPACE_MOUNT` | empty volume | Host directory bind-mounted at `/workspace` for the model to read/write |
+| `TURNSTONE_WORKSPACE` | `/workspace` (image env) | Directory named as the user's workspace in the model's tool descriptions; informational only — see [Working directory](#working-directory) |
 | `SKIP_PERMISSIONS` | — | Set to any value to auto-approve all tool calls (dev only) |
 | `MCP_CONFIG` | — | Path to an MCP server config file |
 | `TURNSTONE_IMAGE_TAG` | `latest` | ghcr.io image tag — production stack |
@@ -276,6 +303,35 @@ docker compose build --no-cache # rebuild from scratch
 | `workspace` | `/workspace` (unless `WORKSPACE_MOUNT` is set) |
 | `caddy-data` / `caddy-config` | Caddy's local CA and config (dev stack) |

+## Working directory
+
+Node processes run with `/data` as their working directory (the image's
+`WORKDIR`), and that is where the model's shell commands execute and
+relative file paths resolve — **not** `/workspace`. The shell and file
+tool descriptions state both paths (the working directory, and the
+workspace named by `TURNSTONE_WORKSPACE`), so the model knows to look in
+`/workspace` for your files without being told each session.
+
+To make tools start inside the mount instead, override the working
+directory on the node services:
+
+```yaml
+services:
+  turnstone-node:
+    working_dir: /workspace
+```
+
+Two caveats before overriding:
+
+- **SQLite fallback**: when a node runs without PostgreSQL, its fallback
+  database `.turnstone.db` is created in the process working directory.
+  Changing `working_dir` on an existing SQLite-fallback deployment makes
+  the node create a fresh database inside the mount and your prior state
+  appears lost (it is still in the `turnstone-data` volume under `/data`).
+  The stock compose stacks use PostgreSQL and are unaffected.
+- Migrations (`entrypoint.sh`) run in the same working directory, so the
+  same SQLite caveat applies to them.
+
 ## Cleanup

 ```bash
@@ -177,7 +177,8 @@ Runs a complete multi-turn conversation:

 1. Appends the user message.
 2. Checks `_cancelled` event — stops if set (timeout cleanup).
-3. Calls the model API (non-streaming).
+3. Calls the model through the production streaming provider path and drains
+   the result.
 4. If tool calls are returned, executes them (with stdout suppressed) and
   logs each call to `self.tool_call_log`.
 5. Repeats up to `max_turns` or until the model responds without tool calls.
@@ -190,14 +191,15 @@ Parallel tool calls are capped at 10 per turn to prevent degenerate repetition.

 Each test runs in a `ThreadPoolExecutor(max_workers=1)` with a per-test
 timeout (`--test-timeout`). Each attempt gets its own `OpenAI` client with
-a matching httpx read timeout. On timeout, three layers of defense prevent
-zombie connections:
+a matching per-read HTTP transport timeout. Because a trickling stream can
+continually reset that read timeout, three layers bound the harness and stop
+follow-on work:

-1. **httpx timeout**: Per-request read timeout aborts the HTTP call and
-   releases the server slot.
+1. **Executor wall clock**: The harness stops waiting after `--test-timeout`.
 2. **`_cancelled` event**: Prevents the orphan thread from starting new turns.
-3. **`run_client.close()`**: Closes the connection pool to abort any
-   in-flight request.
+3. **`run_client.close()`**: Retires the connection pool and prevents reuse.
+   HTTPX2 does not promise that cross-thread client closure immediately aborts
+   an active body read; that read unwinds on its next wire event or read timeout.

 ### Retry Logic

@@ -214,7 +216,8 @@ Each test case runs in isolation:
 1. A fresh temp directory is created.
 2. Setup files are written to the temp directory.
 3. The working directory is changed to the temp directory.
-4. A per-attempt `OpenAI` client is created with httpx timeout matching `--test-timeout`.
+4. A per-attempt `OpenAI` client is created with an HTTP transport timeout
+   matching `--test-timeout`.
 5. A new `HeadlessSession` is created with the current developer prompt.
 6. `send_headless()` runs the user prompt through the conversation loop.
 7. The tool log is scored against expected actions.
@@ -16,15 +16,18 @@ The permission model has two layers:
 2. **Permissions** (granular) — named permission strings checked per-endpoint by
   `require_permission()`.

-**Built-in roles** (seeded by migration 008):
+**Built-in roles** (seeded by migration 008 and extended by later feature
+migrations):

 | Role | Permissions |
 |------|-------------|
-| admin | read, write, approve, admin.users, admin.roles, admin.orgs, admin.policies, admin.skills, admin.audit, admin.usage, admin.schedules, admin.watches, tools.approve, workstreams.create, workstreams.close |
-| operator | read, write, workstreams.create, workstreams.close |
+| admin | Admin-default baseline: ordinary admin, lifecycle, tool-approval, coordinator, project, and persona capabilities. Explicit opt-in capabilities such as `model.skills.write` remain ungranted. |
+| operator | read, write, workstreams.create, workstreams.close, conversation.modify |
 | viewer | read |

-Custom roles can be created with any subset of the valid permissions.
+Custom roles can be created with any subset of the valid permissions. Built-in
+role permission overrides can grant or revoke individual capabilities, so the
+admin console is authoritative for the effective set on a deployment.
 The `persona.create` / `persona.read` / `persona.write` family gates
 persona administration; migration `063` seeds all three onto
 `builtin-admin`, and any role can be granted them through the standard
@@ -67,7 +70,7 @@ etc.) since workstream templates were merged into the skills system in v0.8.0.
  workstreams, concatenated in alphabetical order by name. Use name prefixes
  (e.g. `01-safety`, `02-style`) to control ordering.
 - **Explicit selection**: `--skill <name>` CLI flag, `skill` field on
-  `POST /v1/api/workstreams/new`, console creation modal dropdown, scheduled task
+  `POST /v1/api/workstreams/new`, console launcher dropdown, scheduled task
  config, and channel adapter config. An explicit skill *replaces* defaults.
 - **Variables**: Three built-in placeholders resolved at load time:
  `{{model}}` (active model name), `{{ws_id}}` (workstream ID),
@@ -17,7 +17,9 @@ evaluation:
   read-only tool access. Runs on a daemon thread and delivers its verdict
   progressively.

-The verdict is purely advisory -- the user always makes the final decision.
+The verdict is advisory by default. The opt-in Smart Approvals mode can use a
+completed, high-confidence LLM `approve` verdict to make the decision
+automatically under the fail-closed rules below.

 The heuristic verdict is attached to the `approve_request` SSE event immediately.
 The LLM verdict arrives later via an `intent_verdict` SSE event, allowing the
@@ -28,55 +30,77 @@ persisted to the `intent_verdicts` table for audit and future calibration.

 ## Configuration

-### config.toml
+### Server and console

-```toml
-[judge]
-enabled = true
-model = ""                    # empty = same as session model
-provider = ""                 # empty = same as session provider
-base_url = ""
-api_key = ""
-smart_approvals = false       # auto-approve high-confidence "approve" LLM verdicts (opt-in)
-confidence_threshold = 0.95   # Smart Approvals auto-approve bar (LLM recommendation=approve)
-max_context_ratio = 0.5       # max % of judge context window for history
-timeout = 120.0               # seconds (generous for local models)
-read_only_tools = true        # judge can use read_file/list_directory
-cancel_on_approval = false    # stop judging remaining tool calls once user decides
+Server and console workstreams read database-backed `judge.*` settings from
+the settings registry. Edit them at **Admin → Judge** or through the admin
+settings API; changes take effect for the next judge batch without a restart.
+The principal settings are:
+
+```text
+judge.enabled = true
+judge.model = ""                    # empty = same alias as the session
+judge.smart_approvals = false       # opt-in automatic approval
+judge.confidence_threshold = 0.95   # Smart Approvals confidence bar
+judge.max_context_ratio = 0.5       # fraction of judge context used for history
+judge.timeout = 120.0               # per judge turn and Smart Approvals wait
+judge.parallel_evaluations = 1      # concurrent calls within one batch, 1-16
+judge.read_only_tools = true        # permit read_file/list_directory evidence
+judge.cancel_on_approval = false    # stop unfinished calls when the gate resolves
 ```

+`parallel_evaluations = 1` preserves serial evaluation. Raising it reduces the
+latency of wide tool-call batches. The selected judge model alias's
+`max_concurrency` remains the process-wide generation ceiling, so it can reduce
+the actual overlap across judge batches and other roles using that alias.
+
 ### Smart Approvals

-With `smart_approvals = true` (off by default) a tool call is approved
-automatically — no operator prompt — when the intent judge's **LLM** verdict
-recommends `approve` with confidence at or above `confidence_threshold`. Every
-other outcome still reaches a human: `review` / `deny` recommendations,
-confidence below the threshold, judge errors or timeouts (`llm_fallback`), and
-any call the deterministic heuristic rules explicitly flagged `deny` or
-`critical`. That heuristic floor blocks only those explicit danger verdicts — it
-is **not** a general "never lower the heuristic" rule: the heuristic's default
-for an unmatched tool is `review`, and letting a confident LLM `approve` upgrade
-a `review` is exactly what Smart Approvals is for. Only `deny` / `critical`
-findings are off-limits to auto-approval. Requires the judge to be enabled;
-auto-approved calls are tagged `smart_approval` in the dashboard and audit trail.
-Smart Approvals applies to the web and coordinator surfaces, not the interactive
-CLI.
+With `smart_approvals = true` (off by default), a pending batch is approved
+automatically — no operator prompt — only when **every** call has a completed
+LLM verdict recommending `approve` at or above `confidence_threshold`. The
+decision is batch-atomic: one uncertain sibling sends the entire parallel batch
+to a human rather than executing the safe-looking subset piecemeal.

-All fields are optional. The judge is enabled by default; use `enabled = false`
-(or `--no-judge` on the command line) to disable it.
+Every other outcome reaches a human: `review` / `deny` recommendations,
+confidence below the threshold, judge errors or timeouts (`llm_fallback`), a
+missing/duplicate call ID, an unjudged sibling, and any call the deterministic
+heuristic rules explicitly flagged `deny` or `critical`. That heuristic floor
+blocks only explicit danger verdicts — it is **not** a general "never lower the
+heuristic" rule. The heuristic's default for an unmatched tool is `review`, and
+letting a confident LLM upgrade that default is the feature's purpose.
+
+The Smart Approvals enabled flag, threshold, and bounded verdict wait are
+captured as one immutable snapshot when each gate batch starts. A settings
+reload takes effect on the next batch, while concurrent main-loop and
+task-agent gates cannot mix fields from different reload generations. Stop
+wakes a batch still waiting for verdicts and is linearized against the final
+auto-approval commit: if Stop wins, no `smart_approval` decision or audit row
+is recorded for tools that did not cross the gate.
+
+The verdict wait is capped by the snapshot's `judge.timeout`; the judge may
+continue evaluating advisory verdicts after that gate falls back to a human.
+
+Requires the judge to be enabled. Auto-approved calls are tagged
+`smart_approval` in the dashboard and audit trail. Smart Approvals applies to
+the web and coordinator surfaces, not the interactive CLI.
+
+The judge is enabled by default. Disable `judge.enabled` in the admin Judge
+settings, or use `--no-judge` in the interactive CLI.

 ### CLI flags

 ```
 --judge / --no-judge           Enable/disable (default: enabled)
--judge-model MODEL            Model for judge
--judge-provider PROVIDER      Provider for judge
+--judge-model ALIAS            Registered model alias for judge
 --judge-timeout SECONDS        LLM judge timeout (default: 120)
+--judge-parallel-evaluations N Concurrent evaluations per batch, 1-16 (default: 1)
 --judge-confidence FLOAT       Confidence threshold, 0-1 (default: 0.95)
 ```

-(Smart Approvals is configured via `[judge] smart_approvals` / the admin Judge
-settings, not a CLI flag — the interactive CLI prompts for approval directly.)
+The same five values can be placed in the CLI's `config.toml` `[judge]`
+section. Smart Approvals is configured through the server/console admin Judge
+settings, not a CLI flag—the interactive CLI prompts for approval directly.

 CLI flags override `config.toml` values.

@@ -87,20 +111,19 @@ CLI flags override `config.toml` values.
 - **Default (self-consistency)**: When `model` is empty, the session model
  evaluates its own tool calls. Research shows self-consistency achieves
  comparable accuracy to multi-agent debate at a fraction of the cost.
- **Cross-model**: Use a different model for the judge (e.g. local model for
-  the session, commercial model for the judge). Set `model` and `provider`
-  in the `[judge]` config section, or use `--judge-model` / `--judge-provider`
-  CLI flags.
- **Cross-provider**: When both `model` and `provider` are set, the judge
-  creates its own LLM client. You can optionally specify `base_url` and
-  `api_key` for non-default endpoints.
- **Google models**: The judge supports `google` as a provider. Note that
-  read-only tools are disabled for Google models (the Gemini API requires
-  `thought_signature` in tool call round-trips which the judge's normalized
-  format does not preserve).
+- **Cross-model**: Register the desired model in the Models tab, then set
+  `judge.model` to that alias (or pass `--judge-model ALIAS` to the CLI).
+- **Cross-provider**: A model alias carries its provider, endpoint, and
+  credential configuration together, so a judge alias may use a different
+  provider from the session without separate judge connection settings.
+- **Google models**: The judge supports `google` aliases, including read-only
+  evidence tools. Provider-native reasoning state such as Gemini
+  `thought_signature` stays attached to the pinned model lane across evidence
+  turns.

-The judge creates a fresh HTTP client for each evaluation run and closes it
-when done, avoiding stale connection issues across runs.
+The judge creates one fresh HTTP client per active batch worker and closes each
+when that worker finishes, avoiding cross-thread client sharing and stale
+connections across runs.

 If the LLM judge fails or returns no verdict, a fallback verdict with tier
 `llm_fallback` is delivered via the callback, ensuring the UI always receives
@@ -232,17 +255,25 @@ calls for approval, it calls `_evaluate_intent()` which:
 4. Attaches each heuristic verdict to its item as `_heuristic_verdict`
 5. The daemon thread runs the LLM judge and delivers results via `ui.on_intent_verdict()`

-The daemon evaluates items sequentially, so a large parallel batch can outlive
-its approval gate. With `cancel_on_approval = false` (the default) the daemon
-runs every item to completion: verdicts that land after the operator decided
-still stream to the UI and persist, stamped with the decision. The daemon is
-aborted only when the next tool batch supersedes it or the session closes —
-then each unfinished item degrades to an `llm_fallback` verdict. With
-`cancel_on_approval = true` the abort additionally fires the moment the gate
-resolves, trading verdict completeness for inference savings — recommended
-when the judge shares a single local inference backend with the session model,
-where a large batch's remaining judge calls would otherwise compete with the
-next turn's completion.
+The daemon coordinates up to `parallel_evaluations` independent workers for
+one batch. Completed verdicts stream to the UI as workers finish, and every call
+still receives exactly one LLM or `llm_fallback` verdict. The default of 1 keeps
+the historical serial behavior; a higher value collapses a wide batch toward
+`ceil(batch size / workers)` judge-call intervals. A smaller positive model
+alias capacity also bounds the worker count, avoiding surplus threads queued at
+the same admission gate.
+
+With `cancel_on_approval = false` (the default) the daemon runs every item to
+completion: verdicts that land after the operator decided still stream to the
+UI and persist, stamped with the decision. A newer main-loop batch, session
+close, or explicit Stop retires the old generation; unfinished items degrade
+to `llm_fallback` verdicts. A judge/model binding or parallelism edit prevents
+reuse on the next batch, while already-started calls stay pinned to the binding
+and worker count they began with. With `cancel_on_approval = true`, an ordinary
+gate decision additionally aborts unfinished work, trading verdict completeness
+for inference savings—recommended when the judge shares a single local
+inference backend with the session model. Explicit Stop always cancels every
+live judge generation, regardless of this preference.

 Verdicts that arrive after a *newer batch* has replaced the judge generation
 are withheld from the live surfaces (a reused call_id must never ride a stale
@@ -258,6 +289,15 @@ siblings would otherwise make each other's verdicts look stale); per-cycle
 generation checks enforce staleness instead, and `judge.cancel_on_approval`
 fires per gate exactly like the main loop.

+Several parallel task agents can therefore leave several approval cycles live
+on one workstream. Each cycle owns its event, result, verdict set, and
+`cycle_id`; a decision targets exactly one cycle by `cycle_id` or member
+`call_id` (selector-less legacy clients resolve the oldest). Workstream Stop or
+close performs a workstream-wide denial sweep over all cycles belonging to the
+cancelled operation. A force-cancel successor's newly registered cycle carries
+a fresh operation witness and is not accidentally denied by the predecessor's
+late sweep.
+
 ---

 ## Storage and Audit
@@ -415,13 +455,12 @@ Redaction types: `api_key`, `private_key`, `password`, `secret`.

 ### Configuration

-```toml
-[judge]
-output_guard = true    # enable output evaluation (default)
-redact_secrets = true  # auto-redact detected credentials (default)
+```text
+judge.output_guard = true    # enable output evaluation (default)
+judge.redact_secrets = true  # auto-redact detected credentials (default)
 ```

-Configurable at runtime via the admin Settings tab.
+Configure both at runtime through the admin Judge settings.

 ### Merge semantics (heuristic + LLM judge)

@@ -111,6 +111,7 @@ In the admin MCP form, choose **Sign-in passthrough** and set **Audience** (requ
 The captured credential is a single per-user secret that can mint for every `oauth_obo` server, so treat it like any long-lived credential:

 - **Cut off one user:** unlink their OIDC identity in the admin console (**Users → OIDC identities → delete**). This revokes the captured credential **and** purges their minted cache rows, so future mints fail and cached tokens are dropped. (Warmed in-memory sessions on server nodes self-expire at the access-token TTL; there is no cross-node per-user session-kill.) Removing the user's access at the IdP is the authoritative cut-off.
+- The same unlink also purges that user's synthetic `__model_obo__:` gateway-token rows and requests eviction from every registered host's in-process mint memo. Shared `entra_app` model tokens live under the `__app__` pseudo-user and are intentionally not user-deprovisioned; revoking the app credential prevents new mints, while a cached app bearer lasts until `expires_at`.
 - **Flush a server's minted tokens** (e.g. after narrowing its audience): the server row's **flush cache** action drops all users' cached tokens for that server. This is **not** a revocation — users re-mint on next use from their still-valid sign-in. It is surfaced honestly (audit `mcp_server.oauth.obo_cache_flushed`, response `effect: cache_flush_remints`) so it is never mistaken for cutting access.
 - Per-server revocation in the `oauth_user` sense does not exist for `oauth_obo` — the credential is issuer-scoped and IdP-governed. Revoke at the IdP.

@@ -128,7 +129,7 @@ The captured credential is a single per-user secret that can mint for every `oau

 4. **Step-up scope**: when a tool call hits `403` with `WWW-Authenticate: error="insufficient_scope"`, Turnstone emits `mcp_insufficient_scope` with the parsed scope set; the dashboard offers a "Connect with additional scopes" affordance that opens `/v1/api/mcp/oauth/start?server=<name>&scopes=<extra>` so the union of original + new scopes flows into the AS authorize request.

-5. **User revoke** (settings modal): `DELETE /v1/api/mcp/oauth/connections/{server_name}` runs the authoritative local delete + best-effort RFC 7009 upstream revoke (fire-and-forget, capped at 256 concurrent in-flight tasks). `oauth_obo` servers are excluded: their rows are mint cache, not consents — deleting one only forces a re-mint — so the connections list hides them and the endpoint refuses them with `409` (revocation for sign-in passthrough happens at the identity layer: unlink the identity or revoke at the IdP).
+5. **User revoke** (settings modal): `DELETE /v1/api/mcp/oauth/connections/{server_name}` runs the authoritative local delete + best-effort RFC 7009 upstream revoke (fire-and-forget, capped at 256 concurrent in-flight tasks). `oauth_obo` servers and synthetic model-auth rows are excluded: their rows are mint caches, not consents — deleting one only forces a re-mint — so the connections list hides them and the endpoint refuses them with `409` (revocation for sign-in passthrough happens at the identity layer: unlink the identity or revoke at the IdP).

 6. **Admin bulk-revoke** (Phase 9): `POST /v1/api/admin/mcp-servers/{name}/bulk-revoke` drops every user's token for the server. Upstream RFC 7009 revoke is intentionally **not** attempted in bulk (avoids N upstream HTTP calls per admin click); tokens at the AS expire naturally. Use the per-user revoke endpoint if you need guaranteed upstream invalidation.

@@ -20,7 +20,7 @@ Each memory has three dimensions:
 | Type        | Purpose                                                    |
 |-------------|------------------------------------------------------------|
 | `user`      | User preferences, conventions, working style               |
-| `project`   | Project-specific knowledge, architecture, patterns         |
+| `general`   | General knowledge, architecture, patterns                  |
 | `feedback`  | Corrections, lessons learned, things to avoid              |
 | `reference` | Reference material, documentation, specifications         |

@@ -31,25 +31,37 @@ Each memory has three dimensions:
 | `global`      | Visible to all workstreams and users                            |
 | `workstream`  | Visible only within the originating workstream                  |
 | `user`        | Follows the authenticated user across workstreams               |
-| `coordinator` | Coordinator sessions only; follows the user across coordinators |
+| `coordinator` | Coordinator sessions only; follows the acting user              |
+| `project`     | Shared by workstreams attached to one active project             |

 A memory's identity is the tuple `(name, scope, scope_id)`. Saving a memory
 with the same identity upserts -- updating content while preserving the ID.

-### Coordinator scope
+### Inherited target and coordinator scope

-Coordinator sessions are isolated to a single scope: `coordinator`, keyed by
-the coordinator's creator `user_id`. It is durable -- every coordinator
-session the same user runs (including concurrent ones) shares one
-orchestration namespace, so procedures and lessons survive close/reopen.
+Name-based operations use one inherited target when `scope` is omitted:
+
+- An attached active project selects `project` for `save`, `get`, and
+  `delete`.
+- Read-only project access permits `get`, but `save` and `delete` fail. They do
+  not fall back to a broader namespace.
+- Without a project, interactive sessions select `global`; coordinator
+  sessions select `coordinator`.
+
+A valid explicit scope selects exactly that scope. `search` and `list` are the
+only actions that span every visible scope when `scope` is omitted.
+
+Each coordinator's private `coordinator` namespace is keyed by the acting
+user's `user_id`. It is durable -- every coordinator session that user runs
+(including concurrent ones) shares one orchestration namespace, so procedures
+and lessons survive close/reopen.

 Isolation is bidirectional and enforced by session kind, not by secrecy of
 the scope id:

- A coordinator session can read and write **only** `coordinator`-scope rows.
-  It never sees `global`/`workstream`/`user` memories, so content written by
-  interactive sessions (which routinely ingest untrusted MCP/attachment
-  output) cannot reach a coordinator's system message.
+- A coordinator session sees its acting user's `coordinator` scope and, when
+  attached, the shared `project` scope. It never sees
+  `global`/`workstream`/`user` memories.
 - Interactive sessions -- including a coordinator's own children, which share
  its `user_id` -- are rejected from the `coordinator` scope on every memory
  action. Children cannot plant rows the parent coordinator would read.
@@ -64,12 +76,13 @@ coordinator cannot be constructed, so the scope id is always a real user.

 On every conversation turn, the system:

-1. Fetches up to `fetch_limit` memories visible in the current scope
-2. Extracts context from the last 3 user messages
-3. Scores memories against that context using a BM25 index
-4. Injects the top `relevance_k` memories into the system message as
+1. Resolves the acting principal and their live project access
+2. Fetches up to `fetch_limit` memories across that visibility envelope
+3. Extracts context from the last 3 user messages
+4. Scores memories against that context using a BM25 index
+5. Injects the top `relevance_k` memories into the system message as
   `<memories>` XML tags
-5. Appends a hint telling the model how many memories are in scope
+6. Appends a hint telling the model how many memories are in scope

 This means the model always has its most relevant memories available without
 explicit recall -- but can still use `memory(action='search')` for deeper
@@ -107,19 +120,23 @@ All fields are optional. Defaults are shown above.

 ## Tool Usage

-The `memory` tool supports four actions:
+The `memory` tool supports five actions:

 ### save

 Store or update a memory.

+Every save is a complete write for the relevance summary: `description` must
+be supplied and contain non-whitespace text on both creation and update.
+Content-only updates are rejected.
+
 ```json
 {
  "action": "save",
  "name": "project_architecture",
  "content": "The project uses a hexagonal architecture with...",
  "description": "Core architecture patterns",
-  "type": "project",
+  "type": "general",
  "scope": "global"
 }
 ```
@@ -128,9 +145,26 @@ Store or update a memory.
 |---------------|----------|-------------|------------------------------------------|
 | `name`        | yes      | --          | Snake_case identifier (max 256 chars)    |
 | `content`     | yes      | --          | Memory content (max `max_content` chars) |
-| `description` | no       | `""`        | Short description for relevance matching |
-| `type`        | no       | `"project"` | One of: user, project, feedback, reference |
-| `scope`       | no       | `"global"`  | One of: global, workstream, user         |
+| `description` | yes      | --          | Non-empty relevance summary, required on create and update |
+| `type`        | no       | `"general"` | One of: user, general, feedback, reference |
+| `scope`       | no       | inherited   | Kind-valid scope; see inherited target above |
+
+### get
+
+Retrieve the full content of one memory by name.
+
+```json
+{
+  "action": "get",
+  "name": "project_architecture",
+  "scope": "project"
+}
+```
+
+| Parameter | Required | Default   | Description                |
+|-----------|----------|-----------|----------------------------|
+| `name`    | yes      | --        | Memory name to retrieve    |
+| `scope`   | no       | inherited | Exact scope to query       |

 ### search

@@ -140,7 +174,7 @@ Find memories by query (BM25 full-text search).
 {
  "action": "search",
  "query": "authentication patterns",
-  "type": "project",
+  "type": "general",
  "limit": 10
 }
 ```
@@ -167,7 +201,7 @@ Remove a memory by name.
 | Parameter  | Required | Default    | Description              |
 |------------|----------|------------|--------------------------|
 | `name`     | yes      | --         | Memory name to delete    |
-| `scope`    | no       | `"global"` | Scope of the memory      |
+| `scope`    | no       | inherited | Exact scope to delete    |

 ### list

@@ -197,6 +231,12 @@ Four endpoints on the server for programmatic memory access.

 List memories with optional filters.

+Without `scope`, the response is restricted to `global` plus the authenticated
+caller's `user` namespace. The public API accepts only `global`, `user`, and
+`workstream`; internal `project` and `coordinator` namespaces remain available
+through the session tool and admin API. Explicit `workstream` access requires
+its persisted owner (or a service token).
+
 **Query parameters:**

 | Parameter  | Type   | Required | Default | Description                  |
@@ -204,10 +244,10 @@ List memories with optional filters.
 | `type`     | string | no       | `""`    | Filter by memory type        |
 | `scope`    | string | no       | `""`    | Filter by scope              |
 | `scope_id` | string | no       | `""`    | Filter by scope ID           |
-| `limit`    | int    | no       | `100`   | Max results (capped at 200)  |
+| `limit`    | int    | no       | `100`   | Max results (1-200)          |

-When `scope=user` and `scope_id` is omitted, the authenticated user's ID is
-used automatically.
+When `scope=user`, the authenticated user's ID is used automatically and a
+different supplied ID is rejected. `scope=workstream` requires `scope_id`.

 **Response:** `200`

@@ -218,7 +258,7 @@ used automatically.
      "memory_id": "a1b2c3d4-e5f6-...",
      "name": "project_architecture",
      "description": "Core architecture patterns",
-      "type": "project",
+      "type": "general",
      "scope": "global",
      "scope_id": "",
      "content": "The project uses a hexagonal architecture...",
@@ -236,6 +276,9 @@ used automatically.

 Save or upsert a structured memory.

+`description` is mandatory for both creates and updates and must contain
+non-whitespace text. The API rejects content-only updates.
+
 **Request body:**

 ```json
@@ -243,7 +286,7 @@ Save or upsert a structured memory.
  "name": "deployment_process",
  "content": "Deploy via GitHub Actions. Staging auto-deploys on push to main.",
  "description": "CI/CD deployment workflow",
-  "type": "project",
+  "type": "general",
  "scope": "global",
  "scope_id": ""
 }
@@ -253,8 +296,8 @@ Save or upsert a structured memory.
 |--------------|--------|----------|-------------|--------------------------------------|
 | `name`       | string | yes      | --          | Memory name (max 256 chars)          |
 | `content`    | string | yes      | --          | Memory content (max 65536 chars)     |
-| `description`| string | no       | `""`        | Short description for search ranking |
-| `type`       | string | no       | `"project"` | One of: user, project, feedback, reference |
+| `description`| string | yes      | --          | Non-empty relevance summary, required on create and update |
+| `type`       | string | no       | unset       | user, general, feedback, or reference |
 | `scope`      | string | no       | `"global"`  | One of: global, workstream, user     |
 | `scope_id`   | string | no       | `""`        | Scope qualifier (auto-resolved for user scope) |

@@ -265,7 +308,7 @@ Save or upsert a structured memory.
  "memory_id": "a1b2c3d4-e5f6-...",
  "name": "deployment_process",
  "description": "CI/CD deployment workflow",
-  "type": "project",
+  "type": "general",
  "scope": "global",
  "scope_id": "",
  "content": "Deploy via GitHub Actions...",
@@ -281,7 +324,10 @@ same `(name, scope, scope_id)` already existed.

 | Status | Condition                          |
 |--------|------------------------------------|
-| 400    | Missing name, empty content, invalid type/scope, content too long |
+| 400    | Invalid input, scope, scope ID, or limit |
+| 403    | Cross-user or non-owner workstream access |
+| 404    | Explicit workstream does not exist |
+| 500    | Storage mutation failed |

 ---

@@ -290,12 +336,15 @@ same `(name, scope, scope_id)` already existed.
 Search memories by query. Uses POST for the request body but is non-mutating
 (requires only `read` scope).

+An omitted scope searches the same caller-bound `global` + `user` envelope as
+the list endpoint. It never means every row in the table.
+
 **Request body:**

 ```json
 {
  "query": "authentication",
-  "type": "project",
+  "type": "general",
  "scope": "",
  "scope_id": "",
  "limit": 20
@@ -308,7 +357,7 @@ Search memories by query. Uses POST for the request body but is non-mutating
 | `type`     | string | no       | `""`    | Filter by type                 |
 | `scope`    | string | no       | `""`    | Filter by scope                |
 | `scope_id` | string | no       | `""`    | Filter by scope ID             |
-| `limit`    | int    | no       | `20`    | Max results (capped at 50)     |
+| `limit`    | int    | no       | `20`    | Max results (1-50)             |

 **Response:** `200`

@@ -319,7 +368,7 @@ Search memories by query. Uses POST for the request body but is non-mutating
      "memory_id": "a1b2c3d4-e5f6-...",
      "name": "auth_patterns",
      "description": "Authentication architecture",
-      "type": "project",
+      "type": "general",
      "scope": "global",
      "scope_id": "",
      "content": "JWT tokens with HS256...",
@@ -337,6 +386,9 @@ Search memories by query. Uses POST for the request body but is non-mutating

 Delete a memory by name and scope.

+Deletes are atomic: the row used for the success result and audit event is the
+row actually removed. A storage failure returns `500`, not a false `404`.
+
 **Path parameters:**

 | Parameter | Type   | Description          |
@@ -391,7 +443,7 @@ List memories across all scopes (no automatic scope resolution).
      "memory_id": "a1b2c3d4-e5f6-...",
      "name": "project_architecture",
      "description": "Core architecture patterns",
-      "type": "project",
+      "type": "general",
      "scope": "global",
      "scope_id": "",
      "content": "The project uses...",
@@ -440,7 +492,7 @@ Get a single memory by ID.
  "memory_id": "a1b2c3d4-e5f6-...",
  "name": "project_architecture",
  "description": "Core architecture patterns",
-  "type": "project",
+  "type": "general",
  "scope": "global",
  "scope_id": "",
  "content": "The project uses...",
@@ -497,13 +549,13 @@ with TurnstoneServer("http://localhost:8080", token="tok_xxx") as client:
        "api_conventions",
        "All endpoints use /v1/ prefix. JSON responses.",
        description="API design patterns",
-        mem_type="project",
+        mem_type="general",
        scope="global",
    )
    print(mem.memory_id)

    # Search memories
-    results = client.search_memories("authentication", mem_type="project", limit=10)
+    results = client.search_memories("authentication", mem_type="general", limit=10)
    for m in results.memories:
        print(f"{m['name']}: {m['description']}")

@@ -524,7 +576,7 @@ with TurnstoneConsole("http://localhost:9090", token="tok_xxx") as admin:
    result = admin.list_memories(scope="global", limit=100)

    # Search
-    result = admin.search_memories("architecture", mem_type="project")
+    result = admin.search_memories("architecture", mem_type="general")

    # Get by ID
    mem = admin.get_memory("a1b2c3d4-e5f6-...")
@@ -548,14 +600,14 @@ const mem = await client.saveMemory({
  name: "api_conventions",
  content: "All endpoints use /v1/ prefix. JSON responses.",
  description: "API design patterns",
-  type: "project",
+  type: "general",
  scope: "global",
 });

 // Search memories
 const results = await client.searchMemories({
  query: "authentication",
-  type: "project",
+  type: "general",
  limit: 10,
 });

@@ -124,16 +124,78 @@ allow_private_network = true
 or via `TURNSTONE_OIDC_ALLOW_PRIVATE_NETWORK=true` (the env var wins
 when both are set).

-The opt-in admits private-range (RFC 1918), unique-local, CGNAT
-(100.64/10 — tailnets), and loopback addresses. Link-local, multicast,
-and reserved ranges stay refused even with the opt-in — cloud metadata
-services (169.254.169.254) live there, and no legitimate IdP does. The
-HTTPS requirement and the same-origin endpoint checks are unaffected.
+The opt-in admits private-range (RFC 1918), unique-local, site-local,
+CGNAT (100.64/10, where overlay VPNs commonly assign hosts), and
+loopback addresses. Link-local, multicast, reserved ranges and known
+cloud-metadata endpoints stay refused even with the opt-in — no
+legitimate IdP lives there. An address is judged by what it actually
+reaches, so an IPv6 transition address (NAT64, 6to4, Teredo) wrapping
+an internal IPv4 is treated exactly as that IPv4 would be. The HTTPS
+requirement and the same-origin endpoint checks are unaffected.

 This knob only affects the login-flow IdP configured here. OAuth
 endpoints advertised by remote MCP servers are untrusted input and are
 always held to the strict public-address rule.

+### Model gateway credentials
+
+The same OIDC registration can authenticate model gateways. A model definition
+with `auth_mode = "entra_obo"` (Entra grant profile) or `auth_mode =
+"rfc8693_obo"` (RFC 8693 token-exchange profile) redeems the driving user's
+captured credential for its exact `obo_audience`; `auth_mode = "entra_app"`
+uses the registration's client ID and secret with Entra client credentials.
+All three bind the result through the provider SDK's native credential option
+rather than injecting an override header. The grant mode is never inferred:
+missing user context or a failed OBO mint cannot switch a delegated definition
+to client credentials.
+
+Each dynamic mode pairs with the grant profile whose dialect it names:
+`entra_obo` and `entra_app` require `obo_grant_profile = "entra"`;
+`rfc8693_obo` requires `obo_grant_profile = "rfc8693"`. The pairing is
+enforced when a write chooses a `(auth_mode, obo_audience)` pair — a same-pair
+edit of a row saved before the pairing rule keeps working — and at runtime a
+mismatched legacy row refuses to mint with `cause=grant_profile_mismatch` and
+no IdP traffic. RFC 8693 client-credentials is not implemented.
+
+The delegated modes need the MCP encryption key, a credential captured for the
+driving user, and delegated/admin-consented permission to the audience.
+`rfc8693_obo` additionally carries `obo_scopes`, the space-separated scope
+list its exchange leg requests: exchange-capable IdPs that gate audiences
+behind optional scopes refuse the exchange without it ("Requested audience not
+available"), which is why the scope-less Entra-named mode could never mint on
+that profile (issue #955). Scopes are stored shape-checked only — whether a
+value satisfies the IdP stays the IdP's call at mint time. Turning
+`capture_user_credential` off later stops *new* captures but does not
+invalidate credentials already stored, so existing users keep minting.
+`entra_app` requires a confidential-client secret. Configure the permitted
+resource IDs in the runtime setting `model.auth_audience_allowlist` before
+saving dynamic model definitions. De-listing an audience later blocks every
+write that would arm or re-aim a definition at it, but does not stop aliases
+already configured from minting — disabling the row (the `admin.models` disarm
+lever) is what stops minting. See
+[Settings](settings.md#model-backend-authentication) for permissions, failure
+policy, and lane identity rules.
+
+An unrecognised `obo_grant_profile` is warned about at startup and **rejected
+at the write choke points**: configuring an `oauth_obo` MCP server or a dynamic
+model alias returns a 400 that echoes the configured value, so the typo is the
+diagnosis. At runtime an unknown profile never mints — the mint legs resolve by
+exact name; the full cause detail is logged once per audience, and every
+affected call still logs its per-turn fallback or refusal naming the alias,
+the target audience, and the last recorded cause (`cause=` — for example
+`unsupported_grant_profile` or `oidc_not_enabled`) — so a pre-existing row
+degrades loudly, with the reason visible mid-incident even after the
+once-per-process line has rotated out of retained logs, rather than silently
+swapping per-user attribution for the shared static key.
+
+The `[security]` token encryption key is deployment-wide, not per-host: rows are
+encrypted with `MultiFernet` and carry no key id, so every host that reads them
+needs the same keyring. That includes the console, which mints for
+coordinator-hosted sessions. A node that needs the key and lacks it refuses to
+start; the console starts but withholds its coordinator subsystem and shows
+the key requirement as the remediation error instead of failing silently at
+call time.
+
 ### config.toml alternative

 ```toml
@@ -46,8 +46,8 @@ reads only the stamp:
  `creative_mode` set are converted by migration `063` into full
  `writer` stamps, so they resume as writing sessions rather than as
  legacy defaults.
- Forking (`resume_ws` on create) resumes the source's stamped persona; the
-  fork does not re-resolve.
+- Forking (`resume_ws` on create) clones the source's stamped persona into the
+  new workstream; the fork does not re-resolve it.

 ## Seed personas

@@ -15,11 +15,19 @@ down to a small number of real database connections.

 ## Why PgBouncer works well with turnstone

-All turnstone database operations are short-burst queries: acquire a
-connection, execute 1–3 statements, commit, release. No operation holds
-a connection for more than a few milliseconds. This makes **transaction
-pooling mode** ideal — PgBouncer assigns a real connection only for the
-duration of each transaction, then returns it to the pool.
+Most turnstone database operations are short-burst queries: acquire a
+connection, execute a small transaction, commit, release. Workstream forks are
+the deliberate exception: they clone the source's checkpoint-bounded history
+and configuration and retain its attachment references in one transaction.
+PostgreSQL runs that clone at `SERIALIZABLE` isolation and retries serialization
+or deadlock conflicts as a whole. A large fork can therefore hold its assigned
+server connection longer than an ordinary message write.
+
+This still makes **transaction pooling mode** the right fit — no operation
+depends on server-session state, and PgBouncer returns the connection as soon
+as the transaction finishes. Size and monitor the server pool with concurrent
+fork traffic in mind rather than assuming every transaction completes in a few
+milliseconds.

 | Cluster size | Client connections (max) | PgBouncer server connections needed |
 |--------------|------------------------|-------------------------------------|
@@ -143,9 +151,11 @@ PgBouncer (which then multiplexes to PostgreSQL):
 | `TURNSTONE_DB_URL` | — | Connection URL (point at PgBouncer, not PostgreSQL directly) |

 The default pool of 2 + 3 overflow = 5 connections per process is
-intentionally small to support large clusters. You should not need to
-increase this — turnstone's database operations are all short-burst
-context-managed queries that hold connections for milliseconds.
+intentionally small to support large clusters. Most deployments should not
+need to increase it. If operators create many large forks concurrently, watch
+PgBouncer's `cl_waiting` and PostgreSQL transaction latency before changing
+the per-process pool; adding client-side connections cannot help once the
+PgBouncer server pool is saturated.

 SQLAlchemy `pool_pre_ping` is enabled, so stale connections (e.g. after
 PgBouncer restarts) are automatically detected and replaced.
@@ -177,6 +187,32 @@ Key metrics to watch:
 - **`sv_active`** — active server (PostgreSQL) connections. Should stay
  below PostgreSQL `max_connections`.

+Short `cl_waiting` spikes during large workstream forks can be normal. Sustained
+waiters accompanied by long serializable transactions indicate fork/storage
+load, not an SSE or HTTP client-pool problem.
+
+---
+
+## Upgrade note: deferred workstream creation
+
+The workstream lifecycle now uses durable, hidden `state='creating'`
+reservations while session construction, upload validation, and optional fork
+cloning complete. Older server processes do not understand that private state:
+against the same database they may resolve, list, open, or prune a reservation
+before its new owner publishes it.
+
+For the upgrade that introduces deferred creation, drain create traffic and
+upgrade all server processes sharing the database as one cohort. Do not resume
+creates until no older server process remains. The change needs no manual
+schema migration, but it is not safe to treat mixed lifecycle implementations
+as an ordinary rolling-upgrade state.
+
+A `creating` row should be transient and absent from normal APIs and cluster
+events. If one persists after a process crash, inspect the corresponding
+`ws.create.*` and `session_mgr.commit_create.*` logs before cleanup. Do not
+promote it to `idle` manually: its history, configuration, attachment
+references, or lifecycle publication may be incomplete.
+
 ---

 ## Troubleshooting
@@ -50,6 +50,7 @@ with TurnstoneServer("http://localhost:8080") as client:
 import asyncio
 from turnstone.sdk import AsyncTurnstoneServer

+
 async def main():
    async with AsyncTurnstoneServer("http://localhost:8080") as client:
        await client.login(username="alice", password="s3cret")
@@ -58,6 +59,7 @@ async def main():
            if event.type == "content":
                print(event.text, end="", flush=True)

+
 asyncio.run(main())
 ```

@@ -69,17 +71,18 @@ Both `TurnstoneServer` (sync) and `AsyncTurnstoneServer` (async) expose:
 |----------|--------|---------|
 | **Workstreams** | `list_workstreams()` | `ListWorkstreamsResponse` |
 | | `dashboard()` | `DashboardResponse` |
-| | `create_workstream(*, name, model, auto_approve, skill, persona, initial_message, attachments)` | `CreateWorkstreamResponse` |
+| | `create_workstream(*, name, model, auto_approve, resume_ws, skill, persona, initial_message, project_id, attachments, ...)` | `CreateWorkstreamResponse` |
 | | `close_workstream(ws_id)` | `StatusResponse` |
 | **Attachments** | `upload_attachment(ws_id, filename, data, *, mime_type=...)` | `UploadAttachmentResponse` |
 | | `list_attachments(ws_id)` | `ListAttachmentsResponse` |
 | | `get_attachment_content(ws_id, attachment_id)` | `bytes` |
 | | `delete_attachment(ws_id, attachment_id)` | `StatusResponse` |
-| **Chat** | `send(message, ws_id)` | `SendResponse` |
-| | `approve(*, ws_id, approved, feedback, always)` | `StatusResponse` |
+| **Chat** | `send(message, ws_id, *, attachment_ids=None, client_send_id=None)` | `SendResponse` |
+| | `approve(*, ws_id, approved, feedback, always, cycle_id, call_id)` | `ApproveResponse` |
 | | `command(*, ws_id, command)` | `StatusResponse` |
-| | `cancel(ws_id, *, force=False)` | `StatusResponse` |
-| **Streaming** | `stream_events(ws_id)` | `Iterator[ServerEvent]` |
+| | `cancel(ws_id, *, force=False)` | `CancelResponse` |
+| **History** | `get_history(ws_id, *, limit=100)` | `WorkstreamHistoryResponse` |
+| **Streaming** | `stream_events(ws_id, *, last_event_id=None, history_token=None)` | `Iterator[ServerEvent]` |
 | | `stream_global_events()` | `Iterator[ServerEvent]` |
 | **High-level** | `send_and_wait(message, ws_id, *, timeout, on_event)` | `TurnResult` |
 | **Saved** | `list_saved_workstreams()` | `ListSavedWorkstreamsResponse` |
@@ -100,7 +103,7 @@ Both `TurnstoneConsole` (sync) and `AsyncTurnstoneConsole` (async) expose:
 | | `workstreams(*, state, node, search, sort, page, per_page)` | `ClusterWorkstreamsResponse` |
 | | `node_detail(node_id)` | `NodeDetailResponse` |
 | | `snapshot()` | `ClusterSnapshotResponse` |
-| | `create_workstream(*, node_id, name, model, initial_message, skill, persona)` | `ConsoleCreateWsResponse` |
+| | `create_workstream(*, node_id, name, model, initial_message, skill, persona, resume_ws)` | `ConsoleCreateWsResponse` |
 | **Schedules** | `list_schedules()` | `ListSchedulesResponse` |
 | | `create_schedule(*, name, schedule_type, initial_message, ...)` | `ScheduleInfo` |
 | | `get_schedule(task_id)` | `ScheduleInfo` |
@@ -125,12 +128,12 @@ SSE events are deserialized into typed dataclasses. Use `event.type` to discrimi
 | Type | Class | Key Fields |
 |------|-------|------------|
 | `connected` | `ConnectedEvent` | `model`, `model_alias`, `skip_permissions` |
-| `history` | `HistoryEvent` | `messages` |
+| `user_turn` | `UserTurnEvent` | `ws_id`, `content`, `attachments`, `sender`, `source`, `client_send_ids`, `_event_id` |
 | `content` | `ContentEvent` | `text` |
 | `reasoning` | `ReasoningEvent` | `text` |
 | `tool_info` | `ToolInfoEvent` | `items` |
-| `approve_request` | `ApproveRequestEvent` | `items` |
-| `tool_result` | `ToolResultEvent` | `call_id`, `name`, `output`, `is_error` |
+| `approve_request` | `ApproveRequestEvent` | `cycle_id`, `items` |
+| `tool_result` | `ToolResultEvent` | `call_id`, `name`, `output`, `is_error`, `preview`, `accepted`, `effect_status`, `_event_id` |
 | `tool_output_chunk` | `ToolOutputChunkEvent` | `call_id`, `chunk` |
 | `status` | `StatusEvent` | `prompt_tokens`, `total_tokens`, `pct`, `effort`, `cache_creation_tokens`, `cache_read_tokens` |
 | `error` | `ErrorEvent` | `message` |
@@ -138,14 +141,92 @@ SSE events are deserialized into typed dataclasses. Use `event.type` to discrimi
 | `stream_end` | `StreamEndEvent` | — |
 | `state_change` | `StateChangeEvent` | `state` ∈ `running`/`thinking`/`attention`/`idle`/`error` |
 | `in_progress_snapshot` | `InProgressSnapshotEvent` | `content`, `reasoning` (one-shot mid-stream refresh resume) |
-| `approval_resolved` | `ApprovalResolvedEvent` | `approved`, `feedback` |
+| `approval_resolved` | `ApprovalResolvedEvent` | `cycle_id`, `call_ids`, `approved`, `feedback`, `always` |
 | `cancelled` | `CancelledEvent` | — |
+| `history_resync` | `HistoryResyncEvent` | `reason`, optional `ws_id` |
+
+The Python server `send()` and console `coordinator_send()` methods accept an
+optional `client_send_id`; TypeScript `send()` accepts the equivalent
+`options.clientSendId`. Values match `[A-Za-z0-9_-]{1,128}`. The value is an
+opaque optimistic-UI correlation token, not an idempotency key: reusing it
+still creates distinct accepted turns and events.
+Every upgraded listener on the shared workstream receives `UserTurnEvent`.
+Originating panes use `client_send_ids` only to settle the exact optimistic
+bubble, while peers render the accepted row once by `_event_id`. A
+`message_queued` event carrying the token can establish acceptance even if the
+POST acknowledgement is lost. History projects the same correlation alongside
+the accepted user row. These tokens are not credentials: when sender and viewer
+identities are both known, only a matching sender may settle local optimistic
+state; a peer event still renders its canonical row.
+
+The typed projection is negotiated with `?user_turn=1` on the per-workstream
+SSE URL. Python `stream_events()` / `send_and_wait()` and TypeScript
+`streamEvents()` / `sendAndWait()` set it automatically. Raw consumers that
+omit it receive a backward-compatible `replay_truncated` repair signal instead
+of the user row and must rebuild from `/history`; its pre-row cursor keeps the
+repair retryable if that history request fails.
+
+The browser-only final-tool upsert capability is `?tool_turn=1`. The bundled
+Python and TypeScript SDK streaming helpers and channel adapters intentionally
+do not negotiate it yet: they retain the executor-receipt `tool_result`
+contract and do not own a transcript reducer. `ToolResultEvent` can deserialize
+the accepted fields for direct/custom capable clients. Raw capable clients must
+deduplicate `_event_id` and replace the newest matching call occurrence; raw
+incapable clients receive the pre-row `tool_turn_projection_unsupported` repair
+frame and rebuild from history. That staging deliberately prices in two costs
+for incapable consumers. A raw client that treats every `replay_truncated`
+frame as a rebuild trigger refetches `/history` once per accepted tool row —
+one fetch per tool call on a long agentic turn; a client that wants tool
+results incrementally should negotiate `tool_turn=1` and reduce, and the
+bundled helpers (which ignore the frame rather than rebuild) stay correct
+because their receipt-only view never depends on the accepted projection.
+Second, only the accepted event carries post-execution output transforms, so a
+receipt-rendering consumer (for example, a channel adapter posting the
+executor receipt into a thread) keeps the pre-transform text; the accepted
+projection is a transcript-consistency mechanism, not a wire confidentiality
+boundary — see the API reference note on the preliminary `tool_result`.
+
+Current servers bootstrap conversation history through
+`GET /v1/api/workstreams/{ws_id}/history` before the SSE stream; they do not
+emit a `history` event. `HistoryEvent` remains deserializable only for
+compatibility with older servers. `get_history()` exposes the current REST
+bootstrap response, including its optional cursor and one-shot handoff token.
+
+### Caller-managed history handoff
+
+The SDK supplies typed handshake primitives but intentionally does not own a
+transcript renderer or reconnect policy. After rendering a successful history
+response, pass its cursor and token to exactly one initial stream:
+
+```python
+from turnstone.sdk import HistoryResyncEvent
+
+history = client.get_history(ws_id)
+render(history.messages)
+
+for event in client.stream_events(
+    ws_id,
+    last_event_id=history.cursor,
+    history_token=history.handoff_token,
+):
+    if isinstance(event, HistoryResyncEvent):
+        # Stop this stream. The caller chooses when to fetch, render, and
+        # reconnect with a new history response.
+        break
+    apply_live_event(event)
+```
+
+`history_resync` means numeric replay cannot prove that the rendered limited
+tail came from the same total accepted conversation-row prefix. Stop the
+stream, fetch and render history again, and use only the new cursor/token pair.
+A 503 history response raises `TurnstoneAPIError`; it is not authoritative, so
+retain any existing transcript and do not open a tokenless replacement stream.

 **Global events** (from `stream_global_events()`):

 | Type | Class | Key Fields |
 |------|-------|------------|
-| `ws_state` | `WsStateEvent` | `ws_id`, `state`, `tokens`, `activity` |
+| `ws_state` | `WsStateEvent` | `ws_id`, `state`, `tokens`, `activity`, `persistence_state` |
 | `ws_activity` | `WsActivityEvent` | `ws_id`, `activity`, `activity_state` |
 | `ws_rename` | `WsRenameEvent` | `ws_id`, `name` |
 | `ws_closed` | `WsClosedEvent` | `ws_id` |
@@ -156,24 +237,30 @@ SSE events are deserialized into typed dataclasses. Use `event.type` to discrimi
 |------|-------|------------|
 | `node_joined` | `NodeJoinedEvent` | `node_id` |
 | `node_lost` | `NodeLostEvent` | `node_id` |
-| `cluster_state` | `ClusterStateEvent` | `ws_id`, `node_id`, `state`, `tokens` |
-| `ws_created` | `ClusterWsCreatedEvent` | `ws_id`, `node_id`, `name` |
+| `cluster_state` | `ClusterStateEvent` | `ws_id`, `node_id`, `state`, `tokens`, `persistence_state` |
+| `ws_created` | `ClusterWsCreatedEvent` | `ws_id`, `node_id`, `name`, `persistence_state` |
 | `ws_closed` | `ClusterWsClosedEvent` | `ws_id` |
 | `ws_rename` | `ClusterWsRenameEvent` | `ws_id`, `name` |
 | `snapshot` | `ClusterSnapshotEvent` | `nodes`, `overview`, `timestamp` |

+Operator-facing workstream rows and rich state events expose only the sanitized
+`persistence_state`: `healthy`, `pending`, `retrying`, or `conflict`. SDK types
+treat it as optional for compatibility with older nodes; an omitted value means
+`healthy`. Retry counts, storage errors, commit keys, and conversation content
+are never part of this status surface.
+
 ### TurnResult

 The `send_and_wait()` method returns a `TurnResult` that aggregates the full response:

 ```python
 result = client.send_and_wait("Hello", ws_id, timeout=60)
-result.content      # Full text response
-result.reasoning    # Chain-of-thought (if shown)
-result.tool_results # List of (tool_name, output) tuples
-result.errors       # Any error messages
-result.ok           # True if no errors and not timed out
-result.timed_out    # True if timeout expired
+result.content  # Full text response
+result.reasoning  # Chain-of-thought (if shown)
+result.tool_results  # List of (tool_name, output) tuples
+result.errors  # Any error messages
+result.ok  # True if no errors and not timed out
+result.timed_out  # True if timeout expired
 ```

 ### Attachments
@@ -183,9 +270,7 @@ Upload files to a workstream and attach them to the next user turn:
 ```python
 # Upload separately, then send a message — attachments auto-attach
 with open("screenshot.png", "rb") as f:
-    att = client.upload_attachment(ws.ws_id, "screenshot.png",
-                                   f.read(),
-                                   mime_type="image/png")
+    att = client.upload_attachment(ws.ws_id, "screenshot.png", f.read(), mime_type="image/png")
 client.send("What's wrong in this screenshot?", ws.ws_id)

 # Or attach at workstream-creation time (multipart upload)
@@ -195,9 +280,7 @@ with open("notes.txt", "rb") as f:
    ws = client.create_workstream(
        name="triage",
        initial_message="Summarize the notes",
-        attachments=[AttachmentUpload(data=f.read(),
-                                      filename="notes.txt",
-                                      mime_type="text/plain")],
+        attachments=[AttachmentUpload(data=f.read(), filename="notes.txt", mime_type="text/plain")],
    )
 ```

@@ -206,6 +289,26 @@ Limits: images ≤ 4 MiB (png/jpeg/gif/webp), text ≤ 512 KiB (UTF-8),
 client so cluster-routed callers bind attachments to the owning node
 before the request lands.

+### Forking a workstream
+
+`resume_ws` is the API's compatibility name for an atomic fork. It creates a
+new workstream ID while the source remains unchanged:
+
+```python
+fork = client.create_workstream(
+    resume_ws=ws.ws_id,
+    name="analysis-branch",
+    initial_message="Try the alternative plan.",
+)
+assert fork.resumed
+```
+
+The server transaction clones the source's checkpoint-bounded history, saved
+session configuration, persona, project, and attachment references. Do not
+combine `resume_ws` with `attachments`; fork first, then upload to the new ID.
+To rehydrate the original ID rather than branch it, call the server's
+`POST /v1/api/workstreams/{ws_id}/open` endpoint.
+
 ### Error Handling

 Non-2xx responses raise `TurnstoneAPIError`:
@@ -217,7 +320,7 @@ try:
    client.send("hi", "bad_ws_id")
 except TurnstoneAPIError as e:
    print(e.status_code)  # 404
-    print(e.message)      # "Unknown workstream"
+    print(e.message)  # "Unknown workstream"
 ```

 ---
@@ -242,8 +345,14 @@ const ws = await client.createWorkstream({ name: "demo" });
 const result = await client.sendAndWait("Hello!", ws.ws_id);
 console.log(result.content);

-// Stream events
-for await (const event of client.streamEvents(ws.ws_id)) {
+// Render history, then use its one-shot hints on the initial stream.
+const history = await client.getHistory(ws.ws_id);
+render(history.messages);
+for await (const event of client.streamEvents(ws.ws_id, {
+  lastEventId: history.cursor ?? undefined,
+  historyToken: history.handoff_token ?? undefined,
+})) {
+  if (event.type === "history_resync") break; // caller refetches and reconnects
  if (event.type === "content") {
    process.stdout.write(event.text);
  }
@@ -319,7 +428,7 @@ turnstone/sdk/               Python SDK (sub-package)
  _base.py                   Shared httpx async client, auth, error handling
  _sync.py                   Background event loop for sync wrappers
  _types.py                  TurnResult + TurnstoneAPIError
-  events.py                  38 SSE event dataclasses with type registry
+  events.py                  Typed SSE event dataclasses with type registry
  server.py                  AsyncTurnstoneServer + TurnstoneServer
  console.py                 AsyncTurnstoneConsole + TurnstoneConsole

@@ -64,15 +64,17 @@ Scopes are hierarchical — higher scopes imply all lower ones.

 ### Path-to-scope mapping

-| Method | Path pattern | Required scope |
-|--------|-------------|----------------|
-| GET | Any protected path | `read` |
-| POST | `/api/command` | `write` |
-| POST | `/api/workstreams/new`, `/api/cluster/workstreams/new` | `write` |
-| POST | `/api/workstreams/{ws_id}/{send,cancel,close,delete,open,refresh-title,title,attachments}` | `write` |
-| DELETE | `/api/workstreams/{ws_id}/send` (dequeue), `/api/workstreams/{ws_id}/attachments/{attachment_id}` | `write` |
-| POST | `/api/workstreams/{ws_id}/approve` | `approve` |
-| Any | `/api/admin/*` | `approve` |
+| Method | Path pattern | Required scope | Additional RBAC gate |
+|--------|-------------|----------------|----------------------|
+| GET | Any protected path | `read` | Endpoint-specific where documented |
+| POST | `/api/command` | `write` | Project tenancy on the target workstream |
+| POST | `/api/workstreams/new`, `/api/cluster/workstreams/new` | `write` | `workstreams.create` or `admin.coordinator` |
+| POST | `/api/workstreams/{ws_id}/close` | `write` | `workstreams.close` or `admin.coordinator` |
+| POST | `/api/workstreams/{ws_id}/approve` | `approve` | `tools.approve` or `admin.coordinator` |
+| POST | `/api/workstreams/{ws_id}/{rewind,retry}` | `write` | `conversation.modify` |
+| POST | Other `/api/workstreams/{ws_id}/...` mutation endpoints | `write` | Project tenancy and endpoint-specific gates |
+| DELETE | `/api/workstreams/{ws_id}/send` (dequeue), `/api/workstreams/{ws_id}/attachments/{attachment_id}` | `write` | Project tenancy on the target workstream |
+| Any | `/api/admin/*` | `approve` | Matching `admin.*` permission |

 Public paths bypass authentication entirely: `/`, `/health`, `/metrics`,
 `/static/*`, `/shared/*`, `/docs`, `/openapi.json`, `/api/auth/login`,
@@ -84,7 +86,7 @@ Public paths bypass authentication entirely: `/`, `/health`, `/metrics`,
 > See also: [Governance documentation](governance.md)

 Scopes provide coarse endpoint-level access control. For finer-grained
-enforcement, the governance layer adds 15 named permissions checked
+enforcement, the governance layer adds named permissions checked
 per-endpoint by `require_permission()`. Permissions are bundled into
 roles; users are assigned roles via the `user_roles` join table.

@@ -98,8 +100,8 @@ Three built-in roles are seeded by migration 008:

 | Role | Permissions |
 |------|-------------|
-| admin | All 15 permissions |
-| operator | read, write, workstreams.create, workstreams.close |
+| admin | Admin-default baseline (all ordinary admin and lifecycle permissions; explicitly opt-in capabilities remain ungranted) |
+| operator | read, write, workstreams.create, workstreams.close, conversation.modify |
 | viewer | read |

 Custom roles can be created with any subset of the valid permissions.
@@ -107,6 +109,34 @@ Role creation and update validate permissions against a static allowlist.
 Self-assignment is blocked, and assigning a role requires the caller to
 hold a superset of the target role's permissions.

+### Workstream lifecycle and project boundaries
+
+The remote `/api/command` endpoint is conversation-local. It refuses
+`/new`, `/workstreams`, `/resume`, and `/delete` because those local-CLI
+helpers enumerate or mutate storage outside the HTTP resource gates. Remote
+clients use the dedicated create, open, close, and delete endpoints instead;
+`/rewind` and `/retry` have their own path-keyed, `conversation.modify`-gated
+endpoints.
+
+Passing `resume_ws` to create is an atomic **fork**, not an in-place resume.
+It requires the ordinary create capability and source visibility. A private
+project source is visible only to its workstream creator, project owner/member,
+or authorized service-to-service cluster plumbing; denials use a not-found
+response so guessed IDs do not become an existence oracle. The caller must also
+be allowed to attach a new workstream to the source's current project. The
+destination always inherits that effective project — a caller-supplied
+`project_id` cannot re-file or declassify the conversation.
+
+The canonical preflight atomically captures (and, for a legacy row, installs) a
+private source-incarnation fence. The storage transaction compares that source
+fence, rejects provisional sources, repeats the ACL/project check, and verifies
+the persona/project construction snapshot, destination ownership and
+incarnation, emptiness, and every referenced attachment before committing. A
+source replacement, membership, project, persona, or destination-incarnation
+race aborts the whole fork. Concurrent source-history writes serialize wholly
+before or after the snapshot; no mixed or partially authorized history or
+attachment reference becomes visible.
+
 ---

 ## Login Flows
@@ -441,12 +471,15 @@ Each proxied request gets a fresh JWT (5-minute expiry).  This ensures:
 - **Permission forwarding** — granular RBAC permissions from the
  console JWT are carried through to the server.

-The JWT `src` claim is set to `"console-proxy"`, allowing servers to
-distinguish proxied requests from direct logins in audit logs.
+For ordinary users the JWT `src` claim is set to `"console-proxy"`, allowing
+servers to distinguish proxied requests from direct logins in audit logs.
+Coordinator tokens retain `src="coordinator"` and their signed `coord_ws_id`;
+the console service identity retains `src="console"` only when its validated
+token also carries the unassignable `service` scope.

 When no user context is available (auth disabled, or internal requests),
-the proxy falls back to a `ServiceTokenManager` with service identity
-`console-proxy` and full scopes.
+the proxy falls back to a `ServiceTokenManager` with identity `console-proxy`,
+`src="console"`, and `{read, write, approve, service}` scopes.

 ### Service-to-service authentication

@@ -455,8 +488,8 @@ JWTs when communicating with server nodes:

 | Service | Identity | Scope | Audience | Purpose |
 |---------|----------|-------|----------|---------|
-| Console collector | `console-collector` | `read` | `turnstone-server` | Node health polling |
-| Console proxy (fallback) | `console-proxy` | `approve` | `turnstone-server` | Proxied API calls when no user context |
+| Console collector | `console-collector` | `read`, `service` | `turnstone-server` | Node health polling and global event collection |
+| Console proxy (fallback) | `console-proxy` | `read`, `write`, `approve`, `service` | `turnstone-server` | Proxied API calls when no user context |
 | Channel notify | `system` | `write` | `turnstone-channel` | Notification delivery to channel gateway |

 Service tokens use 1-hour expiry with automatic refresh via
@@ -468,8 +501,8 @@ When the console creates a workstream (the normal path), the
 authenticated user's `user_id` is forwarded in the HTTP payload when
 calling the server's `POST /v1/api/workstreams/new`.  The server
 accepts a `user_id` from the request body **only when the caller is a
-trusted service** — identified by `token_source` matching
-`console-proxy` or `console`.  Regular API callers cannot
+trusted service** — identified by `token_source="console"` together with the
+unassignable `service` scope. `console-proxy`, coordinator, and regular API callers cannot
 override `user_id`; the server always uses their JWT identity.

 Note that the channel gateway uses a distinct JWT audience
@@ -54,6 +54,132 @@ When a per-model override is `NULL` (empty in the UI), the global default is
 used. Switching models via `/model <alias>` re-resolves sampling parameters
 from the new model's overrides or global defaults.

+### Per-model concurrency
+
+Each model definition may set `max_concurrency` to limit simultaneous model
+generations for that alias in one Turnstone process. `0` or an omitted value
+means unlimited. The gate is shared by every role using the alias—interactive
+turns, coordinators, task agents, judges, output guards, perception, compaction,
+and title generation—and a streaming generation holds its slot until the
+stream is fully drained or closed.
+
+Admission is strictly per alias. Two aliases remain independent even when they
+point to the same URL; Turnstone does not infer shared capacity from endpoint
+text. Queue time is excluded from judge/output-guard deadline accounting, and
+each retry releases its slot before backoff and reacquires for the next wire
+attempt. The cap is local to each process, not cluster-wide; account for the
+number of nodes targeting the same inference server. Direct STT/TTS protocol
+calls and Cohere/Jina reranking do not currently consume this generation cap.
+
+### Judge batch parallelism
+
+`judge.parallel_evaluations` controls how many independent tool calls from one
+approval batch the intent judge evaluates concurrently. It is an integer from
+1 through 16 and defaults to 1, preserving serial evaluation until an operator
+opts into wider fan-out. Changes are hot-read at the next batch; work already
+in flight keeps its captured worker count.
+
+This is a per-batch fan-out setting, not another backend capacity limit. The
+judge model alias's `max_concurrency` gate still caps total generations across
+all judge batches and every other role using that alias. Actual overlap is
+therefore bounded by the batch size, `judge.parallel_evaluations`, and available
+alias admission slots. A smaller positive alias cap also narrows the batch's
+worker pool so excess judge threads do not queue ahead of later alias traffic.
+
+### Model backend authentication
+
+Model definitions support four backend credential modes:
+
+| `auth_mode` | Identity sent to the model gateway |
+|-------------|------------------------------------|
+| `static` | The definition's stored `api_key`. |
+| `entra_obo` | A caller-delegated Entra access token minted from that user's captured OIDC credential. |
+| `entra_app` | A shared app-identity token minted with Turnstone's OIDC client credentials. |
+| `rfc8693_obo` | A caller-delegated access token minted from the captured credential via RFC 8693 token exchange, requesting the definition's `obo_scopes`. |
+
+Dynamic modes require an exact `obo_audience` resource identifier. Before an
+admin can save one, an operator must add that literal audience to
+`model.auth_audience_allowlist` (comma- or newline-separated). Wildcards and
+base-URL host matching are intentionally unsupported, and a row whose
+effective mode is `static` refuses to store a new non-empty `obo_audience` on
+either create or update — an audience cannot be staged for a later flip
+(clearing a stale value, or re-saving it unchanged, stays allowed).
+`obo_scopes` follows the same staging rule with the mode set inverted: only
+`rfc8693_obo` reads it, so every other effective mode refuses to store a new
+non-empty value, while clearing or re-saving one unchanged stays open. The
+value itself is optional and shape-checked only — whether it satisfies the
+IdP is decided at mint time. On a row that is (or becomes) dynamic, every
+change except the tuning fields — context window, temperature, max tokens,
+reasoning effort, and the two reasoning-persistence toggles — also requires
+`admin.mcp`; service tokens do not bypass this capability-escalation gate.
+The one exception is de-escalation: a save whose only gated change is
+switching `enabled` off is a pure disable, needs only `admin.models`, and
+skips validation — a de-listed audience must never block disarming its own
+row. The gate is deny-by-default: a field counts as auth-relevant unless it
+is provably neutral, so re-enabling a disabled dynamic row, re-pointing its
+`base_url`, or swapping its provider or alias all escalate.
+
+Validation runs in two tiers, matching the MCP `oauth_obo` write rules. Row
+validity — the audience is allow-listed — applies to every gated write that
+touches a dynamic configuration, so a revoked audience can be neither silently
+re-pointed at a new `base_url` nor re-armed by an enable flip. Deployment
+posture — the token encryption key installed, single sign-on configured, and
+the grant profile valid and able to carry the mode — is checked when a write
+*chooses* the mode/audience pair and when it re-enables a disabled dynamic
+row (arming is the flip that resumes minting, so it must meet what minting
+needs); other edits to an existing row stay open if the deployment's posture
+changed after it was saved (its mints warn at runtime instead). Refusals name
+their cause and echo the configured value.
+
+One asymmetry to be aware of: the write path counts a transient discovery
+outage (`enabled=false`, retryable) as configured, but the mints themselves
+require discovery to have completed — a config saved during an outage starts
+minting only once any authenticated request heals discovery. Until then calls
+warn and follow the fail-open/fail-closed policy above.
+
+Every dynamic mode pairs with exactly one grant profile: `entra_obo` and
+`entra_app` require `[oidc] obo_grant_profile = "entra"`, and `rfc8693_obo`
+requires `"rfc8693"`. The pairing is enforced at the posture tier, so a row
+saved before the rule existed keeps accepting same-pair edits; its mints
+refuse at runtime with `cause=grant_profile_mismatch` and no IdP traffic.
+Judge, output-guard, perception, utility, and sub-agent lanes inherit the
+session's effective user for the delegated modes. The perception memo is
+partitioned by that principal as well as alias and content hash, so a result
+authorized as one user cannot be served to another. Scheduled and wake-driven
+work retains the workstream owner even when no user is connected. Eval and
+optimizer lanes are registry-less development tools and therefore do not use
+dynamic model authentication.
+
+`entra_app` is an explicit model-definition choice; Turnstone never changes a
+failed or ownerless delegated call into a client-credentials grant. A
+delegated-mode call with no effective user always refuses. A dynamic alias
+without a real static key also always refuses instead of issuing its
+SDK-construction placeholder. When a real static key is explicitly configured,
+mint failures may use it by default; set `model.auth_fail_closed = true` to
+prohibit even that fallback. A refusal is not routed through the model
+fallback chain.
+
+Dynamic token caches are encrypted in `mcp_user_tokens`, shared across nodes,
+and memoized on each host. Unlinking a user's OIDC identity purges their
+delegated-mode rows and memo entries. `entra_app` rows belong to the shared
+`__app__` identity and are not user-deprovisioned; after client-credential
+revocation, an already-minted app bearer remains usable until its recorded
+expiry.
+
+Each model call resolves its dynamic credential against the immutable model
+definition snapshot that supplied that call's provider, client, endpoint, and
+model ID. An admin edit can therefore never pair an old `base_url` with a new
+audience, grant mode, or static-key fallback input. The principal and token
+remain per-call/live; the connection and model-owned auth configuration move
+together as one binding on the next operation. The deployment-wide
+`model.auth_fail_closed` switch is intentionally read live on every mint, so an
+operator can tighten fallback policy immediately without rebuilding sessions.
+
+`obo_audience` and `obo_scopes` are literal and capped at 2048 characters
+each. Environment-variable expansion is deliberately not applied, so the
+allow-list decision cannot vary by node or expand beyond the persisted
+boundary.
+
 ### Responses output controls (per-model)

 Models whose capability table declares Responses output controls expose two
@@ -124,7 +250,7 @@ initialization:

 | Section | Settings |
 |---------|----------|
-| `model` | default_alias, temperature, max_tokens, reasoning_effort, task_alias, task_effort |
+| `model` | default_alias, auth_audience_allowlist, auth_fail_closed, temperature, max_tokens, reasoning_effort, task_alias, task_effort |
 | `session` | instructions, retention_days, compact_max_tokens, auto_compact_pct |
 | `tools` | timeout, truncation, agent_max_turns, skip_permissions, search, search_threshold, search_max_results |
 | `server` | workstream_idle_timeout, max_workstreams |
@@ -132,7 +258,7 @@ initialization:
 | `mcp` | config_path, registry_url |
 | `ratelimit` | enabled, requests_per_second, burst, trusted_proxies |
 | `health` | backend_probe_interval, backend_probe_timeout, circuit_breaker_threshold, circuit_breaker_cooldown |
-| `judge` | enabled, model, provider, base_url, api_key, confidence_threshold, max_context_ratio, timeout, read_only_tools, output_guard, redact_secrets, cancel_on_approval |
+| `judge` | enabled, model, smart_approvals, confidence_threshold, max_context_ratio, timeout, parallel_evaluations, read_only_tools, output_guard, output_guard_budget_seconds, output_guard_llm, output_guard_model, output_guard_llm_timeout, redact_secrets, cancel_on_approval |
 | `interface` | close_tab_action, theme |
 | `skills` | discovery_url |
 | `memory` | relevance_k, fetch_limit, max_content, nudge_cooldown, nudges |
@@ -309,13 +435,11 @@ Reset a setting to its registry default by removing it from storage.

 ## Secret Settings

-Settings with `is_secret=True` (currently only `judge.api_key`) are blocked
-from the write API with a `403` response. This prevents accidental exposure
-through the admin UI or audit logs. Secret settings must be configured via
-`config.toml` or environment variables.
-
-The list endpoint masks secret values: stored secrets appear as `"***"`
-rather than their actual value.
+The registry currently defines no production secret system setting. The generic
+machinery nevertheless treats any future `is_secret=True` entry as write-only:
+list and write responses return `"***"`, and submitting that sentinel preserves
+the stored value. Model API keys are fields on model definitions—not
+`judge.*` system settings—and use the Models tab's separate write-only flow.

 ---

@@ -336,10 +460,40 @@ reload.
 **Behavior after reload:**

 - New workstreams pick up updated values immediately (via `session_factory`)
- Existing sessions keep their frozen configuration (settings are captured at
-  workstream creation time, not read on every turn)
+- Most workstream/session settings remain the snapshot captured at creation or
+  resume. Component docs call out deliberate live-read exceptions; for
+  example, Smart Approval settings are snapshotted coherently at the start of
+  each approval batch.
 - Settings marked `restart_required=True` need a server restart to take effect

+### Model-definition reloads
+
+The Models tab has a separate live-reload contract from ordinary ConfigStore
+settings. Existing sessions remember the concrete registry generation that
+supplied their active alias and re-resolve that alias at the start of the next
+send. Endpoint, provider, backend model ID, capabilities, extra parameters, and
+backend-auth configuration are replaced as one immutable binding. In-flight
+turns, judges, and task agents finish or cancel against the binding they
+started with; an admin edit never tears one request across two definitions.
+The alias's admission gate is retained and resized in place, so a concurrency
+edit preserves in-flight accounting and does not reset cached judges or the
+output-guard rate limiter.
+
+Sampling and other saved workstream configuration remain workstream state. A
+model-definition edit does not silently rewrite a live workstream's chosen
+temperature, reasoning effort, max tokens, skill, or persona. Use
+`/model <alias>` (or create/fork a workstream) when an explicit session-level
+model switch is intended.
+
+If a live workstream's alias is deleted, its next send first attempts the
+configured fallback chain. Without a usable fallback, the operator-facing
+error names the removed alias and points interactive users to `/model`; adding
+the alias back causes the next send to rebind without a process restart. If a
+replacement client cannot be constructed, Turnstone logs one
+`session.model_refresh_client_construction_failed` warning per registry
+generation and retries only after another model reload, avoiding a rebuild
+storm on every send.
+
 ---

 ## Migration from config.toml
@@ -1,7 +1,7 @@
 ---
 name: import-conversation-history
 description: Use this skill when the user wants to import or migrate conversation history from another LLM chat or coding tool (e.g. ChatGPT, Claude.ai, Cursor, Copilot Chat, Aider, Gemini, a custom JSON export) into Turnstone. The skill teaches Turnstone's destination contracts — workstream identity, the OpenAI-shaped message rows, tool-call/result pairing, provider-fidelity blobs, attachments, and archive-vs-resumable choice — so the agent can map any source format onto them. Trigger phrases: "import my chats", "migrate this transcript into Turnstone", "bring my Claude.ai history over", "load this export as a workstream".
-version: 1.0.0
+version: 1.1.0
 ---

 # Importing Conversation History into Turnstone
@@ -12,7 +12,7 @@ Source formats vary; the destination does not. Your job is to translate whatever

 Two questions to settle with the user before writing anything:

-1. **Archive or resumable?** An archive ("saved" workstream — `state="closed"`) is read-only history. A resumable workstream (`state="idle"`) lets the user continue the conversation; this only works cleanly when the source LLM matches a Turnstone-supported provider/model and tool definitions still resolve.
+1. **Archive or resumable?** An archive is left closed and is read-only history. A resumable import is also kept closed and unloaded while rows are written, then explicitly opened after validation; this only works cleanly when the source LLM matches a Turnstone-supported provider/model and tool definitions still resolve.
 2. **One workstream per source thread, or merge?** Default to one-to-one unless the user explicitly asks to merge.

 Default to **archive** when in doubt — resuming a foreign transcript with mismatched tool schemas or stale provider signatures will fail at the next turn.
@@ -25,13 +25,13 @@ Two tables carry the conversation:

 | Column | Required | Notes |
 |---|---|---|
-| `ws_id` | yes | 32-char lowercase hex. Auto-generate with `secrets.token_hex(16)` if you don't already have one. **First 4 hex chars are the routing bucket** — see "Identity & Routing" below. |
+| `ws_id` | yes | 32-char lowercase hex. Auto-generate with `secrets.token_hex(16)` if you don't already have one. The router hashes the **full ID** — see "Identity & Routing" below. |
 | `name` | yes | Short title. Pull from source thread title; fall back to first ~60 chars of first user message. |
-| `state` | yes | `"closed"` for archive, `"idle"` for resumable. Never set `"running"` on import. |
+| `state` | yes | Register as `"closed"` while importing. Leave it closed for an archive; explicitly open it after commit for a resumable import. Never set `"running"` or `"creating"` directly. |
 | `kind` | yes | `"interactive"` for normal threads. Do NOT use `"coordinator"` for imports — that's reserved for cluster-spawned coordinator workstreams. |
 | `parent_ws_id` | no | Leave NULL. Only set if you're importing a coordinator-spawned subtree and re-parenting it; rare. |
 | `user_id` | yes | Owner. Must exist in `users`; importer must know which Turnstone user owns the imported history. |
-| `node_id` | yes (multi-node) | Denormalized cache of the node that owns this `ws_id`'s bucket. Single-node deployments can leave it NULL or set it to the only node. |
+| `node_id` | no | Nullable creation-time service/liveness hint. It is not the routing key or durable owner and may become stale after membership changes. Let a routed create stamp it; a direct shared-storage import may leave it NULL. |
 | `alias` | no | Human-typeable short name. Optional; must be unique cluster-wide if set. |
 | `title` | no | Auto-titled later by the LLM; safe to leave NULL on import. |
 | `skill_id`, `skill_version` | yes | Default `""` and `0` unless the source thread was scoped to a Turnstone skill. |
@@ -55,25 +55,65 @@ The internal format is **OpenAI-shaped**, even when the source was Anthropic or
 ## Identity & Routing (`ws_id`)

 - `ws_id` is **32-char lowercase hex** (i.e. `secrets.token_hex(16)`).
- The **routing bucket** is `int(ws_id[:4], 16)` — the first 4 hex chars place this workstream on a specific node via the consistent hash ring.
- For multi-node imports: either insert through the console's routing proxy (which forwards to the owning node), or generate `ws_id`s and write directly to each node's database in batches grouped by bucket.
- For single-node imports: bucket math is irrelevant; any `ws_id` works.
+- Ordinary placement is rendezvous (Highest Random Weight, HRW) selection over
+  the **full `ws_id`** and the current live server set. For each node, Turnstone
+  computes 32-bit FNV-1a over the node ID, a NUL separator, and the full
+  workstream ID; it then applies the node weight and selects the highest score.
+  A live per-workstream override takes precedence.
+- The live set comes from recent `services` heartbeats. Placement can therefore
+  change when nodes join, leave, change weight, or an override changes. There
+  is no stable prefix-derived placement to pre-compute or persist.
+- `workstreams.node_id` is stamped at creation and is not updated as HRW
+  placement changes. It supports display and liveness-safe cleanup; the console
+  router does not use it as the ordinary ownership decision.
+- For multi-node imports, create through the console routing proxy when the
+  lifecycle must be published, or write the history once through the cluster's
+  configured **shared storage backend**. Never partition rows across node-local
+  databases by ID prefix or by a one-time HRW result: a later membership change
+  can route the same full ID to another node.
+- For single-node imports, HRW placement is degenerate; any valid `ws_id` works.
 - **Do not reuse the source platform's IDs as `ws_id`** unless they happen to be 32-char hex. Generate fresh; if you need the old ID for traceability, store it in `workstream_config` under a key like `import.source_id`.

 ## Recommended Import Path

 Three options, in order of preference:

-### 1. Storage protocol (recommended for full history)
+### 1. Quiesced storage import (recommended for full history)

-Use `turnstone.core.storage.Storage.save_messages_bulk(rows)`. This is the canonical bulk-insert primitive and bypasses the LLM round-trip entirely.
+Use the current `turnstone.core.storage.StorageBackend` protocol against the
+same shared backend as the cluster. The destination must remain absent from all
+in-memory session managers while rows are changing: a loaded `ChatSession`
+holds its own trajectory and will not observe conversation rows inserted behind
+it.
+
+The safe sequence is:
+
+1. Normalize and validate the complete source transcript before writing.
+2. Call `register_workstream(..., state="closed")` and require a `True` return;
+   `False` means the caller-selected ID already exists, so abort rather than
+   appending to an unrelated workstream.
+3. Insert the ordered conversation rows and attachment references.
+4. Load the saved rows back and run the validation checklist below.
+5. Leave an archive closed. For a resumable import, only now invoke the normal
+   `POST /v1/api/workstreams/{ws_id}/open` endpoint on the currently routed
+   node so the session hydrates from the complete transcript.
+
+Do **not** create the destination through the web/SDK create endpoint before a
+direct bulk import. Create publishes an empty live session. If that already
+happened, close the workstream and confirm the manager-authoritative live probe
+returns false before writing, then explicitly open it again after validation.
+
+For attachment-free history, `save_messages_bulk(rows)` is the canonical
+single-transaction insert primitive and bypasses the LLM round-trip entirely.
+New attachment bytes require the per-row path described under
+[Attachments](#attachments).

 ```python
-from turnstone.core.storage import get_storage  # construct via the same path the server uses
+from turnstone.core.storage import get_storage  # initialized by the host/import entry point

-storage = get_storage(...)  # see turnstone.core.storage.__init__ for the project's wiring
+storage = get_storage()

-storage.create_workstream(  # or whatever the project's exposed creator is — check turnstone/core/storage/_protocol.py
+inserted = storage.register_workstream(
    ws_id=ws_id,
    user_id=user_id,
    name=name,
@@ -81,6 +121,8 @@ storage.create_workstream(  # or whatever the project's exposed creator is — c
    kind="interactive",
    ...
 )
+if not inserted:
+    raise RuntimeError(f"destination already exists: {ws_id}")

 storage.save_messages_bulk([
    {"ws_id": ws_id, "role": "user", "content": "Hello"},
@@ -94,7 +136,19 @@ storage.save_messages_bulk([
 ])
 ```

-`save_messages_bulk` handles `timestamp` and the workstream's `updated` column internally, so you don't need to compute them per row. **Verify the exact creator signature** by reading `turnstone/core/storage/_protocol.py` — table layout has shifted across migrations and the Storage protocol is the source of truth.
+`save_messages_bulk` handles `timestamp` and the workstream's `updated` column
+internally, so you don't need to compute them per row. Verify the exact
+`register_workstream` and message signatures in
+`turnstone/core/storage/_protocol.py`; the Storage protocol, not the physical
+table layout, is the source of truth.
+
+**Multi-node note:** this path assumes `get_storage()` is connected to the
+cluster's shared backend. Do not open a node-local database selected from the
+current HRW result, and do not pre-create a live session through the console
+routing proxy. After the shared-storage import commits, resolve the current
+route and open the closed workstream on that node. Any stored `node_id`
+describes creation-time placement, not a permanent shard that should receive a
+separate copy.

 ### 2. SDK `create_workstream(resume_ws=...)` (when the source is already a Turnstone workstream)

@@ -181,27 +235,48 @@ If the source thread had image or file attachments:

 - **Size limits**: images ≤ 4 MiB, text documents ≤ 512 KiB. Reject or downsample anything bigger.
 - **Allowed types**: server validates magic bytes for images and UTF-8-decodes for text. Binary blobs that aren't images won't pass.
- **Lifecycle**: pending → reserved → consumed. For imports, the cleanest path is to upload as pending and immediately consume by attaching to the relevant `conversations.id`.
+- **Blob identity**: `attachment_id` is the lowercase SHA-256 hex digest of the
+  bytes. `workstream_attachments` stores that content-addressed blob and its
+  refcount; it has no workstream or message foreign key.
+- **Message link**: the sole message-to-blob link is the ordered JSON ID list in
+  `conversations.attachments`.
+- **No persisted staging lifecycle**: pending upload bytes live only in a
+  node's in-memory attachment buffer. The old persisted
+  `pending → reserved → consumed` lifecycle does not apply to storage imports.

-Two import paths:
+For new attachment bytes, preserve row order by calling `save_message()` for
+each turn. It returns the `conversations.id`; for every attachment referenced by
+that turn, call `save_attachment()` with its content hash and bytes, then call
+`set_message_attachments(ws_id, message_id, ordered_ids)`. Each
+`save_attachment()` call accounts for one reference, while
+`set_message_attachments()` records the ordered link.

-1. **Bulk-insert + post-attach**: insert messages first, get back the assistant/user `conversations.id`, then write `workstream_attachments` rows linking the file to `message_id`.
-2. **SDK multipart create**: `create_workstream(attachments=[...], initial_message=...)` for the *first* turn only — the server reserves and consumes them onto that turn. Doesn't help for mid-thread attachments.
+`save_messages_bulk(..., attachment_ids=[...])` is appropriate only when those
+content-addressed blobs already exist: the bulk transaction retains their
+references and writes the ordered lists. Do not first call `save_attachment()`
+for a new reference and then pass the same reference to `save_messages_bulk()`;
+both paths retain it and would double-count the refcount.

-For full-history imports with multiple attachments at different turns, path (1) is the only option.
+SDK multipart create remains useful only for attachments on a new first turn;
+it publishes a live session and is not the full-history import path.

 ## Validation Checklist

 Before declaring success, verify:

 - [ ] `ws_id` is 32-char lowercase hex.
- [ ] `workstreams` row exists with the right `user_id`, `state`, `kind`.
+- [ ] The workstream remained closed and absent from every live manager while rows were written; archives stay closed and resumable imports are opened only after validation.
+- [ ] `workstreams` row exists with the right `user_id` and `kind`.
 - [ ] Conversation rows are inserted **in order** (autoincrement `id` will reflect insert order).
 - [ ] Every assistant `tool_calls[].id` has a matching `role="tool"` row with the same `tool_call_id`.
 - [ ] `tool_calls[].function.arguments` is a JSON-encoded **string**, not a parsed object.
 - [ ] First message is typically `role="user"` (not `system`) — Turnstone composes its own system prompt at runtime.
 - [ ] No empty assistant rows (`content=NULL` AND `tool_calls=NULL` is invalid).
- [ ] If multi-node: the `ws_id`'s bucket maps to a node that exists; `workstreams.node_id` matches.
+- [ ] Every attachment ID is the SHA-256 of its stored bytes; each turn's ordered IDs are in `conversations.attachments`, and blob refcounts match message references.
+- [ ] If multi-node: the row is in shared storage and the node selected by
+  `ConsoleRouter.route(ws_id)` from the current live set can load it.
+  `workstreams.node_id`, when present, is treated as a creation-time hint rather
+  than asserted equal to the current HRW result.
 - [ ] Round-trip test: run `Storage.load_messages(ws_id)` and confirm the reconstructed list matches what you inserted (modulo timestamps).

 ## Anti-patterns
@@ -211,15 +286,20 @@ Before declaring success, verify:
 - **Don't fabricate `tool_call_id`s without re-pairing.** Mismatched ids silently break the replay chain on the next turn.
 - **Don't skip the `tool_name` field on `role="tool"` rows.** Some load paths use it for display and audit; NULL there will render as "unknown tool".
 - **Don't write through the LLM (`send()` per turn) for full history.** It's expensive, rewrites assistant turns, and rate-limits will bite long imports.
+- **Don't shard imported rows by an ID prefix or a one-time HRW result.** HRW
+  uses the full ID and live membership; placement may move. In a cluster, write
+  one copy to shared storage and let request routing select the live node.

 ## Quick Reference

 | Task | Path |
 |---|---|
 | Generate ws_id | `secrets.token_hex(16)` |
-| Bulk insert messages | `Storage.save_messages_bulk(rows)` |
+| Multi-node placement | Full-ID 32-bit FNV-1a HRW over live servers; store rows once in shared storage |
+| Bulk insert attachment-free messages | `Storage.save_messages_bulk(rows)` |
+| Attach new bytes | `save_message()` → `save_attachment()` per reference → `set_message_attachments()` |
 | Archive (read-only) | `state="closed"`, skip `provider_data` |
-| Resumable | `state="idle"`, populate `provider_data` if same provider |
+| Resumable | Register closed, import and validate while unloaded, then explicitly open; populate `provider_data` if same provider |
 | Tool call id | OpenAI shape: `{"id": ..., "type": "function", "function": {"name": ..., "arguments": "<json string>"}}` |
 | Tool result row | `role="tool"`, `tool_name`, `tool_call_id`, `content` |
 | Source role → Turnstone role | See "Role Mapping" table |
@@ -228,6 +308,8 @@ Before declaring success, verify:
 ## Files to read before writing the importer

 - `turnstone/core/storage/_schema.py` — authoritative table definitions.
- `turnstone/core/storage/_protocol.py` — `save_message`, `save_messages_bulk`, `load_messages` signatures.
+- `turnstone/core/storage/_protocol.py` — `register_workstream`, message, attachment, and load signatures.
+- `turnstone/core/rendezvous.py` — authoritative full-ID FNV-1a HRW scoring.
+- `turnstone/console/router.py` — live-node discovery, override precedence, and routing behavior.
 - `turnstone/core/session.py` (around the message-save section) — how the runtime constructs in-memory message dicts; mirror this shape on import to round-trip cleanly.
 - `turnstone/api/server_schemas.py` — Pydantic shapes for the SDK paths if you go through HTTP.
@@ -54,15 +54,11 @@ docker compose exec caddy \
  cat /data/caddy/pki/authorities/local/root.crt   # import into your OS/browser
 ```

-**Can Caddy get its cert from the console's internal CA instead?** Technically
-yes — the console exposes a real ACME directory (`/acme/directory`) with
-auto-approval, so Caddy's `tls { ca http://console:8090/acme/directory }` would
-mint a cert for any name. It's not recommended as the default: lacme's ACME
-responder is built for turnstone's own client (interop with Caddy's client is
-unverified), it couples Caddy startup to the console, and the browser must trust
-a private CA either way — so it buys nothing over `tls internal`. For a publicly
-trusted cert (no warning), point Caddy at Let's Encrypt with a real domain
-instead.
+**Can Caddy get its cert from the console's internal CA instead?** Not directly.
+The console's ACME signing routes require Turnstone's rotating enrollment JWT,
+which a standard Caddy ACME issuer does not attach. Keep `tls internal`, or use a
+public ACME CA for a publicly trusted certificate. An authenticated gateway or
+Caddy plugin would be required to use Turnstone's responder.

 ---

@@ -106,13 +102,15 @@ An mTLS listener rejects plain-HTTP probes at the socket, so
 it presents the node's own cert as the client cert and pins the cluster
 CA, using the PEM files the server writes at boot under
 `$TURNSTONE_TLS_PEM_DIR` (default `<tmpdir>/turnstone-tls`). The probe
-dials `localhost` for the TLS attempt — the internal CA issues DNS SANs
-only, so a literal-IP URL would fail verification. Cert renewal rewrites
+dials `localhost` for the TLS attempt, which every service certificate carries
+as a DNS SAN. Cert renewal rewrites
 the PEM dir alongside the live listener swap, so the probe's client cert
 never outlives the served cert. With TLS disabled the plain probe succeeds
 and the PEM directory is never consulted. On bare metal with multiple
 nodes per host, set `TURNSTONE_TLS_PEM_DIR` per node (each boot clears
 stale `lacme-pem-*` dirs under its root).
+The production TLS Compose overlay inherits this healthcheck from the base
+service; it remains enabled under mTLS.

 ---

@@ -125,6 +123,13 @@ stale `lacme-pem-*` dirs under its root).
 | `tls.enabled` | `false` | Master switch for internal mTLS |
 | `tls.acme_directory` | `""` | External ACME CA URL for console frontend cert |

+### ACME topology environment
+
+| Variable | Default | Description |
+|----------|---------|-------------|
+| `TURNSTONE_ACME_EXTERNAL_URL` | request-derived | Canonical externally reachable responder base, including `/acme` (for example `http://192.0.2.1:8090/acme`). Set it on the console so advertised URLs are routable and on in-cluster clients so their enrollment JWT is allowed only at that configured destination. A public path prefix is valid only when a reverse proxy maps it to Turnstone's internal `/acme` mount. |
+| `TURNSTONE_CONSOLE_HTTP_BIND` | `127.0.0.1` | Production TLS-overlay bind for the console's plain-HTTP bootstrap/API port. For cross-host enrollment, use a trusted LAN/VPN interface and firewall it to enrolling nodes. |
+
 ### Bootstrap Config (config.toml)

 These are needed before storage is available:
@@ -144,7 +149,7 @@ sslkey = ""          # path to client key
 | CA common name | "Turnstone CA" | |
 | CA validity | 10 years | |
 | Cert validity | 48 hours | Short-lived, auto-renewed |
-| Renewal interval | 24 hours | Half of validity |
+| Renewal interval | 12 hours | Leaves retry headroom before expiry |
 | ACME auto-approve | true | Internal network, no challenge validation |

 ---
@@ -177,10 +182,14 @@ turnstone-admin tls-ca-cert --out ca.pem --console-url http://console:8080
 # Request a cert for a domain
 turnstone-admin tls-issue worker-1.internal --out /certs --console-url http://console:8080

-# List issued certs
+# List managed cluster certs
 turnstone-admin tls-list --console-url http://console:8080
 ```

+`tls-ca-cert` preserves the supplied scheme. An `https://` console URL is
+verified with the system trust store; an explicitly supplied `http://` URL is
+TOFU and prints a fingerprint that must be checked out of band.
+
 ### Console URL Discovery

 If `--console-url` is not provided, the CLI discovers it from the `services`
@@ -193,7 +202,9 @@ table in the shared database. The console registers itself on startup.
 The **TLS** tab in the console admin panel (System group) shows:
 - CA status (common name, certificate count)
 - Certificate table (domain, SANs, issued, expires)
- Force-renew and delete actions per certificate
+- Force-renew for the console-owned internal identity; remote nodes renew and
+  hot-reload their own keys
+- Delete for expired, remotely managed certificate rows

 ---

@@ -248,8 +259,15 @@ const client = new TurnstoneServer({
   `TURNSTONE_CONSOLE_URL` (a bare-metal node outside the compose network can't
   resolve the in-cluster `console` name, so it points this at the console's
   published ACME endpoint)
-3. Fetches CA root cert from `http://console/acme/ca.pem` (plain HTTP, TOFU)
-4. Requests a service cert via ACME (plain HTTP, JWS-signed). The cert's
+3. Fetches the CA root from the configured console scheme. Direct deployments
+   use `http://console/acme/ca.pem` (plain HTTP, TOFU); an explicitly configured
+   HTTPS proxy is preserved and verified with the system trust store.
+4. Requests a service cert via ACME with a dedicated, short-lived Turnstone
+   service JWT pinned to configured responder origins. lacme emits ACME JWS
+   messages, but its lightweight responder deliberately does not validate their
+   signatures or nonces; the service JWT is the enrollment authorization gate.
+   Direct HTTP bootstrap therefore still requires a trusted LAN/VPN (or an
+   independently trusted HTTPS proxy). The cert's
   primary domain / SAN is the node's **advertised host** (the host of
   `TURNSTONE_ADVERTISE_URL`, e.g. `node-1`) — the name peers actually dial,
   not the container hostname. This makes mTLS hostname verification succeed
@@ -264,7 +282,12 @@ const client = new TurnstoneServer({

 1. Read `tls.enabled` from ConfigStore
 2. Initialize CA (load from DB or generate new root key)
-3. Mount ACME responder at `/acme` (serves `/ca.pem` natively)
+3. Mount ACME responder at `/acme` (serves `/ca.pem` natively). When
+   `TURNSTONE_ACME_EXTERNAL_URL` is set, use it for every advertised directory,
+   order, authorization, and certificate URL; otherwise derive URLs from each
+   request as before. Directory, nonce, and CA bootstrap resources stay public;
+   account/order/challenge/finalization/certificate routes require the dedicated
+   enrollment service JWT.
 4. Issue console certs (internal + optional frontend)
 5. Start CA-direct auto-renewal (no network, signs directly), scoped to the
   console's own cert, plus a periodic GC that reclaims cert rows for
@@ -290,6 +313,15 @@ fronted under a second hostname). Symptom if this is wrong: the console
 dashboard shows nodes as unreachable and `openssl s_client` reports the served
 cert's SANs don't include the dialed name.

+The advertised host and extra SANs may be DNS names or literal IPv4/IPv6
+addresses. Turnstone converts IP literals to typed ACME identifiers so the
+certificate contains `IPAddress` SANs that normal IP hostname verification can
+use; DNS spelling is preserved. Bracket an IPv6 address when it appears in a URL
+(for example `TURNSTONE_ADVERTISE_URL=http://[2001:db8::10]:8080`), but use the
+bare address in `TURNSTONE_TLS_SANS`. Unspecified bind addresses (`0.0.0.0` and
+`::`) and scoped IPv6 addresses such as `fe80::1%eth0` are not certificate
+identities. Restart after changing the advertised identity or extra SANs.
+
 ### "No console service found"

 The console registers itself in the `services` table on startup. If the console
@@ -297,6 +329,22 @@ hasn't started or the registration expired (1 hour TTL), nodes can't discover
 it. Set `TURNSTONE_CONSOLE_URL` to a reachable console address (this is also how
 a bare-metal node that can't resolve the in-cluster `console` name enrolls).

+### Cross-host ACME links point at the container
+
+For a node on another host, publish port 8090 on a reachable interface and set
+`TURNSTONE_ACME_EXTERNAL_URL` on the console **and in-cluster nodes** to that full
+responder base, including `/acme` (for example
+`http://192.0.2.1:8090/acme`). The console advertises it; clients use it as a
+trusted enrollment-token destination. Keep
+`TURNSTONE_CONSOLE_URL=http://console:8090` for in-cluster service discovery.
+A remote node whose `TURNSTONE_CONSOLE_URL` already names the public origin can
+derive the same `/acme` base, but setting both values explicitly avoids drift.
+
+Bind only a trusted LAN/VPN interface and firewall it to enrolling nodes. The
+JWT authenticates the client, but a direct plain-HTTP bootstrap remains TOFU and
+does not resist an active on-path attacker. If the network is untrusted, expose
+the responder through an independently trusted HTTPS proxy instead.
+
 ### Browser HTTPS to the console

 The console serves plain HTTP (it's the ACME bootstrap endpoint — see
@@ -1,9 +1,10 @@
 # Tools Reference

-turnstone exposes 17 built-in tools plus any number of external MCP tools to the
-LLM via the OpenAI function-calling interface. Built-in tools are defined as JSON
-files under `turnstone/tools/` and loaded at startup by `turnstone/core/tools.py`.
-MCP tools are discovered from configured MCP servers at startup by
+Turnstone exposes a role-specific built-in tool surface plus any configured MCP
+tools through provider-native or OpenAI-compatible function calling. Built-in
+schemas live under `turnstone/tools/` and are loaded by
+`turnstone/core/tools.py`; metadata selects the interactive, coordinator, and
+task-agent subsets. MCP tools are discovered from configured servers by
 `turnstone/core/mcp_client.py`.

 ---
@@ -28,13 +29,19 @@ schema plus turnstone-specific metadata keys:
 }
 ```

-**Metadata keys** (stripped before sending the schema to the model):
+**Metadata keys** (stripped before sending the schema to the model; the full
+set lives in `_META_KEYS` in `turnstone/core/tools.py`):

-| Key            | Type | Meaning |
-|----------------|------|---------|
-| `task_agent`   | bool | Tool is available to task sub-agents. |
-| `auto_approve` | bool | Tool runs without user confirmation (read-only, safe operations). |
-| `primary_key`  | str  | When the model sends a bare string instead of JSON args, map it to this parameter name. |
+| Key              | Type | Meaning |
+|------------------|------|---------|
+| `task_agent`     | bool | Tool is available to task sub-agents. |
+| `coordinator`    | bool | Tool is available to coordinator sessions. Without `interactive: true` alongside it, this reads as coord-only and the tool is stripped from interactive sessions. |
+| `interactive`    | bool | Opt a `coordinator: true` tool back into interactive sessions (dual-kind tools like `memory`). |
+| `auto_approve`   | bool | Tool runs without user confirmation (read-only, safe operations). |
+| `primary_key`    | str  | When the model sends a bare string instead of JSON args, map it to this parameter name. |
+| `kind_variants`  | dict | Per-kind description / parameter-schema overlays so each session kind sees only the surface it can use (see `memory.json`). |
+| `cwd_note`       | str  | Sentence appended to the description at session build time with `{working_dir}` substituted — declare on tools whose semantics depend on the process working directory (see `bash.json`, `apply_cwd_context`). |
+| `workspace_note` | str  | Companion sentence naming the operator-configured workspace directory, `{workspace_dir}` substituted; dropped when no workspace is configured. |

 ---

@@ -44,10 +51,10 @@ schema plus turnstone-specific metadata keys:

 | Name                | Description |
 |---------------------|-------------|
-| `TOOLS`             | All 29 loaded built-in tool definitions (interactive + coordinator union). Sessions send a kind-specific subset (`INTERACTIVE_TOOLS` or `COORDINATOR_TOOLS`). |
+| `TOOLS`             | The complete loaded built-in union. Sessions send a kind-specific subset (`INTERACTIVE_TOOLS` or `COORDINATOR_TOOLS`). |
 | `TASK_AGENT_TOOLS`  | Tools with `task_agent: true` -- available to task sub-agents. Includes write operations. |
 | `TASK_AUTO_TOOLS`   | Set of all tool names with `auto_approve: true` -- used by task-agent sub-sessions to skip confirmation for matching available tools. |
-| `BUILTIN_TOOL_NAMES`| Frozenset of all 29 built-in tool names (interactive + coordinator union). Used by tool search to distinguish always-on tools from deferrable MCP tools. |
+| `BUILTIN_TOOL_NAMES`| Frozenset of the built-in union. Used by tool search to distinguish built-ins from deferrable MCP tools. |
 | `PRIMARY_KEY_MAP`   | Dict mapping tool name to its `primary_key` parameter name. |

 ---
@@ -56,7 +63,10 @@ schema plus turnstone-specific metadata keys:

 > See also: [Tool Pipeline diagram](diagrams/png/05-tool-pipeline.png)

-Tool execution follows a three-phase pipeline inside `ChatSession._execute_tools()`:
+Tool handling spans a four-phase pipeline. `ChatSession._execute_tools()` owns
+prepare, approval, and execution (phases 1–3); after it returns, the owning
+conversation loop guards the observed results and folds them into the
+trajectory (phase 4).

 ### Phase 1: Prepare

@@ -65,9 +75,8 @@ Tool execution follows a three-phase pipeline inside `ChatSession._execute_tools
 - Parses the JSON arguments (with fallback for malformed JSON).
 - If JSON parsing fails entirely, uses `PRIMARY_KEY_MAP` to map a bare string
  to the correct parameter.
- Dispatches to the matching `_prepare_{func_name}()` handler. There are 17
-  built-in tools plus `tool_search` (synthetic, client-side BM25 fallback) and
-  the generic `_prepare_mcp_tool()` handler for MCP tools.
+- Dispatches to the matching `_prepare_{func_name}()` handler, the synthetic
+  `tool_search` fallback, or the generic `_prepare_mcp_tool()` handler.
 - Validates arguments and builds a preview dict containing:
  - `call_id`, `func_name`, `header`, `preview` (for display)
  - `needs_approval` (bool)
@@ -76,7 +85,10 @@ Tool execution follows a three-phase pipeline inside `ChatSession._execute_tools

 ### Phase 2: Approve

-All prepared items are sent to the UI via `ui.approve_tools(items)`.
+Prepared items are sent to the UI via `ui.approve_tools(items)`. Several
+parallel task agents may leave independent `ApprovalCycle` objects pending on
+one workstream; each round owns a `cycle_id`, event, result, and verdict set.
+Remote clients resolve the exact round by `cycle_id` (or a member `call_id`).

 - The UI displays each tool's header and preview to the user.
 - Items where `needs_approval` is `False` (auto-approved tools) are shown
@@ -88,6 +100,10 @@ All prepared items are sent to the UI via `ui.approve_tools(items)`.
  prompt). This is per-tool, not blanket.
 - If `auto_approve` is `True` on the session (via `--skip-permissions` or workstream
  template), all tools are approved automatically.
+- When Smart Approvals are enabled, one immutable judge/settings snapshot is
+  stamped onto the whole batch. The batch auto-approves only when every gated
+  item has a qualifying verdict; partial or mixed qualification fails closed to
+  the human prompt. Stop is linearized against that terminal decision.

 ### Phase 3: Execute

@@ -107,6 +123,28 @@ Each item's `execute` callable is invoked:
  denials are tracked separately. This removes the need for text-prefix heuristics.
  Other tools deliver results atomically via
  `ui.on_tool_result(call_id, name, output, is_error=...)` only.
+
+Stop propagates to child model scopes, judges, tracked subprocess groups, and
+the approval cycles owned by the cancelled operation. Calls that definitely
+never started receive `EffectStatus.none`; an interrupted call whose external
+outcome was not observed receives `unknown`, `partial`, or `rolled_back` as
+appropriate. These typed receipts preserve effect truth across storage/replay
+without exposing unreviewed model output as a tool result.
+
+### Phase 4: Guard and atomic fold
+
+After `_execute_tools()` returns, the main `send()` loop compacts/truncates
+completed results to the remaining shared budget and then runs the heuristic
+and optional LLM output guard. The task-agent loop deliberately guards the
+observed raw output before applying its size cap, so truncation cannot hide a
+sensitive result from that check.
+
+After guard work, the owning loop rechecks generation ownership. On the main
+conversation path, one generation-fenced commit appends the complete
+tool-result block, advisories, feedback, and queued user turns; its durable
+records run in FIFO order outside the lifecycle lock. A force-cancelled
+predecessor can therefore finish external cleanup, but cannot fold late results
+into its successor's trajectory.
 ---

 ## Tool Approval Flow
@@ -114,7 +152,7 @@ Each item's `execute` callable is invoked:
 **Auto-approved** (no user confirmation needed at runtime):
 - `read_file` -- reads files, no side effects
 - `search` -- grep-style search, no side effects
- `memory` -- structured persistent memory (save/search/delete/list)
+- `memory` -- structured persistent memory (save/get/search/delete/list)
 - `recall` -- searches conversation history
 - `notify` -- sends notifications to linked channels (time-sensitive, auto-approved for urgency)

@@ -289,7 +327,7 @@ Fetch a URL and extract specific information from it.
 | `url`      | string | yes      | The URL to fetch (must start with `http://` or `https://`). |
 | `question` | string | yes      | What to extract or answer from the page content. |

- **What it does**: Fetches the URL, strips HTML to plain text, and uses the LLM to extract the answer to the question from the page content. Every redirect hop is SSRF-screened before it is requested. Private/internal addresses are refused by default; enable `tools.allow_private_network` (console Settings → Tools) to make them approvable for self-hosted setups whose services live on the local network — the approval prompt marks such requests, and a public site redirecting into private space is refused regardless.
+- **What it does**: Fetches the URL, strips HTML to plain text, and uses the LLM to extract the answer to the question from the page content. Every redirect hop is SSRF-screened before it is requested. Private/internal addresses are refused by default; enable `tools.allow_private_network` (console Settings → Tools) to make them approvable for self-hosted setups whose services live on the local network — the approval prompt marks such requests, and a public site redirecting into private space is refused regardless. Cloud metadata endpoints and link-local, multicast and reserved addresses are refused even with the opt-in enabled, including as a redirect target from a private address you approved. An address is judged by what it actually reaches, so an IPv6 transition address (NAT64, 6to4, Teredo) wrapping an internal IPv4 is treated exactly as that IPv4 would be.
 - **Auto-approve**: No -- requires user confirmation (makes network requests).
 - **Agent availability**: `task_agent`.

@@ -395,7 +433,7 @@ Delegate a general-purpose task to an autonomous sub-agent.
 |-----------|--------|----------|-------------|
 | `prompt`  | string | yes      | Complete task description for the sub-agent. |

- **What it does**: Spawns a sub-agent that inherits the `TASK_AGENT_TOOLS` set (read, write, edit, search, bash, web tools, memory tools). The sub-agent runs autonomously to completion. Use for work that requires file modifications or command execution.
+- **What it does**: Spawns a sub-agent that inherits the `TASK_AGENT_TOOLS` set (read, write, edit, search, bash, and web tools). The sub-agent runs autonomously to completion. Use for work that requires file modifications or command execution.
 - **Auto-approve**: No -- requires user confirmation.
 - **Agent availability**: Top-level only.

@@ -409,18 +447,26 @@ Structured persistent memory across sessions with typed, scoped entries.

 | Parameter     | Type    | Required | Description |
 |---------------|---------|----------|-------------|
-| `action`      | string  | yes      | `save`, `search`, `delete`, or `list`. |
-| `name`        | string  | save/delete | Short snake_case identifier for the memory. |
+| `action`      | string  | yes      | `save`, `get`, `search`, `delete`, or `list`. |
+| `name`        | string  | save/get/delete | Short snake_case identifier for the memory. |
 | `content`     | string  | save     | Memory content to store. |
-| `description` | string  | no       | Short description for relevance matching (recommended for `save`). |
-| `type`        | string  | no       | Memory type: `user`, `project`, `feedback`, or `reference`. Default: `project`. |
-| `scope`       | string  | no       | Memory scope: `global`, `workstream`, or `user`. Default: `global`. |
+| `description` | string  | save     | Non-empty description for relevance matching; required on create and update. |
+| `type`        | string  | no       | Memory type: `user`, `general`, `feedback`, or `reference`. Default: `general`. |
+| `scope`       | string  | no       | Memory scope: `global`, `workstream`, `user`, `coordinator`, or `project`. See defaults below. |
 | `query`       | string  | search   | Search query for finding memories. |
 | `limit`       | integer | no       | Max results for `search` or `list`. Default: 20. |

- **What it does**: Manages structured persistent memories in the database. Memories persist across sessions, have a type classification (user preferences, project knowledge, feedback, reference material) and a scope (global across all workstreams, private to a workstream, or following a user). Relevant memories are included in the system prompt on startup.
+- **What it does**: Manages structured persistent memories in the database.
+  Memories persist across sessions, have a type classification, and live in a
+  role-specific visible scope. Unscoped `save`/`get`/`delete` resolve to one
+  target: the attached active project, otherwise `global` for an interactive
+  session or `coordinator` for a coordinator. Read-only project access permits
+  `get` but makes `save`/`delete` fail without falling back. A valid explicit
+  scope selects exactly that scope. Unscoped `search`/`list` cover all visible
+  scopes; use the displayed scope when following a result with `get` or
+  `delete`.
 - **Auto-approve**: Yes.
- **Agent availability**: Not available to sub-agents (top-level only).
+- **Agent availability**: Not available to task agents.

 ---

@@ -435,7 +481,7 @@ Search conversation history for past messages and tool results.

 - **What it does**: Searches conversation history across sessions using FTS5 full-text search. Returns matching messages, tool calls, and tool results with timestamps and workstream context.
 - **Auto-approve**: Yes.
- **Agent availability**: Not available to sub-agents (top-level only).
+- **Agent availability**: Not available to task agents.

 ---

@@ -571,7 +617,11 @@ pre-configure skills at workstream creation.

 ---

-## Summary Table
+## Interactive Tool Summary
+
+This table describes the ordinary interactive surface. Coordinator sessions
+receive their delegation/lifecycle tools instead, and task agents receive the
+metadata-selected `TASK_AGENT_TOOLS` subset.

 | Tool         | Category   | Auto-approve | task_agent | primary_key |
 |--------------|------------|--------------|------------|-------------|
@@ -692,7 +742,7 @@ MCP-compatible service.
 3. **Schema conversion**: Each MCP tool's `inputSchema` is converted to OpenAI
   function-calling format. The tool name is prefixed: `mcp__{server}__{tool}`.

-4. **Merging**: MCP tools are appended after the 17 built-in tools via
+4. **Merging**: MCP tools are appended after the role's built-in tools via
   `merge_mcp_tools()`. Built-in tools appear first, giving them natural LLM priority.
   When dynamic tool search is active, MCP tools are deferred rather than directly
   visible -- the model discovers them via search as needed (see
@@ -779,7 +829,10 @@ MCP tool lists stay up-to-date without restart through two mechanisms:
 1. **Push notifications** -- MCP servers that declare `tools.listChanged: true` in
   their capabilities send `notifications/tools/list_changed` when their tool list
   changes. `MCPClientManager` registers a `message_handler` on each `ClientSession`
-   that triggers an immediate refresh for that server.
+   that triggers an immediate refresh for that server (debounced per server and
+   notification kind, and run off the receive loop). A refresh that fails while
+   the connection stays up is retried automatically on the next health-loop tick
+   until one completes.

 2. **Manual** -- `/mcp refresh` re-fetches tools from all servers immediately.
   `/mcp refresh <server>` targets a single server. If a server has disconnected,
@@ -787,6 +840,10 @@ MCP tool lists stay up-to-date without restart through two mechanisms:
   same controls (refresh / reconnect buttons per server) for cluster-wide
   fan-out.

+Reconnects (health-loop, dispatch-driven, or operator-forced) always end in a
+full catalog rediscovery, so a server that changed its tools while disconnected
+comes back current.
+
 When tools change, `MCPClientManager` rebuilds its merged tool list using copy-on-write
 (new list/dict objects assigned atomically) and notifies all active `ChatSession`
 instances via registered listener callbacks. Each session rebuilds its `_tools`,
@@ -822,6 +879,13 @@ capabilities for the `resources` capability. For servers that declare it:
 2. `list_resource_templates` fetches URI templates (parameterized patterns like
   `db://tables/{table}/rows/{id}`).

+The protocol advertises both lists through one aggregate `resources`
+capability, so a server may implement only one of them. If either request
+returns the JSON-RPC `Method not found` code (`-32601`), turnstone treats that
+half of the catalog as empty and keeps the other half; authentication,
+validation, transport, and all other discovery errors still fail the
+connection or refresh.
+
 Both are stored as `{uri, name, description, mimeType, server}` dicts and
 merged into a unified catalog.

@@ -857,13 +921,16 @@ catalog.

 ### Refresh

-Resource lists stay current through the same three-tier mechanism as tool lists:
+Resource lists stay current through the same mechanisms as tool lists:

 1. **Push** -- Servers declaring `resources.listChanged: true` send
-   `notifications/resources/list_changed`, triggering an immediate refresh.
-2. **Periodic** -- Servers without push are polled on the configured refresh
-   interval (default 4 hours, same timer as tools).
-3. **Manual** -- `/mcp refresh` re-fetches resources alongside tools.
+   `notifications/resources/list_changed`, triggering an immediate refresh
+   (with the same failed-refresh retry on the health-loop tick).
+2. **Manual** -- `/mcp refresh` re-fetches resources alongside tools.
+
+Servers without push support are refreshed whenever they reconnect (every
+reconnect ends in full rediscovery) or when an operator refreshes manually;
+there is no periodic polling.

 ---

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"

 [project]
 name = "turnstone"
-version = "1.8.0a1"
+version = "1.8.0a7"
 description = "Multi-node AI orchestration platform with tool use, agent routing, and cluster simulation."
 readme = "README.md"
 license = "Apache-2.0"
@@ -23,9 +23,14 @@ classifiers = [
    "Topic :: Scientific/Engineering :: Artificial Intelligence",
 ]
 dependencies = [
-    "openai>=2.45",  # GPT-5.6: typed reasoning.mode, prompt_cache_options, and cache_write_tokens
-    "anthropic>=0.108",  # claude-fable-5 support; hard runtime floor is 0.105 (mid-conversation system blocks)
+    # Version 3 moves the default transport to HTTPX2. Keep major upgrades
+    # deliberate because the stream retry boundary depends on that contract.
+    "openai>=3,<4",
+    "anthropic>=0.117",  # tracks the release current at claude-opus-5 onboarding; hard runtime floor is still 0.105 (mid-conversation system blocks) — Opus 5 itself needs no new SDK surface (model ids are opaque strings; "refusal" has been in the StopReason literal since ~0.95). Raise this when adopting fast mode / server-side fallbacks / advisor / mid-conversation tool changes, which DO need newer typed params.
    "httpx>=0.28",
+    # Direct because the provider boundary catches this exception family;
+    # OpenAI v3's transitive dependency alone is not an import contract.
+    "httpx2>=2.7,<3",
    "mcp>=1.27,<2",  # v2 is a breaking rewrite (2.0.0a1 live 2026-06-11; stable ~2026-07-27) — streamablehttp_client removed, 2-tuple transport, snake_case types; migrate deliberately
    "starlette>=1.3.1",  # CVE-2026-54282 (path->authority host spoof) + CVE-2026-54283 (url-encoded form DoS); supersedes the PYSEC-2026-161 host-header path-injection floor
    "uvicorn>=0.34",
@@ -42,7 +47,9 @@ dependencies = [
    "PyJWT>=2.8",
    "bcrypt>=4.0",
    "cryptography>=48.0.1",  # GHSA-537c-gmf6-5ccf: PyPI wheels <48.0.1 bundle a vulnerable statically-linked OpenSSL (2026-06-09 secadv)
-    "lacme>=1.0.5",
+    # Core mTLS/ACME contract. 1.1 moves lacme onto HTTPX2; 1.2 adds typed IP
+    # identifiers and retains the external responder URL for cross-host enrollment.
+    "lacme>=1.2,<2",
    "python-frontmatter>=1.0",
    "pypdfium2>=4",  # PDF text-extract + rasterize for models without native PDF input (core/pdf.py)
    "pillow>=10",  # PNG encoding for the PDF->images rasterize fallback (vision models, core/pdf.py)
@@ -88,10 +95,10 @@ include = [
  "turnstone/console/static/coordinator/*.js",
  "turnstone/shared_static/*.css",
  "turnstone/shared_static/*.js",
-  "turnstone/shared_static/katex-0.17.0/**/*",
+  "turnstone/shared_static/katex-0.18.4/**/*",
  "turnstone/shared_static/hljs-11.11.1/**/*",
-  "turnstone/shared_static/mermaid-11.16.0/**/*",
-  "turnstone/shared_static/hls-1.6.16/**/*",
+  "turnstone/shared_static/mermaid-11.16.1/**/*",
+  "turnstone/shared_static/hls-1.6.17/**/*",
  "turnstone/sdk/py.typed",
  "turnstone/deploy/*.yaml",
  "turnstone/deploy/Caddyfile",
@@ -103,8 +110,14 @@ testpaths = ["tests"]
 markers = [
    "live: requires a running LLM backend",
    "allow_thread_leak: test intentionally leaves a background thread running (opts out of the leaked-thread guard)",
+    "e2e_recovery: opt-in end-to-end SSE recovery harness (real server + real SSE consumers, scripted provider — NOT live, no LLM backend needed); tens of seconds each. CI lanes run ``-m 'not live and not e2e_recovery'``; select with ``-m e2e_recovery``.",
 ]
 filterwarnings = [
+    # The MCP v1 FastMCP integration fixture rebuilds its incomplete generic
+    # Settings model before construction. Keep the new pydantic-settings 2.15
+    # diagnostic fatal so removing that compatibility step cannot regress
+    # into a warning hidden in the full-suite summary.
+    "error:Field 'lifespan' has an incomplete definition",
    # mcp v1 deprecates streamablehttp_client for an entry point whose call
    # shape only settles in v2 — adoption rides the deliberate v2 migration
    # (pin capped <2); silence exactly this message until then.
@@ -186,10 +199,6 @@ ignore_missing_imports = true
 module = ["frontmatter", "frontmatter.*"]
 ignore_missing_imports = true

-[[tool.mypy.overrides]]
-module = ["lacme", "lacme.*"]
-ignore_missing_imports = true
-
 [[tool.mypy.overrides]]
 module = ["pypdfium2", "pypdfium2.*"]
 ignore_missing_imports = true
@@ -0,0 +1,458 @@
+#!/usr/bin/env python3
+"""Browser regression for the two-layer frontend cache contract.
+
+This harness serves a versioned entry module which imports the real,
+unversioned ``shared/interactive.js`` module.  It loads an old pane build in a
+real Chrome profile, switches the server to a same-version replacement whose
+pane has the same byte length and mtime, then performs a normal browser reload
+without clearing or disabling the HTTP cache.
+
+The old fixture advertises ``user_turn=0&tool_turn=0`` in its EventSource URL;
+the current source advertises both capabilities as ``1``.  A passing run
+therefore proves both layers of the contract:
+
+* the package-versioned entry URL revalidates and may return 304; and
+* its unversioned transitive pane import revalidates by content and returns the
+  current bytes rather than surviving from the prior build.
+
+The page also loads representative KaTeX, Highlight.js, and HLS.js assets.
+Their installed version directories are discovered from ``shared_static`` at
+runtime, so a normal vendor-version bump requires no harness edit; reload must
+reuse them under the immutable policy.
+
+Usage::
+
+    uv run python scripts/asset_cache_e2e.py
+"""
+
+from __future__ import annotations
+
+import contextlib
+import json
+import os
+import shutil
+import socket
+import sys
+import tempfile
+import threading
+import time
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Any
+
+ROOT = Path(__file__).resolve().parents[1]
+if str(ROOT) not in sys.path:
+    sys.path.insert(0, str(ROOT))
+
+from recovery_e2e import CDP, _find_chrome, _launch_chrome, _page_ws_url  # noqa: E402
+
+from turnstone import __version__  # noqa: E402
+from turnstone.core.web_helpers import (  # noqa: E402
+    RevalidatingStaticFiles,
+    version_html,
+)
+
+_PAGE_HTML = """<!doctype html>
+<html>
+  <head>
+    <meta charset="utf-8">
+    <title>ASSET-CACHE-PENDING</title>
+    <!-- VENDORED_ASSET_TAGS -->
+    <script>
+      window.__assetCacheUrls = [];
+      class RecordingEventSource {
+        static CONNECTING = 0;
+        static OPEN = 1;
+        static CLOSED = 2;
+        constructor(url) {
+          this.url = String(url);
+          this.readyState = RecordingEventSource.CONNECTING;
+          window.__assetCacheUrls.push(this.url);
+        }
+        close() {
+          this.readyState = RecordingEventSource.CLOSED;
+        }
+      }
+      window.EventSource = RecordingEventSource;
+      window.addEventListener("error", (event) => {
+        document.title = "ASSET-CACHE-FAILED-" + String(event.message || "script").slice(0, 80);
+      });
+      window.addEventListener("unhandledrejection", (event) => {
+        document.title = "ASSET-CACHE-FAILED-" + String(event.reason || "promise").slice(0, 80);
+      });
+    </script>
+  </head>
+  <body>
+    <main id="pane"></main>
+    <script type="module" src="/static/asset_cache_boot.js"
+            onerror="document.title='ASSET-CACHE-FAILED-module-load'"></script>
+  </body>
+</html>
+"""
+
+_BOOT_JS = """import { InteractivePane } from "/shared/interactive.js";
+
+const fixtureWorkstreamId = "00000000-0000-0000-0000-000000000001";
+const pane = new InteractivePane(fixtureWorkstreamId, { base: "" });
+document.getElementById("pane").appendChild(pane.el);
+pane.connectSSE(fixtureWorkstreamId);
+const url = window.__assetCacheUrls.at(-1) || "";
+const generation = url.includes("user_turn=1") && url.includes("tool_turn=1")
+  ? "CURRENT"
+  : url.includes("user_turn=0") && url.includes("tool_turn=0")
+    ? "OLD"
+    : "FAILED-CAPABILITIES";
+window.__assetCacheResult = { generation, url };
+document.title = "ASSET-CACHE-" + generation;
+"""
+
+
+@dataclass
+class CacheState:
+    phase: str = "old"
+    requests: list[dict[str, Any]] = field(default_factory=list)
+    lock: threading.Lock = field(default_factory=threading.Lock)
+
+    def record(self, item: dict[str, Any]) -> None:
+        with self.lock:
+            self.requests.append(item)
+
+    def matching(self, phase: str, path: str) -> list[dict[str, Any]]:
+        with self.lock:
+            return [
+                item for item in self.requests if item["phase"] == phase and item["path"] == path
+            ]
+
+
+class SwitchingStaticFiles:
+    """Select one immutable build snapshot at request dispatch time."""
+
+    def __init__(self, state: CacheState, old_dir: Path, current_dir: Path) -> None:
+        self._state = state
+        self._apps = {
+            "old": RevalidatingStaticFiles(directory=str(old_dir)),
+            "current": RevalidatingStaticFiles(directory=str(current_dir)),
+        }
+
+    async def __call__(self, scope: Any, receive: Any, send: Any) -> None:
+        await self._apps[self._state.phase](scope, receive, send)
+
+
+class RecordingApp:
+    """Record static HTTP validators and status without perturbing streaming."""
+
+    def __init__(self, app: Any, state: CacheState) -> None:
+        self._app = app
+        self._state = state
+
+    async def __call__(self, scope: Any, receive: Any, send: Any) -> None:
+        path = str(scope.get("path", ""))
+        if scope.get("type") != "http" or not path.startswith(("/static/", "/shared/")):
+            await self._app(scope, receive, send)
+            return
+
+        phase = self._state.phase
+        request_headers = {
+            key.decode("latin-1").lower(): value.decode("latin-1")
+            for key, value in scope.get("headers", [])
+        }
+
+        async def record_send(message: dict[str, Any]) -> None:
+            if message["type"] == "http.response.start":
+                response_headers = {
+                    key.decode("latin-1").lower(): value.decode("latin-1")
+                    for key, value in message.get("headers", [])
+                }
+                self._state.record(
+                    {
+                        "phase": phase,
+                        "path": path,
+                        "query": scope.get("query_string", b"").decode("latin-1"),
+                        "if_none_match": request_headers.get("if-none-match"),
+                        "status": message["status"],
+                        "etag": response_headers.get("etag"),
+                        "cache_control": response_headers.get("cache-control"),
+                    }
+                )
+            await send(message)
+
+        await self._app(scope, receive, record_send)
+
+
+def _old_interactive(current: bytes) -> bytes:
+    old = current
+    for needle, replacement in (
+        (b'"user_turn=1"', b'"user_turn=0"'),
+        (b'"&tool_turn=1"', b'"&tool_turn=0"'),
+    ):
+        if old.count(needle) != 1:
+            raise RuntimeError(f"expected exactly one {needle.decode()} capability literal")
+        old = old.replace(needle, replacement, 1)
+    if len(old) != len(current):
+        raise AssertionError("old and current pane fixtures must have equal byte length")
+    return old
+
+
+def _discover_vendor_assets(source_shared: Path) -> tuple[str, ...]:
+    selected = (
+        ("katex", "katex.min.css"),
+        ("katex", "katex.min.js"),
+        ("hljs", "highlight.min.js"),
+        ("hls", "hls.min.js"),
+    )
+    paths = []
+    for library, filename in selected:
+        matches = sorted(source_shared.glob(f"{library}-*/{filename}"))
+        if not matches:
+            raise RuntimeError(f"no vendored {library} asset named {filename} was found")
+        paths.extend(f"/shared/{match.relative_to(source_shared).as_posix()}" for match in matches)
+    return tuple(paths)
+
+
+def _vendor_tags(vendor_paths: tuple[str, ...]) -> str:
+    tags = []
+    for path in vendor_paths:
+        if path.endswith(".css"):
+            tags.append(f'<link rel="stylesheet" href="{path}">')
+        else:
+            tags.append(f'<script src="{path}"></script>')
+    return "\n    ".join(tags)
+
+
+def _prepare_builds(scratch: Path) -> tuple[Path, Path, Path, tuple[str, ...]]:
+    import turnstone
+
+    package_dir = Path(turnstone.__file__).resolve().parent
+    source_shared = package_dir / "shared_static"
+    vendor_paths = _discover_vendor_assets(source_shared)
+    old_shared = scratch / "old" / "shared"
+    current_shared = scratch / "current" / "shared"
+    static_dir = scratch / "static"
+    shutil.copytree(source_shared, old_shared)
+    shutil.copytree(source_shared, current_shared)
+    static_dir.mkdir()
+    (static_dir / "asset_cache_boot.js").write_text(_BOOT_JS, encoding="utf-8")
+
+    current_asset = current_shared / "interactive.js"
+    old_asset = old_shared / "interactive.js"
+    current = current_asset.read_bytes()
+    old_asset.write_bytes(_old_interactive(current))
+
+    # Reproduce the metadata collision which defeated Starlette's default ETag.
+    fixed_mtime_ns = 1_700_000_000_123_456_789
+    for asset in (old_asset, current_asset):
+        os.utime(asset, ns=(fixed_mtime_ns, fixed_mtime_ns))
+    old_stat = old_asset.stat()
+    current_stat = current_asset.stat()
+    if (old_stat.st_size, old_stat.st_mtime_ns) != (
+        current_stat.st_size,
+        current_stat.st_mtime_ns,
+    ):
+        raise AssertionError("pane fixture size/mtime collision was not preserved")
+    return old_shared, current_shared, static_dir, vendor_paths
+
+
+def _make_app(
+    state: CacheState,
+    old_dir: Path,
+    current_dir: Path,
+    static_dir: Path,
+    vendor_paths: tuple[str, ...],
+) -> Any:
+    from starlette.applications import Starlette
+    from starlette.responses import HTMLResponse
+    from starlette.routing import Mount, Route
+
+    async def page(_request: Any) -> HTMLResponse:
+        page_html = _PAGE_HTML.replace("<!-- VENDORED_ASSET_TAGS -->", _vendor_tags(vendor_paths))
+        return HTMLResponse(
+            version_html(page_html),
+            headers={"Cache-Control": "no-store"},
+        )
+
+    app = Starlette(
+        routes=[
+            Route("/asset-cache-e2e", page),
+            Mount(
+                "/static",
+                app=RevalidatingStaticFiles(directory=str(static_dir)),
+            ),
+            Mount(
+                "/shared",
+                app=SwitchingStaticFiles(state, old_dir, current_dir),
+            ),
+        ]
+    )
+
+    return RecordingApp(app, state)
+
+
+def _start_server(app: Any) -> tuple[Any, threading.Thread, socket.socket, str]:
+    import uvicorn
+
+    sock = socket.socket()
+    sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
+    sock.bind(("127.0.0.1", 0))
+    sock.listen(128)
+    port = int(sock.getsockname()[1])
+    server = uvicorn.Server(
+        uvicorn.Config(app, host="127.0.0.1", port=port, log_level="warning", lifespan="off")
+    )
+    thread = threading.Thread(
+        target=server.run,
+        kwargs={"sockets": [sock]},
+        name="asset-cache-e2e-server",
+        daemon=True,
+    )
+    thread.start()
+    deadline = time.monotonic() + 10
+    while not server.started and thread.is_alive() and time.monotonic() < deadline:
+        time.sleep(0.05)
+    if not server.started:
+        server.should_exit = True
+        thread.join(timeout=2)
+        sock.close()
+        raise RuntimeError("asset cache test server did not start")
+    return server, thread, sock, f"http://127.0.0.1:{port}"
+
+
+def _wait_for_generation(cdp: CDP, expected: str, timeout: float = 20) -> dict[str, str]:
+    deadline = time.monotonic() + timeout
+    last_title = ""
+    while time.monotonic() < deadline:
+        last_title = cdp.title()
+        result = cdp.evaluate("window.__assetCacheResult || null")
+        if isinstance(result, dict) and result.get("generation") == expected:
+            return {"generation": str(result["generation"]), "url": str(result["url"])}
+        if last_title.startswith("ASSET-CACHE-FAILED"):
+            raise RuntimeError(last_title)
+        time.sleep(0.1)
+    raise TimeoutError(f"expected {expected}, last title was {last_title!r}")
+
+
+def _one(state: CacheState, phase: str, path: str) -> dict[str, Any]:
+    requests = state.matching(phase, path)
+    if len(requests) != 1:
+        raise AssertionError(f"expected one {phase} request for {path}, got {requests!r}")
+    return requests[0]
+
+
+def _verify_trace(state: CacheState, vendor_paths: tuple[str, ...]) -> tuple[str, list[str]]:
+    entry_path = "/static/asset_cache_boot.js"
+    pane_path = "/shared/interactive.js"
+    old_entry = _one(state, "old", entry_path)
+    current_entry = _one(state, "current", entry_path)
+    old_pane = _one(state, "old", pane_path)
+    current_pane = _one(state, "current", pane_path)
+
+    expected_query = f"v={__version__}"
+    if old_entry["query"] != expected_query or current_entry["query"] != expected_query:
+        raise AssertionError("entry URL did not retain the same package version across builds")
+    if old_entry["status"] != 200 or old_pane["status"] != 200:
+        raise AssertionError("old build did not populate the browser cache")
+    if current_entry["status"] != 304 or not current_entry["if_none_match"]:
+        raise AssertionError(f"versioned entry did not revalidate to 304: {current_entry!r}")
+    if current_pane["status"] != 200 or not current_pane["if_none_match"]:
+        raise AssertionError(
+            f"transitive pane did not revalidate to current bytes: {current_pane!r}"
+        )
+    if old_pane["etag"] == current_pane["etag"]:
+        raise AssertionError("content-derived pane validators did not change")
+    if current_pane["cache_control"] != "no-cache":
+        raise AssertionError("transitive pane lost its revalidation policy")
+
+    vendor_trace = []
+    immutable = "public, max-age=31536000, immutable"
+    for path in vendor_paths:
+        old_vendor = _one(state, "old", path)
+        if old_vendor["query"] or old_vendor["status"] != 200:
+            raise AssertionError(f"versioned vendor URL was rewritten or failed: {old_vendor!r}")
+        if old_vendor["cache_control"] != immutable:
+            raise AssertionError(f"versioned vendor asset was not immutable: {old_vendor!r}")
+        revisits = state.matching("current", path)
+        if revisits:
+            if len(revisits) != 1 or revisits[0]["status"] not in (200, 304):
+                raise AssertionError(f"unexpected vendor reload trace: {revisits!r}")
+            if revisits[0]["cache_control"] != immutable:
+                raise AssertionError(f"vendor reload lost immutable policy: {revisits[0]!r}")
+            vendor_trace.append(f"{path}: revisited-{revisits[0]['status']}")
+        else:
+            vendor_trace.append(f"{path}: cache-hit")
+    verdict = f"ASSET-CACHE-READY-entry304-pane200-user1-tool1-vendor{len(vendor_paths)}"
+    return verdict, vendor_trace
+
+
+def _stop_process(proc: Any) -> None:
+    if proc.poll() is not None:
+        return
+    proc.terminate()
+    with contextlib.suppress(Exception):
+        proc.wait(timeout=5)
+    if proc.poll() is None:
+        proc.kill()
+        with contextlib.suppress(Exception):
+            proc.wait(timeout=2)
+
+
+def main() -> int:
+    chrome = _find_chrome()
+    if not chrome:
+        print("ASSET-CACHE-FAILED-no-chrome")
+        return 2
+
+    with tempfile.TemporaryDirectory(prefix="turnstone-asset-cache-e2e-") as raw_scratch:
+        scratch = Path(raw_scratch)
+        old_dir, current_dir, static_dir, vendor_paths = _prepare_builds(scratch)
+        state = CacheState()
+        server, server_thread, sock, base_url = _start_server(
+            _make_app(state, old_dir, current_dir, static_dir, vendor_paths)
+        )
+        chrome_proc = None
+        cdp = None
+        try:
+            chrome_proc, cdp_port = _launch_chrome(chrome, scratch / "chrome-profile")
+            cdp = CDP(_page_ws_url(cdp_port))
+            cdp.cmd("Page.enable")
+            cdp.cmd("Runtime.enable")
+            cdp.cmd("Network.enable")
+            cdp.cmd("Page.navigate", {"url": f"{base_url}/asset-cache-e2e"})
+            old_result = _wait_for_generation(cdp, "OLD")
+
+            state.phase = "current"
+            cdp.cmd("Page.reload", {"ignoreCache": False})
+            current_result = _wait_for_generation(cdp, "CURRENT")
+
+            if "user_turn=0" not in old_result["url"] or "tool_turn=0" not in old_result["url"]:
+                raise AssertionError(f"old pane did not expose old capabilities: {old_result!r}")
+            if (
+                "user_turn=1" not in current_result["url"]
+                or "tool_turn=1" not in current_result["url"]
+            ):
+                raise AssertionError(
+                    f"reloaded pane did not expose current capabilities: {current_result!r}"
+                )
+
+            verdict, vendor_trace = _verify_trace(state, vendor_paths)
+            cdp.evaluate(f"document.title = {json.dumps(verdict)}")
+            print(verdict)
+            print(f"  old EventSource:     {old_result['url']}")
+            print(f"  current EventSource: {current_result['url']}")
+            for item in vendor_trace:
+                print(f"  vendor: {item}")
+            return 0
+        except Exception as exc:
+            print(f"ASSET-CACHE-FAILED-{type(exc).__name__}: {exc}")
+            return 1
+        finally:
+            if cdp is not None:
+                cdp.close()
+            if chrome_proc is not None:
+                _stop_process(chrome_proc)
+            server.should_exit = True
+            server_thread.join(timeout=10)
+            with contextlib.suppress(OSError):
+                sock.close()
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
@@ -45,6 +45,22 @@ Shell harness (?split=): right (default) · down · three · none — boots the
  document.title stamps SPLIT-READY-<visible cells> on success and
  SPLIT-FAILED-<reason> when a driven split was denied — judge the focused
  cell's top accent bar, the separators, and the .shown tab marker.
+Proxy-brand harness (/proxybrand/livepass.html): back-to-console from a
+  PROXIED node view, driven end to end.  An iframe hosts a node page built
+  from the REAL shell.js rail plus the REAL _JS_PROXY_SHIM (read out of
+  turnstone/console/server.py by text, never imported -- scripts/ has no
+  sys.path guard, so an import would silently pick up site-packages).  The
+  host clicks the brand's child span and, because the shim navigates the
+  FRAME away, reads the frame's post-navigation location from the surviving
+  top page.  document.title stamps PROXYBRAND-READY, or
+  PROXYBRAND-FAILED-<reason>: sub-not-repointed-server (nothing wired),
+  showhome-also-ran (shell.js won the click), nav-<path> (went somewhere
+  other than the console root), sub-not-console-<text>, aria-not-repointed,
+  no-navigation, no-brand, no-sub.  Needs --virtual-time-budget=9000;
+  there is nothing to screenshot.  Read the verdict from <title> --
+  both literals also appear in the host page's inline script, so a bare
+  grep over --dump-dom output false-positives.
+
 Attachments harness (/attachments/livepass.html): the composer attachment
  chips + the sent-message attachment pills, both driven through the REAL
  code paths — createAttachmentController.rehydrate() builds the chips and
@@ -58,6 +74,21 @@ Attachments harness (/attachments/livepass.html): the composer attachment
  thumbnail crop/size, the native audio-control fit at the constrained
  height, the snippet contrast, and how a long filename behaves at the
  340px chip cap.
+Paste-over-HTTP harness (/paste/livepass.html): the REAL Composer's
+  large-text paste path, exercised with a browser-generated, trusted paste
+  event on an explicitly non-secure HTTP origin.  No
+  ``navigator.clipboard`` stub, synthetic ``ClipboardEvent``, or
+  secure-context override is involved.  The page requires
+  ``window.isSecureContext === false``, ``event.isTrusted``, a ``text/plain``
+  clipboard item, canceled inline insertion, and an exact
+  ``pasted-text.txt`` File round-trip before stamping ``PASTE-HTTP-READY``.
+  It fails closed as ``PASTE-HTTP-FAILED-<reason>``.  Serve beyond loopback,
+  open the page by a LAN address (localhost and 127.0.0.1 are treated as
+  trustworthy origins by browsers), then use the browser's normal Copy and
+  Paste commands:
+
+    python3 scripts/livepass.py --serve 8950 --bind 0.0.0.0
+
 Task-agent harness (/taskagent/livepass.html): the task_agent card — a task
  agent's sub-tool steps nested under its conversation row, driven through the
  REAL InteractivePane.handleEvent (parent tool_pending/tool_info -> child
@@ -79,6 +110,30 @@ Task-agent harness (/taskagent/livepass.html): the task_agent card — a task
  success, TASKAGENT-FAILED-... / TASKAGENT-ERROR when routing breaks, so a
  broken card can't screenshot green.

+Copy harness (/copy/livepass.html): the copy-to-clipboard affordances — the
+  per-bubble copy button in .msg-actions and the floating block-copy button
+  over hovered fences / mermaid diagrams / tables (pointer-only; keyboard
+  copies with Enter on the focused block) — driven through the REAL
+  InteractivePane (replayHistory plus a live handleEvent stream turn, so the
+  retry-holder buttons coexist with the persistent copy buttons on the last
+  bubble; the turn ends idle, matching the affordances' idle-only gate).
+  navigator.clipboard is stubbed to a recorder, hover/focus/keys are
+  dispatched synthetically, and every copied payload is compared byte-exact
+  against the SOURCE (fences, pipes, mermaid text, the bubble's raw
+  markdown).  + &theme=light.  document.title stamps
+  COPY-READY-<bubbles>-<blocks> only when every probe copied exact source;
+  COPY-FAILED-<reason> otherwise.  &kbd=1 probes the KEYBOARD path: focus a
+  block, dispatch Enter — the block's source lands on the clipboard, the
+  block carries the outcome flash class, and the floating button stays out
+  of it — stamps COPY-KBD-READY / COPY-KBD-FAILED-<step>.
+  Screenshot states: &flash=1 (visual-only
+  run — no probes; floating button + ✓ state on the fence, holder bar
+  revealed via focus) and &bare=1 (single hover, no decoration).  Known
+  capture artifact: the DARK-theme &flash=1 shot can omit the floating
+  button's pixels (headless software compositor; the DOM state is correct
+  and light theme paints) — judge the dark floating button from &bare=1
+  and the ✓ state from the light shot.  &stepmax=N bisects a paint
+  regression to the interaction that triggers it.
 Perf harness (/perf/livepass.html): long-session performance baseline for the
  interactive pane — mounts the REAL InteractivePane at real scroll geometry
  (fixed-height mount, production CSS chain) and drives production-shaped
@@ -143,6 +198,24 @@ def extract_admin_fragment() -> str:
    return html[start:end]


+def extract_proxy_shim(prefix: str = "/node/livepass-node") -> str:
+    """Pull ``_JS_PROXY_SHIM`` out of console/server.py BY TEXT, not import.
+
+    ``scripts/`` has no ``sys.path`` guard, so ``import turnstone`` from here
+    resolves to whatever is installed in site-packages rather than this
+    checkout -- silently building the page from a DIFFERENT version of the
+    shim than the one you are trying to verify.  Read the source instead.
+    """
+    src = (ROOT / "turnstone/console/server.py").read_text(encoding="utf-8")
+    m = re.search(r'^_JS_PROXY_SHIM = """\\\n(.*?)^"""', src, re.S | re.M)
+    if not m:
+        raise SystemExit(
+            "livepass: could not find _JS_PROXY_SHIM in turnstone/console/server.py "
+            "-- the constant was renamed or reshaped; update extract_proxy_shim()."
+        )
+    return m.group(1).replace('"PREFIX_PLACEHOLDER"', json.dumps(prefix))
+
+
 def inject(template: str, marker: str, payload: str) -> str:
    begin = template.index(f"<!-- {marker}:BEGIN -->") + len(f"<!-- {marker}:BEGIN -->")
    end = template.index(f"<!-- {marker}:END -->")
@@ -347,6 +420,28 @@ CONSOLE_TEMPLATE = """<!doctype html>
    <div id="toast" role="status" aria-live="polite"></div>
    <script>
      (function () {
+        // Freeze window.fetch BEFORE the module scripts evaluate: auth.js
+        // fires a boot-time whoami at import, and a non-OK answer from the
+        // fixture server would CLEAR the permissions grant seeded below
+        // mid-pass. A never-settling fetch keeps the seed authoritative;
+        // everything the passes drive flows through the authFetch fixture
+        // (reinstated after auth.js's window bridge runs — see the load
+        // handler).
+        window.fetch = function () {
+          return new Promise(function () {});
+        };
+        // Grant the operator scopes admin.js gates on: _modelAuthEditable()
+        // reads this exact key THROUGH the real auth.js hasPermission
+        // (loaded below, before admin.js) — without the grant, or without
+        // auth.js supplying window.hasPermission, the auth-constraints
+        // stub below is dead code: _fetchModelAuthConstraints returns
+        // before authFetch and every pass renders the Backend-auth section
+        // in its read-only degraded state. The headless profile is fresh
+        // per pass, so nothing else seeds it.
+        sessionStorage.setItem(
+          "turnstone_permissions",
+          "admin.models,admin.mcp",
+        );
        function reply(data) {
          return Promise.resolve({
            ok: true,
@@ -372,9 +467,15 @@ CONSOLE_TEMPLATE = """<!doctype html>
          enabled: true, temperature: null, max_tokens: null,
          reasoning_effort: null, surface_persisted_reasoning: true,
          replay_reasoning_to_model: false,
+          auth_mode: "static", obo_audience: "", obo_scopes: "",
        };
        window.__putCount = 0;
-        window.authFetch = function (url, opts) {
+        // Held under a private name too: auth.js's legacy window bridge
+        // (Object.assign(window, {authFetch})) runs at module-import time
+        // and clobbers the plain window.authFetch assigned here — the load
+        // handler reinstates the fixture from this name after the modules
+        // have evaluated.
+        window.__consoleAuthFetch = window.authFetch = function (url, opts) {
          var method = (opts && opts.method) || "GET";
          if (method === "PUT" && url.indexOf("/model-definitions/def1") >= 0) {
            window.__putCount++;
@@ -404,8 +505,26 @@ CONSOLE_TEMPLATE = """<!doctype html>
                supports_effort: true,
              },
            });
+          if (url.indexOf("/model-definitions/auth-constraints") >= 0)
+            // Fetched by the shelf ON OPEN (showCreateModelModal /
+            // showEditModelModal), so this stub is exercised by any pass that
+            // opens the model editor — no tab-switch plumbing needed. Omitting
+            // it would render the Backend-auth block in its degraded
+            // no-suggestions state and quietly stop exercising the section.
+            return reply({
+              auth_audience_allowlist: ["api://example-gateway"],
+              auth_grant_profile: "entra",
+              dynamic_auth_modes: ["entra_app", "entra_obo", "rfc8693_obo"],
+              scopes_auth_modes: ["rfc8693_obo"],
+              app_identity_auth_modes: ["entra_app"],
+              auth_mode_profiles: {
+                entra_app: "entra", entra_obo: "entra",
+                rfc8693_obo: "rfc8693",
+              },
+            });
          if (url.indexOf("/model-definitions/def1") >= 0) return reply(MODEL);
-          if (url.indexOf("/model-definitions") >= 0) return reply({ models: [] });
+          if (url.indexOf("/model-definitions") >= 0)
+            return reply({ models: [], default_alias: "fable-5" });
          if (url.indexOf("/api/models") >= 0)
            return reply({ models: [
              { alias: "fable-5", model: "claude-fable-5" },
@@ -432,10 +551,22 @@ CONSOLE_TEMPLATE = """<!doctype html>
    </script>
    <script type="module" src="shared/utils.js"></script>
    <script type="module" src="shared/hatch.js"></script>
+    <!-- The REAL auth.js, loaded (and therefore parsed) before admin.js's
+         permission shims run any pass: it owns the sessionStorage parse
+         contract and assigns the window.hasPermission /
+         window.whenPermissionsReady globals the shims probe at call time.
+         Without it the seeded permissions grant is never READ, the
+         Backend-auth section renders read-only/hidden, and the
+         auth-constraints stub above is dead code in every pass. -->
+    <script type="module" src="shared/auth.js"></script>
    <script src="console-static/admin.js"></script>
    <script src="console-static/governance.js"></script>
    <script>
      window.addEventListener("load", function () {
+        // Reinstate the fixture fetch now the modules (and auth.js's
+        // window bridge) have evaluated — passes run after load, so every
+        // shelf-open fetch flows through the fixture, not the bridge.
+        window.authFetch = window.__consoleAuthFetch;
        var q = new URLSearchParams(location.search);
        if (q.get("theme") === "light")
          document.documentElement.dataset.theme = "light";
@@ -658,6 +789,108 @@ SHELL_TEMPLATE = """<!doctype html>
 # call the same window.buildAttachmentPreview).  The page frame is harness-only
 # chrome and not under review; the chips row and the pill row are.
 # --------------------------------------------------------------------------
+
+# The PROXIED NODE page: the real L-shell (so the rail brand is the real
+# element, with the real shell.js click listener on it) plus the real proxy
+# shim injected exactly where proxy_index puts it -- first thing inside
+# <body>, ahead of the deferred shell.js module.  caps mirror a NODE, not
+# the console: brandSub "server" is what the shim has to overwrite, and
+# leaving it "console" would make the host's /console/i check vacuous.
+PROXYBRAND_FRAME_TEMPLATE = """<!doctype html>
+<html lang="en">
+  <head>
+    <meta charset="utf-8" />
+    <title>proxied node</title>
+    <link rel="stylesheet" href="shared/base.css" />
+    <link rel="stylesheet" href="shared/ui-base.css" />
+    <link rel="stylesheet" href="static/style.css" />
+    <link rel="stylesheet" href="shared/shell.css" />
+  </head>
+  <body>
+    <!-- SHIM:BEGIN -->
+    <!-- SHIM:END -->
+    <div id="header" style="display: none"><div id="status-bar"></div></div>
+    <div id="main" style="padding: 18px">
+      <h2 style="margin: 0 0 8px">Node dashboard</h2>
+    </div>
+    <div id="view-admin" style="display: none"></div>
+    <script>
+      window.TURNSTONE_SHELL_CAPS = { cluster: false, brandSub: "server" };
+      window.TS_APP = {
+        boot() {},
+        getClusterState() { return { nodes: {} }; },
+        onRender() {},
+      };
+      window.TS_ADMIN = {};
+      // Record on the PARENT, which survives the frame's navigation.
+      // A flag on the frame's own window dies with the document, so the
+      // host would read undefined and pass -- a check that cannot fail.
+      window.showHome = function () {
+        try { window.parent.__showHomeRan = true; } catch (e) {}
+      };
+    </script>
+    <script type="module" src="shared/shell.js"></script>
+  </body>
+</html>
+"""
+
+# The HOST page.  The shim navigates the FRAME to "/", which would destroy
+# any verdict stamped inside it -- so the surviving top page reads the
+# frame's post-navigation location and stamps its own title instead.  No
+# landing page at "/" is needed (the harness root serves a directory
+# listing) and no CDP client either.
+PROXYBRAND_HOST_TEMPLATE = """<!doctype html>
+<html lang="en">
+  <head>
+    <meta charset="utf-8" />
+    <title>proxybrand livepass</title>
+    <style>
+      html, body { margin: 0; height: 100%; }
+      iframe { width: 100%; height: 100%; border: 0; }
+    </style>
+  </head>
+  <body>
+    <iframe id="frame" src="frame.html"></iframe>
+    <script>
+      const frame = document.getElementById("frame");
+      let phase = 0;
+      const fail = (r) => { phase = 9; document.title = "PROXYBRAND-FAILED-" + r; };
+
+      frame.addEventListener("load", () => {
+        if (phase === 9) return;
+        if (phase === 0) {
+          const doc = frame.contentDocument;
+          const brand = doc.querySelector(".rail-brand .brand-home");
+          if (!brand) return fail("no-brand");
+          const sub = brand.querySelector(".brand-sub");
+          if (!sub) return fail("no-sub");
+          const text = sub.textContent.trim();
+          if (text === "server") return fail("sub-not-repointed-server");
+          if (!/console/i.test(text)) return fail("sub-not-console-" + text);
+          if (brand.getAttribute("aria-label") !== "Back to console")
+            return fail("aria-not-repointed");
+          phase = 1;
+          // Click the CHILD span, as a real user does: the shim must match
+          // via contains(), not target identity.
+          sub.click();
+          setTimeout(() => { if (phase === 1) fail("no-navigation"); }, 2000);
+          return;
+        }
+        const path = frame.contentWindow.location.pathname;
+        const ranShowHome = !!window.__showHomeRan;
+        phase = 2;
+        if (path !== "/") return fail("nav-" + path);
+        if (ranShowHome) return fail("showhome-also-ran");
+        // Sticky, mirroring fail(): a third load must not re-stamp.
+        phase = 9;
+        document.title = "PROXYBRAND-READY";
+      });
+    </script>
+  </body>
+</html>
+"""
+
+
 ATTACH_TEMPLATE = """<!doctype html>
 <html lang="en">
  <head>
@@ -813,6 +1046,165 @@ ATTACH_TEMPLATE = """<!doctype html>
 """


+# --------------------------------------------------------------------------
+# Native paste-over-HTTP harness.  Unlike the copy-affordance harness, this
+# must never stub clipboard access or dispatch a script-created paste event:
+# its purpose is to prove that the production ClipboardEvent path still sees
+# user-agent clipboard data on a non-secure origin.
+# --------------------------------------------------------------------------
+PASTE_TEMPLATE = """<!doctype html>
+<html lang="en">
+  <head>
+    <meta charset="utf-8" />
+    <meta name="viewport" content="width=device-width, initial-scale=1" />
+    <title>PASTE-HTTP-BOOTING</title>
+    <link rel="stylesheet" href="shared/base.css" />
+    <link rel="stylesheet" href="shared/ui-base.css" />
+    <link rel="stylesheet" href="shared/chat.css" />
+    <style>
+      body {
+        margin: 0; padding: 32px;
+        background: var(--bg); color: var(--fg);
+        font-family: var(--font-sans, system-ui, sans-serif);
+      }
+      main { width: min(760px, 100%); margin: 0 auto; }
+      #paste-source {
+        box-sizing: border-box; width: 100%; height: 80px;
+      }
+      #composer-mount, #probe-status { margin-top: 16px; }
+      #probe-status { white-space: pre-wrap; }
+    </style>
+  </head>
+  <body>
+    <main>
+      <h1>Native paste on plain HTTP</h1>
+      <p>
+        This passes only when a trusted paste exposes clipboard text to the
+        real Composer on a non-secure HTTP origin.
+      </p>
+      <p>1. Select this 2001-character fixture, then copy it normally.</p>
+      <textarea id="paste-source" aria-label="Paste test source"></textarea>
+      <button id="select-source" type="button">Select source text</button>
+      <p>2. Focus the composer and paste normally.</p>
+      <div id="composer-mount"></div>
+      <pre id="probe-status" role="status" aria-live="polite">Booting…</pre>
+    </main>
+    <script type="module">
+      import { Composer } from "./shared/composer.js";
+      import { PASTE_ATTACHMENT_CHARS } from "./shared/composer_paste_text.js";
+
+      const prefix = "TURNSTONE-PASTE-HTTP:";
+      const seedText =
+        prefix + "x".repeat(PASTE_ATTACHMENT_CHARS + 1 - prefix.length);
+      const source = document.getElementById("paste-source");
+      const status = document.getElementById("probe-status");
+      source.value = seedText;
+
+      let copyTrusted = false;
+      let paste = null;
+      let attachment = null;
+      let fileSettled = false;
+      let failed = false;
+
+      function paint() {
+        status.textContent = document.title + "\\n" + JSON.stringify({
+          origin: location.origin,
+          secureContext: window.isSecureContext,
+          copyTrusted: copyTrusted,
+          paste: paste,
+          attachment: attachment,
+        }, null, 2);
+      }
+
+      function fail(reason) {
+        if (failed) return;
+        failed = true;
+        document.title = "PASTE-HTTP-FAILED-" + reason;
+        paint();
+      }
+
+      function finish() {
+        if (failed || !paste || !fileSettled) return;
+        const checks = [
+          ["copy-untrusted", copyTrusted],
+          ["paste-untrusted", paste.trusted],
+          ["no-clipboard-data", paste.hasClipboardData],
+          ["no-text-plain", paste.hasPlainText],
+          ["paste-not-canceled", paste.defaultPrevented],
+          ["attach-count", attachment.calls === 1],
+          ["filename", attachment.name === "pasted-text.txt"],
+          ["mime", attachment.type === "text/plain"],
+          ["size", attachment.size === new Blob([seedText]).size],
+          ["content", attachment.contentMatches],
+          ["inline-insert", paste.inputValue === ""],
+        ];
+        const firstFailure = checks.find(function (entry) { return !entry[1]; });
+        if (firstFailure) return fail(firstFailure[0]);
+        document.title = "PASTE-HTTP-READY";
+        paint();
+      }
+
+      document.addEventListener("copy", function (event) {
+        copyTrusted = event.isTrusted;
+        paint();
+      }, true);
+
+      document.addEventListener("paste", function (event) {
+        const observed = {
+          trusted: event.isTrusted,
+          hasClipboardData: !!event.clipboardData,
+          hasPlainText: !!event.clipboardData &&
+            Array.from(event.clipboardData.types || []).includes("text/plain"),
+        };
+        setTimeout(function () {
+          paste = Object.assign(observed, {
+            defaultPrevented: event.defaultPrevented,
+            inputValue: composer.inputEl.value,
+          });
+          if (!attachment) fail("no-attachment");
+          else finish();
+        }, 0);
+      });
+
+      const composer = new Composer(document.getElementById("composer-mount"), {
+        onSend: function () {},
+        placeholder: "Paste the copied fixture here…",
+        attachments: {
+          onAttach: function (file) {
+            attachment = {
+              calls: attachment ? attachment.calls + 1 : 1,
+              name: file.name,
+              type: file.type,
+              size: file.size,
+              contentMatches: false,
+            };
+            file.text().then(function (text) {
+              attachment.contentMatches = text === seedText;
+              fileSettled = true;
+              finish();
+            }, function () { fail("file-read"); });
+            return true;
+          },
+        },
+      });
+
+      document.getElementById("select-source").addEventListener("click", function () {
+        source.focus();
+        source.select();
+      });
+
+      if (location.protocol !== "http:") fail("not-http");
+      else if (window.isSecureContext) fail("secure-context");
+      else {
+        document.title = "PASTE-HTTP-WAITING";
+        paint();
+      }
+    </script>
+  </body>
+</html>
+"""
+
+
 # --------------------------------------------------------------------------
 # Task-agent harness — the task_agent card: a task agent's sub-tool steps
 # nested under its conversation row.  Driven through the REAL
@@ -821,7 +1213,24 @@ ATTACH_TEMPLATE = """<!doctype html>
 # is exercised, not just the leaf builders.  The page frame is harness-only
 # chrome; the .conv-batch / task_agent card is what's under review.
 # --------------------------------------------------------------------------
-TASKAGENT_TEMPLATE = """<!doctype html>
+# The host seams a mounted InteractivePane provides, stubbed once for every
+# harness that drives the REAL pane (taskagent, copy).  A new required seam
+# gets added HERE — a harness left with a stale stub set does not fail at
+# review time, it throws HARNESS ERROR at run time.
+PANE_STUB_JS = """\
+        // Drive the REAL pane; stub only the host seams a mounted pane provides.
+        const pane = new InteractivePane("demo-ws");
+        pane.messagesEl = messages;
+        pane.inputEl = document.createElement("textarea");
+        pane.sendBtn = document.createElement("button");
+        pane.isNearBottom = () => false;
+        pane.scrollToBottom = () => {};
+        pane.removeEmptyState = () => {};
+        pane.removeThinkingIndicator = () => {};
+        pane.setBusy = () => {};"""
+
+TASKAGENT_TEMPLATE = (
+    """<!doctype html>
 <html lang="en">
  <head>
    <meta charset="utf-8" />
@@ -871,16 +1280,9 @@ TASKAGENT_TEMPLATE = """<!doctype html>

      const messages = document.getElementById("messages");
      try {
-        // Drive the REAL pane; stub only the host seams a mounted pane provides.
-        const pane = new InteractivePane("demo-ws");
-        pane.messagesEl = messages;
-        pane.inputEl = document.createElement("textarea");
-        pane.sendBtn = document.createElement("button");
-        pane.isNearBottom = () => false;
-        pane.scrollToBottom = () => {};
-        pane.removeEmptyState = () => {};
-        pane.removeThinkingIndicator = () => {};
-        pane.setBusy = () => {};
+"""
+    + PANE_STUB_JS
+    + """
        const ev = (e) => pane.handleEvent(e);

        // ?recall=1: exercise the RECALL path — replayHistory rebuilding the
@@ -1029,6 +1431,266 @@ TASKAGENT_TEMPLATE = """<!doctype html>
  </body>
 </html>
 """
+)
+
+
+# --------------------------------------------------------------------------
+# Copy harness — the copy-to-clipboard affordances over the REAL pane.  The
+# bubbles come from the REAL replayHistory / handleEvent paths so the copy
+# sources are the ones production stashes (_copySource, the mermaid / table
+# data attributes), and the probes drive the REAL buttons and key path and
+# compare what landed on the (stubbed) clipboard byte-exact against the
+# source.
+# --------------------------------------------------------------------------
+COPY_TEMPLATE = (
+    """<!doctype html>
+<html lang="en">
+  <head>
+    <meta charset="utf-8" />
+    <meta name="viewport" content="width=device-width, initial-scale=1" />
+    <title>copy livepass</title>
+    <link rel="stylesheet" href="shared/base.css" />
+    <link rel="stylesheet" href="shared/ui-base.css" />
+    <link rel="stylesheet" href="shared/chat.css" />
+    <link rel="stylesheet" href="shared/conversation.css" />
+    <link rel="stylesheet" href="shared/cards.css" />
+    <link rel="stylesheet" href="shared/interactive.css" />
+    <style>
+      /* Harness-only framing (NOT under review) — a plausible pane context. */
+      body {
+        padding: 24px; margin: 0; background: var(--bg); color: var(--ink);
+        font-family: var(--font-sans, system-ui, sans-serif);
+      }
+      .demo-frame { max-width: 720px; margin: 0 auto; }
+      .demo-label {
+        font: 11px var(--font-mono, monospace); color: var(--ink-3);
+        text-transform: uppercase; letter-spacing: 0.08em; margin: 0 0 8px;
+      }
+    </style>
+  </head>
+  <body>
+    <div class="demo-frame">
+      <div class="demo-label">conversation — copy affordances (real InteractivePane)</div>
+      <div class="messages" id="messages"></div>
+    </div>
+    <script>
+      window.toast = { error: function (m) { console.log("toast:", m); } };
+      window.authFetch = function () {
+        return Promise.resolve({
+          ok: true,
+          json: function () { return Promise.resolve({}); },
+          text: function () { return Promise.resolve(""); },
+        });
+      };
+      // Deterministic clipboard: record instead of writing.  localhost is a
+      // secure context so copyTextToClipboard takes the async-API branch and
+      // hits this stub; force isSecureContext for any odd serving setup.
+      window.__copied = [];
+      try {
+        Object.defineProperty(window, "isSecureContext", { value: true });
+      } catch (e) { /* already true */ }
+      try {
+        Object.defineProperty(navigator, "clipboard", {
+          value: {
+            writeText: function (t) {
+              window.__copied.push(t);
+              return Promise.resolve();
+            },
+          },
+          configurable: true,
+        });
+      } catch (e) {
+        document.title = "COPY-FAILED-clipboard-stub";
+      }
+    </script>
+    <script type="module">
+      import { InteractivePane } from "./shared/interactive.js";
+      const q = new URLSearchParams(location.search);
+      if (q.get("theme") === "light")
+        document.documentElement.dataset.theme = "light";
+
+      const FENCE_SRC = 'def stash(depth):\\n    total = 0\\n    for k in range(depth):\\n        total += k\\n    return total';
+      const TABLE_SRC = '| node | state |\\n|---|:--:|\\n| flat | idle |\\n| blck | busy |';
+      const MERMAID_SRC = 'graph TD\\n  A --> B\\n  B --> C';
+      const MD_ONE =
+        'First reply with a fence and a table.\\n\\n' +
+        '```python\\n' + FENCE_SRC + '\\n```\\n\\n' +
+        TABLE_SRC + '\\n\\nTrailing prose under the table.';
+      const MD_TWO =
+        'Second reply with a diagram.\\n\\n' +
+        '```mermaid\\n' + MERMAID_SRC + '\\n```\\n\\n' +
+        'And `inline code` after it.';
+      const MD_LIVE =
+        'Streamed reply: the **live** turn, so the retry holder lands here.';
+
+      const messages = document.getElementById("messages");
+      const fail = (r) => { document.title = "COPY-FAILED-" + r; };
+      try {
+"""
+    + PANE_STUB_JS
+    + """
+
+        pane.replayHistory([
+          { role: "user", content: "Show me the stash helper and the node table." },
+          { role: "assistant", content: MD_ONE },
+          { role: "user", content: "Now the flow as a diagram, please." },
+          { role: "assistant", content: MD_TWO },
+        ]);
+        // A live streamed turn on top — the retry holder must land on this
+        // bubble WITHOUT stripping its (or any) copy button.
+        pane.handleEvent({ type: "state_change", state: "running" });
+        for (let k = 0; k < MD_LIVE.length; k += 16)
+          pane.handleEvent({ type: "content", text: MD_LIVE.slice(k, k + 16) });
+        pane.handleEvent({ type: "stream_end" });
+        pane.handleEvent({ type: "state_change", state: "idle" });
+
+        const hover = (el) =>
+          el.dispatchEvent(new MouseEvent("mouseover", { bubbles: true }));
+        const fabEl = () => document.querySelector(".block-copy-btn");
+
+        // Let the streamed bubble's rAF render + retry attach settle.
+        setTimeout(async () => {
+          try {
+          const bubbles = messages.querySelectorAll(".msg.assistant");
+          const bars = messages.querySelectorAll(
+            ".msg.assistant .msg-actions .msg-copy-btn",
+          );
+          if (bubbles.length !== 3) return fail("bubbles" + bubbles.length);
+          if (bars.length !== 3) return fail("bars" + bars.length);
+          const last = bubbles[bubbles.length - 1];
+          if (!last.querySelector(".msg-retry-btn"))
+            return fail("no-retry-on-holder");
+          if (!last.querySelector(".msg-copy-btn"))
+            return fail("holder-lost-copy");
+
+          // Block probes: hover reveals the floating button; a click must
+          // land the byte-exact SOURCE on the clipboard.
+          const probes = [
+            [messages.querySelector(".msg.assistant pre"), FENCE_SRC, "fence"],
+            [messages.querySelector(".table-wrap"), TABLE_SRC, "table"],
+            [messages.querySelector(".mermaid-container"), MERMAID_SRC, "mermaid"],
+          ];
+          // &bare=1 — diagnostic state: no probe clicks, no repositioning;
+          // one hover on the fence and stop.  Splits "the probe cycle
+          // corrupts the button's paint" from "it never paints here".
+          if (q.get("bare") === "1") {
+            hover(probes[0][0]);
+            document.title = "COPY-BARE";
+            return;
+          }
+
+          // &kbd=1 — the keyboard path: Enter on a FOCUSED block copies
+          // that block's source directly.  Blocks are focusable (tabindex=0
+          // from the fence / table / mermaid renders), the outcome flashes
+          // on the block itself, and the floating button — pointer-only —
+          // must stay out of it entirely (never created, never revealed).
+          if (q.get("kbd") === "1") {
+            const tw = messages.querySelector(".table-wrap");
+            if (!tw) return fail("kbd-no-block");
+            tw.focus();
+            const focused = document.activeElement === tw;
+            tw.dispatchEvent(
+              new KeyboardEvent("keydown", { key: "Enter", bubbles: true }),
+            );
+            await new Promise((r) => setTimeout(r, 0));
+            const copied =
+              window.__copied[window.__copied.length - 1] === TABLE_SRC;
+            const flashed = tw.classList.contains("is-copied");
+            const fabStaysOut =
+              !fabEl() || !fabEl().classList.contains("is-visible");
+            document.title =
+              focused && copied && flashed && fabStaysOut
+                ? "COPY-KBD-READY"
+                : "COPY-KBD-FAILED-" +
+                  [
+                    focused ? "" : "focus",
+                    copied ? "" : "copy",
+                    flashed ? "" : "flash",
+                    fabStaysOut ? "" : "fab",
+                  ]
+                    .filter(Boolean)
+                    .join("-");
+            return;
+          }
+
+          // &flash=1 — the VISUAL state, screenshot-only: skip the probes so
+          // the fence hover is the floating button's FIRST show.  Returning
+          // the button to an already-visited position stops it PAINTING in
+          // headless captures (visible + hit-testable, no pixels — a stale
+          // compositor tile; bisected via &stepmax).  Function and pixels
+          // are therefore split: the probe run (no flash) is the verdict,
+          // this state is the picture.
+          if (q.get("flash") === "1") {
+            bars[bars.length - 1].focus();
+            hover(probes[0][0]);
+            const fab = fabEl();
+            if (!fab) return fail("no-fab-visual");
+            fab.classList.add("is-copied");
+            fab.title = "Copied";
+            // Freeze: the capture pipeline synthesizes a pointer event
+            // outside the block at screenshot time, which would hide the
+            // button (correct in production).  Capture-phase stops starve
+            // the module's delegated listeners for the capture.
+            for (const t of ["mouseover", "scroll"])
+              document.addEventListener(t, (e) => e.stopPropagation(), true);
+            document.title = "COPY-VISUAL";
+            return;
+          }
+          // &stepmax=N — diagnostic: stop after the Nth interaction (hovers
+          // and clicks count) and stamp COPY-STEP-N, so a paint regression
+          // can be bisected to the interaction that triggers it.
+          let step = 0;
+          const stepMax = parseInt(q.get("stepmax") || "999", 10);
+          const gate = () => {
+            step += 1;
+            if (step > stepMax) {
+              document.title = "COPY-STEP-" + (step - 1);
+              throw { __stop: true };
+            }
+          };
+          let done = 0;
+          for (const [el, want, name] of probes) {
+            if (!el) return fail("no-" + name);
+            gate();
+            hover(el);
+            const fab = fabEl();
+            if (!fab || !fab.classList.contains("is-visible"))
+              return fail("fab-hidden-" + name);
+            gate();
+            fab.click();
+            await new Promise((r) => setTimeout(r, 0));
+            const got = window.__copied[window.__copied.length - 1];
+            if (got !== want) {
+              console.log("copy mismatch", name, JSON.stringify(got));
+              return fail("source-" + name);
+            }
+            done += 1;
+          }
+
+          // Bubble probe: the whole raw markdown, fences and pipes intact.
+          gate();
+          bars[0].click();
+          await new Promise((r) => setTimeout(r, 0));
+          if (window.__copied[window.__copied.length - 1] !== MD_ONE)
+            return fail("bubble-source");
+
+          document.title = "COPY-READY-" + bars.length + "-" + done;
+          } catch (e) {
+            if (!(e && e.__stop)) {
+              console.log("copy harness error", e);
+              fail("error");
+            }
+          }
+        }, 400);
+      } catch (e) {
+        messages.textContent = "HARNESS ERROR: " + e.message;
+        fail("error");
+      }
+    </script>
+  </body>
+</html>
+"""
+)


 # --------------------------------------------------------------------------
@@ -1470,12 +2132,24 @@ def build(out: Path) -> None:
    (att / "livepass.html").write_text(ATTACH_TEMPLATE, encoding="utf-8")
    print(f"{att}/livepass.html — composer chips + message attachment pills")

+    paste = out / "paste"
+    paste.mkdir(parents=True, exist_ok=True)
+    symlink(paste / "shared", ROOT / "turnstone/shared_static")
+    (paste / "livepass.html").write_text(PASTE_TEMPLATE, encoding="utf-8")
+    print(f"{paste}/livepass.html — trusted native paste on insecure HTTP")
+
    ta = out / "taskagent"
    ta.mkdir(parents=True, exist_ok=True)
    symlink(ta / "shared", ROOT / "turnstone/shared_static")
    (ta / "livepass.html").write_text(TASKAGENT_TEMPLATE, encoding="utf-8")
    print(f"{ta}/livepass.html — task_agent card (real Pane.handleEvent routing)")

+    cp = out / "copy"
+    cp.mkdir(parents=True, exist_ok=True)
+    symlink(cp / "shared", ROOT / "turnstone/shared_static")
+    (cp / "livepass.html").write_text(COPY_TEMPLATE, encoding="utf-8")
+    print(f"{cp}/livepass.html — copy affordances (bubble bars + block button)")
+
    pf = out / "perf"
    pf.mkdir(parents=True, exist_ok=True)
    symlink(pf / "shared", ROOT / "turnstone/shared_static")
@@ -1483,6 +2157,17 @@ def build(out: Path) -> None:
    (pf / "livepass.html").write_text(PERF_TEMPLATE, encoding="utf-8")
    print(f"{pf}/livepass.html — long-session perf baseline (real InteractivePane)")

+    pb = out / "proxybrand"
+    pb.mkdir(parents=True, exist_ok=True)
+    symlink(pb / "shared", ROOT / "turnstone/shared_static")
+    symlink(pb / "static", ROOT / "turnstone/ui/static")
+    shim = "<script>" + extract_proxy_shim() + "</script>"
+    (pb / "frame.html").write_text(
+        inject(PROXYBRAND_FRAME_TEMPLATE, "SHIM", shim), encoding="utf-8"
+    )
+    (pb / "livepass.html").write_text(PROXYBRAND_HOST_TEMPLATE, encoding="utf-8")
+    print(f"{pb}/livepass.html — back-to-console brand (real shell.js + real shim)")
+

 class _PerfStore:
    """Rendezvous for the perf page's POSTed JSON report."""
@@ -1713,6 +2398,12 @@ def main() -> None:
    ap = argparse.ArgumentParser(description=__doc__.splitlines()[0])
    ap.add_argument("--out", type=Path, default=Path("/tmp/livepass"))
    ap.add_argument("--serve", type=int, metavar="PORT")
+    ap.add_argument(
+        "--bind",
+        default="127.0.0.1",
+        metavar="HOST",
+        help="listen address for --serve (use 0.0.0.0 for a manual insecure-origin pass)",
+    )
    ap.add_argument("--perf", action="store_true", help="run the perf baseline and exit")
    ap.add_argument(
        "--perf-n",
@@ -1730,8 +2421,9 @@ def main() -> None:
        import functools

        handler = functools.partial(_HarnessHandler, directory=str(args.out))
-        print(f"serving {args.out} on http://localhost:{args.serve}/ — Ctrl+C stops")
-        http.server.ThreadingHTTPServer(("127.0.0.1", args.serve), handler).serve_forever()
+        display_host = "localhost" if args.bind == "127.0.0.1" else args.bind
+        print(f"serving {args.out} on http://{display_host}:{args.serve}/ — Ctrl+C stops")
+        http.server.ThreadingHTTPServer((args.bind, args.serve), handler).serve_forever()


 if __name__ == "__main__":
@@ -17,6 +17,19 @@ Checks E1–E7 mirror the Entra harness:
  E6 unconsented audience C → NOT token, credential SURVIVES
  E7 cache flush → re-mint

+M1-M3 drive the MODEL-backend mint (``mint_obo_access_token``, #898/#955) on
+the same captured credential — the path an ``auth_mode=rfc8693_obo`` model
+alias takes, distinct from the classified MCP path above:
+  M1 model mint audience A with the alias's exchange scopes → token carries A
+     (the #955 fix: model definitions now carry per-row ``obo_scopes``, so
+     the exchange leg requests the audience's scope exactly as MCP rows do)
+  M2 warm re-mint serves the synthetic ``__model_obo__`` cache row —
+     identity-keyed on the owning alias, audience + scopes in the row's
+     own columns — with zero IdP calls
+  M3 an entra-leg mode (``entra_obo``) on this rfc8693 deployment refuses
+     BEFORE any IdP traffic, recording cause=grant_profile_mismatch — the
+     mode/profile pairing that replaced the pre-#955 overload
+
 Env (set by keycloak_e2e.sh):
  KC_TOKEN_ENDPOINT, KC_ISSUER, KC_CLIENT_ID, KC_CLIENT_SECRET,
  KC_USER, KC_PASSWORD, AUD_A, SCOPE_A, AUD_B, SCOPE_B, AUD_C
@@ -40,7 +53,13 @@ from turnstone.core.mcp_crypto import (
    MCPTokenCipherConfig,
    MCPTokenStore,
 )
-from turnstone.core.mcp_oauth import get_obo_access_token_classified
+from turnstone.core.mcp_oauth import (
+    get_obo_access_token_classified,
+    mint_obo_access_token,
+    model_mint_refusal_cause,
+    model_obo_cache_server,
+    model_obo_cause_key,
+)
 from turnstone.core.oidc import OIDCConfig
 from turnstone.core.storage._sqlite import SQLiteBackend

@@ -235,6 +254,87 @@ async def _run(cfg: dict[str, str], refresh_token: str) -> None:
            "VERIFIED" if r7.kind == "token" and client.posts > posts_before else "FAILED",
            f"E7 flush→re-mint: kind={r7.kind} kc_calls={client.posts - posts_before} (want >=1)",
        )
+
+        # M1-M3 — MODEL backend mint on the rfc8693 profile: same captured
+        # credential and legs as E1-E7, but through mint_obo_access_token —
+        # the path an auth_mode=rfc8693_obo alias takes, carrying the
+        # per-alias exchange scopes MCP rows always had (#955). The mint's
+        # cache and cause records are identity-keyed on the owning alias, so
+        # the harness names one per mode-variant exactly as a deployment
+        # would define separate rows.
+        posts_before = client.posts
+        m1 = await mint_obo_access_token(
+            app_state=app_state,
+            user_id=USER,
+            alias="model-a",
+            audience=cfg["AUD_A"],
+            scopes=cfg.get("SCOPE_A", ""),
+            grant_leg="rfc8693",
+        )
+        m1_kc_calls = client.posts - posts_before
+        if m1:
+            ok1, why1 = aud_carries(m1, cfg["AUD_A"])
+            record(
+                "VERIFIED" if ok1 and m1_kc_calls > 0 else "FAILED",
+                f"M1 model mint (rfc8693_obo, scoped exchange): token={redact(m1)} "
+                f"aud_ok={ok1} ({why1}) kc_calls={m1_kc_calls} (want >=1)",
+            )
+        else:
+            record(
+                "FAILED",
+                f"M1 model mint (rfc8693_obo): no token (kc_calls={m1_kc_calls}) — "
+                "the #955 scope wire-through should mint here",
+            )
+
+        # M2 — warm re-mint serves the synthetic __model_obo__ cache row —
+        # identity-keyed on the owning alias, audience + scopes in the row's
+        # own columns — with zero IdP calls, and the row is named so
+        # deprovisioning can find it by prefix.
+        posts_before = client.posts
+        m2 = await mint_obo_access_token(
+            app_state=app_state,
+            user_id=USER,
+            alias="model-a",
+            audience=cfg["AUD_A"],
+            scopes=cfg.get("SCOPE_A", ""),
+            grant_leg="rfc8693",
+        )
+        cache_row = storage.get_mcp_user_token(USER, model_obo_cache_server("model-a"))
+        if m1:
+            record(
+                "VERIFIED"
+                if m2 and client.posts == posts_before and cache_row is not None
+                else "FAILED",
+                f"M2 model cache-hit: token={redact(m2)} kc_calls="
+                f"{client.posts - posts_before} (want 0) synthetic_row="
+                f"{'present' if cache_row is not None else 'MISSING'}",
+            )
+        else:
+            record("FAILED", "M2 model cache-hit: blocked behind M1 — M1 failed, see above")
+
+        # M3 — the mode/profile pairing refusal that replaced the pre-#955
+        # overload: an entra-leg mode on this rfc8693 deployment must yield
+        # None with ZERO IdP calls and record the grant_profile_mismatch
+        # cause the session heartbeat reads (under its own alias — a
+        # deployment defines the entra-mode variant as its own row).
+        posts_before = client.posts
+        m3 = await mint_obo_access_token(
+            app_state=app_state,
+            user_id=USER,
+            alias="model-a-entra",
+            audience=cfg["AUD_A"],
+            grant_leg="entra",
+        )
+        m3_cause = model_mint_refusal_cause(
+            "model_obo", model_obo_cause_key("model-a-entra", grant_leg="entra"), USER
+        )
+        record(
+            "VERIFIED"
+            if m3 is None and client.posts == posts_before and m3_cause == "grant_profile_mismatch"
+            else "FAILED",
+            f"M3 mode/profile mismatch refusal: token={redact(m3)} (want absent) "
+            f"kc_calls={client.posts - posts_before} (want 0) cause={m3_cause!r}",
+        )
    finally:
        await inner.aclose()

@@ -185,8 +185,8 @@ esac

 echo ""
 echo "NOTE: If you added a NEW library (not just updating a version), also update"
-echo "  the _ASSET_RE regex in turnstone/core/web_helpers.py — its negative lookahead"
-echo "  skips vendored directories to avoid double-versioning static asset URLs."
+echo "  _VERSIONED_VENDOR_DIR in turnstone/core/web_helpers.py — it controls both"
+echo "  HTML version rewriting and immutable static-response caching."
 echo ""
 echo "Verify the update:"
 echo "  git diff --stat"
@@ -2,7 +2,7 @@
  "openapi": "3.1.0",
  "info": {
    "title": "turnstone Server API",
-    "version": "1.7.0rc1",
+    "version": "1.8.0a7",
    "description": "Single-node workstream management, chat interaction, and real-time streaming."
  },
  "paths": {
@@ -55,7 +55,7 @@
        "tags": [
          "Workstreams"
        ],
-        "description": "Accepts two content types. Default is `application/json` with a `CreateWorkstreamRequest` body. Alternatively, `multipart/form-data` with one `meta` field (JSON-encoded `CreateWorkstreamRequest` shape) plus zero-or-more `file` parts saves each file as an attachment under the new workstream. When `initial_message` is also set, attachments are resolved onto that turn before the worker thread dispatches; otherwise they remain pending for a follow-up `POST /v1/api/workstreams/{ws_id}/send`.",
+        "description": "Accepts two content types. Default is `application/json` with a `CreateWorkstreamRequest` body. Alternatively, `multipart/form-data` with one `meta` field (JSON-encoded `CreateWorkstreamRequest` shape) plus zero-or-more `file` parts saves each file as an attachment under the new workstream. When `initial_message` is also set, attachments are resolved onto that turn before the worker thread dispatches; otherwise they remain pending for a follow-up `POST /v1/api/workstreams/{ws_id}/send`. Setting `resume_ws` atomically forks the visible source history, configuration, project, persona, and attachment references into a distinct destination; it does not reopen or mutate the source. Attachments and `resume_ws` cannot be combined. Creation stays unpublished until validation and the optional fork transaction complete.",
        "requestBody": {
          "required": true,
          "content": {
@@ -87,6 +87,26 @@
              }
            }
          },
+          "403": {
+            "description": "Error 403",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
+          },
+          "404": {
+            "description": "Error 404",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
+          },
          "409": {
            "description": "Error 409",
            "content": {
@@ -106,6 +126,36 @@
                }
              }
            }
+          },
+          "429": {
+            "description": "Error 429",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
+          },
+          "500": {
+            "description": "Error 500",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
+          },
+          "503": {
+            "description": "Error 503",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
          }
        }
      }
@@ -117,6 +167,7 @@
        "tags": [
          "Workstreams"
        ],
+        "description": "Unloads the live workstream while preserving storage. Returns 409 when any accepted live conversation row still requires persistence reconciliation; the workstream remains loaded and its history journal is retained.",
        "parameters": [
          {
            "name": "ws_id",
@@ -167,6 +218,16 @@
                }
              }
            }
+          },
+          "409": {
+            "description": "Error 409",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
          }
        }
      }
@@ -228,6 +289,16 @@
                }
              }
            }
+          },
+          "409": {
+            "description": "Error 409",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
          }
        }
      },
@@ -325,7 +396,7 @@
            "content": {
              "application/json": {
                "schema": {
-                  "$ref": "#/components/schemas/StatusResponse"
+                  "$ref": "#/components/schemas/ApproveResponse"
                }
              }
            }
@@ -339,6 +410,16 @@
                }
              }
            }
+          },
+          "409": {
+            "description": "Error 409",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
          }
        }
      }
@@ -390,6 +471,26 @@
                }
              }
            }
+          },
+          "409": {
+            "description": "Error 409",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
+          },
+          "503": {
+            "description": "Error 503",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
          }
        }
      }
@@ -412,7 +513,7 @@
          }
        ],
        "requestBody": {
-          "required": true,
+          "required": false,
          "content": {
            "application/json": {
              "schema": {
@@ -427,7 +528,7 @@
            "content": {
              "application/json": {
                "schema": {
-                  "$ref": "#/components/schemas/StatusResponse"
+                  "$ref": "#/components/schemas/CancelResponse"
                }
              }
            }
@@ -462,6 +563,7 @@
        "tags": [
          "Chat"
        ],
+        "description": "Claims the workstream mutation slot, durably truncates the requested tail, then emits clear_ui. Concurrent sends are ordered after the cut; a storage failure returns 503 without changing live history.",
        "parameters": [
          {
            "name": "ws_id",
@@ -512,6 +614,16 @@
                }
              }
            }
+          },
+          "503": {
+            "description": "Error 503",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
          }
        }
      }
@@ -523,6 +635,7 @@
        "tags": [
          "Chat"
        ],
+        "description": "Uses one workstream worker claim for the durable truncation and the replacement generation, so another send cannot enter between them. A storage failure returns 503 without changing live history.",
        "parameters": [
          {
            "name": "ws_id",
@@ -563,6 +676,16 @@
                }
              }
            }
+          },
+          "503": {
+            "description": "Error 503",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
          }
        }
      }
@@ -574,7 +697,7 @@
        "tags": [
          "Streaming"
        ],
-        "description": "Opens a Server-Sent Events stream scoped to a single workstream. Returns text/event-stream. See API reference for event types.",
+        "description": "Opens a Server-Sent Events stream scoped to a single workstream. After rendering REST history, pass its opaque handoff_token once as ?history_token=; it names the exact accepted conversation-row prefix used for that render. A history_resync event closes this stream and requires a fresh history read; numeric event replay is not a substitute. Native Last-Event-ID reconnects take priority. Pass ?user_turn=1 to opt into typed accepted-user events; otherwise those rows become a backward-compatible strong-repair frame. Pass ?tool_turn=1 to receive the final accepted tool row as a typed tool_result with accepted=true; without it, accepted tool rows use the same pre-row strong-repair projection. Returns text/event-stream. See API reference for event types.",
        "parameters": [
          {
            "name": "ws_id",
@@ -583,6 +706,42 @@
            "schema": {
              "type": "string"
            }
+          },
+          {
+            "name": "last_event_id",
+            "in": "query",
+            "required": false,
+            "schema": {
+              "type": "integer"
+            },
+            "description": "Numeric per-workstream event cursor for manual reconnects."
+          },
+          {
+            "name": "history_token",
+            "in": "query",
+            "required": false,
+            "schema": {
+              "type": "string"
+            },
+            "description": "Opaque one-shot token naming the accepted prefix rendered from REST history."
+          },
+          {
+            "name": "user_turn",
+            "in": "query",
+            "required": false,
+            "schema": {
+              "type": "integer"
+            },
+            "description": "Set to 1 to receive typed user_turn events instead of history-repair frames."
+          },
+          {
+            "name": "tool_turn",
+            "in": "query",
+            "required": false,
+            "schema": {
+              "type": "integer"
+            },
+            "description": "Set to 1 to receive final accepted tool_result projections."
          }
        ],
        "responses": {
@@ -609,7 +768,7 @@
        "tags": [
          "Streaming"
        ],
-        "description": "Server-Sent Events stream for node-level state broadcasts. Emits a node_snapshot event on connect (workstreams, health, aggregate), followed by real-time delta events (ws_state, ws_activity, ws_created, ws_closed, ws_rename, health_changed, aggregate). Pass ?expected_node_id=X for identity verification (returns 409 on mismatch).",
+        "description": "Server-Sent Events stream for node-level state broadcasts. Emits a node_snapshot event on connect (workstreams, health, aggregate), followed by real-time delta events (ws_state, ws_activity, ws_created, ws_closed, ws_rename, health_changed, aggregate). Pass ?expected_node_id=X for identity verification (returns 409 on mismatch). Every event's SSE id is an opaque '{boot_epoch}-{counter}' string; presenting it on reconnect (Last-Event-ID header or ?last_event_id=) replays missed events, or emits a replay_truncated event (reason: ring_evicted with lost_count + earliest_available_id, or boot_epoch when the cursor predates this server process) followed by a fresh node_snapshot. Treat the id as opaque \u2014 its format may change.",
        "responses": {
          "200": {
            "description": "Success"
@@ -882,7 +1041,7 @@
        "tags": [
          "Workstreams"
        ],
-        "description": "Returns the tail of the conversation in OpenAI-like message format. Persisted-but-not-loaded workstreams (closed / evicted) serve history without rehydrating. Lifted from the coord-only surface in the Stage 2 history/detail verb lift \u2014 interactive previously only exposed history through the SSE replay on ``/events``.",
+        "description": "Returns the tail of the conversation in OpenAI-like message format. Persisted-but-not-loaded workstreams (closed / evicted) are rehydrated before history is served so every successful response participates in the REST-to-SSE handoff. Lifted from the coord-only surface in the Stage 2 history/detail verb lift \u2014 interactive previously only exposed history through the SSE replay on ``/events``. Messages are the requested limit-bounded tail of one authoritative total accepted conversation-row prefix: user, assistant, tool, and system rows, including projected compaction checkpoints and cancellation markers. The opaque handoff_token names the exact prefix used for the render and is passed once on initial SSE registration. Admission of a later row changes the token; durable acknowledgement does not. If the durable prefix cannot be loaded, the endpoint returns 503 with `History temporarily unavailable`; that response is not authoritative and supplies no usable handoff token.",
        "parameters": [
          {
            "name": "ws_id",
@@ -1710,7 +1869,7 @@
    },
    "/v1/api/memories": {
      "get": {
-        "summary": "List structured memories",
+        "summary": "List structured memories. Without a scope, returns global plus the authenticated user's memories; workstream scope is owner-bound.",
        "operationId": "v1_api_memories_get",
        "tags": [
          "Memories"
@@ -1732,7 +1891,7 @@
            "schema": {
              "type": "string"
            },
-            "description": "Filter by scope"
+            "description": "Filter by public scope: global, workstream, or user"
          },
          {
            "name": "scope_id",
@@ -1764,6 +1923,46 @@
                }
              }
            }
+          },
+          "400": {
+            "description": "Error 400",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
+          },
+          "403": {
+            "description": "Error 403",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
+          },
+          "404": {
+            "description": "Error 404",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
+          },
+          "500": {
+            "description": "Error 500",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
          }
        }
      },
@@ -1803,13 +2002,43 @@
                }
              }
            }
+          },
+          "403": {
+            "description": "Error 403",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
+          },
+          "404": {
+            "description": "Error 404",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
+          },
+          "500": {
+            "description": "Error 500",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
          }
        }
      }
    },
    "/v1/api/memories/search": {
      "post": {
-        "summary": "Search structured memories by query",
+        "summary": "Search structured memories by query. Without a scope, searches global plus the authenticated user's memories.",
        "operationId": "v1_api_memories_search_post",
        "tags": [
          "Memories"
@@ -1834,6 +2063,46 @@
                }
              }
            }
+          },
+          "400": {
+            "description": "Error 400",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
+          },
+          "403": {
+            "description": "Error 403",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
+          },
+          "404": {
+            "description": "Error 404",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
+          },
+          "500": {
+            "description": "Error 500",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
          }
        }
      }
@@ -1884,6 +2153,26 @@
              }
            }
          },
+          "400": {
+            "description": "Error 400",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
+          },
+          "403": {
+            "description": "Error 403",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
+          },
          "404": {
            "description": "Error 404",
            "content": {
@@ -1893,6 +2182,16 @@
                }
              }
            }
+          },
+          "500": {
+            "description": "Error 500",
+            "content": {
+              "application/json": {
+                "schema": {
+                  "$ref": "#/components/schemas/ErrorResponse"
+                }
+              }
+            }
          }
        }
      }
@@ -2234,6 +2533,22 @@
            "title": "Message",
            "type": "string"
          },
+          "client_send_id": {
+            "anyOf": [
+              {
+                "maxLength": 128,
+                "minLength": 1,
+                "pattern": "^[A-Za-z0-9_-]+$",
+                "type": "string"
+              },
+              {
+                "type": "null"
+              }
+            ],
+            "default": null,
+            "description": "Opaque browser correlation token echoed in the accepted `user_turn` event and history row. It is not an idempotency key; repeated sends with the same value remain distinct turns.",
+            "title": "Client Send Id"
+          },
          "attachment_ids": {
            "anyOf": [
              {
@@ -2260,16 +2575,23 @@
      "SendResponse": {
        "properties": {
          "status": {
-            "description": "'ok', 'busy', 'queued', or 'queue_full'",
+            "description": "'ok' (fresh turn dispatched), 'queued' (folded into the live turn's interjection queue, or \u2014 when `deferred` is true \u2014 parked for dispatch after the current command window), 'queue_full', 'attachments_busy' (attachments can't ride a queued turn; retry when idle), or 'cross_user_interjection' (another participant's turn is in flight; carried on the 409 body).",
            "examples": [
              "ok",
-              "busy",
              "queued",
-              "queue_full"
+              "queue_full",
+              "attachments_busy",
+              "cross_user_interjection"
            ],
            "title": "Status",
            "type": "string"
          },
+          "deferred": {
+            "default": false,
+            "description": "Set on `queued` responses: the message is parked on the workstream's deferred-send list (a slash-command window holds the worker slot, or earlier deferred sends are still pending) and dispatches as an ordinary full-fidelity send afterwards \u2014 it is NOT in a live turn's interjection queue. `DELETE .../send` retracts it until dispatch. Node-local and in-memory: a node restart before dispatch drops it (at-most-once intake).",
+            "title": "Deferred",
+            "type": "boolean"
+          },
          "attached_ids": {
            "description": "Attachment ids actually attached to this turn. Subset of the request's `attachment_ids` (or the auto-consumed pending set). Empty when the send carries no attachments.",
            "items": {
@@ -2359,6 +2681,32 @@
            "description": "Auto-approve the tools in this batch going forward",
            "title": "Always",
            "type": "boolean"
+          },
+          "cycle_id": {
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "type": "null"
+              }
+            ],
+            "default": null,
+            "description": "Resolve this exact approval cycle",
+            "title": "Cycle Id"
+          },
+          "call_id": {
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "type": "null"
+              }
+            ],
+            "default": null,
+            "description": "Resolve the approval cycle containing this tool call",
+            "title": "Call Id"
          }
        },
        "required": [
@@ -2367,10 +2715,35 @@
        "title": "ApproveRequest",
        "type": "object"
      },
+      "ApproveResponse": {
+        "properties": {
+          "status": {
+            "default": "ok",
+            "description": "Request outcome",
+            "title": "Status",
+            "type": "string"
+          },
+          "cycle_id": {
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "type": "null"
+              }
+            ],
+            "default": null,
+            "description": "Approval cycle that was resolved, or null when none was pending",
+            "title": "Cycle Id"
+          }
+        },
+        "title": "ApproveResponse",
+        "type": "object"
+      },
      "CommandRequest": {
        "properties": {
          "command": {
-            "description": "Slash command (e.g. /clear, /new, /resume)",
+            "description": "Workstream-local slash command (for example /clear or /instructions). Lifecycle commands such as /new and /resume are local-CLI-only; remote clients use the dedicated workstream endpoints.",
            "title": "Command",
            "type": "string"
          },
@@ -2399,6 +2772,24 @@
        "title": "CancelRequest",
        "type": "object"
      },
+      "CancelResponse": {
+        "properties": {
+          "status": {
+            "default": "ok",
+            "description": "Request outcome",
+            "title": "Status",
+            "type": "string"
+          },
+          "dropped": {
+            "additionalProperties": true,
+            "description": "Best-effort, credential-redacted snapshot of pending work affected by cancellation; keys are omitted when not observable",
+            "title": "Dropped",
+            "type": "object"
+          }
+        },
+        "title": "CancelResponse",
+        "type": "object"
+      },
      "RewindRequest": {
        "properties": {
          "turns": {
@@ -2428,15 +2819,43 @@
            "title": "Model",
            "type": "string"
          },
+          "judge_model": {
+            "default": "",
+            "description": "Optional judge model alias for this workstream. Empty uses the server's configured judge model.",
+            "title": "Judge Model",
+            "type": "string"
+          },
          "auto_approve": {
            "default": false,
            "description": "Auto-approve all tool calls",
            "title": "Auto Approve",
            "type": "boolean"
          },
+          "auto_approve_tools": {
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "items": {
+                  "type": "string"
+                },
+                "type": "array"
+              }
+            ],
+            "default": "",
+            "description": "Tool names to auto-approve even when auto_approve is false, accepted as either a comma-separated string or an array of strings.",
+            "title": "Auto Approve Tools"
+          },
+          "user_id": {
+            "default": "",
+            "description": "Optional workstream owner override. Honored only for trusted service identities (currently the console); ordinary callers remain bound to their authenticated user id.",
+            "title": "User Id",
+            "type": "string"
+          },
          "resume_ws": {
            "default": "",
-            "description": "Workstream ID to resume atomically during creation (empty = fresh start)",
+            "description": "Source workstream ID or alias to fork atomically into the new workstream (empty = fresh start)",
            "title": "Resume Ws",
            "type": "string"
          },
@@ -2473,7 +2892,7 @@
          },
          "client_type": {
            "default": "",
-            "description": "Client surface type (web, cli, chat). Defaults to web for server-created sessions.",
+            "description": "Client surface type (web, cli, chat, scheduled). Defaults to web for server-created sessions.",
            "title": "Client Type",
            "type": "string"
          },
@@ -2547,13 +2966,13 @@
          },
          "resumed": {
            "default": false,
-            "description": "Whether a previous workstream was resumed",
+            "description": "Whether the requested source was successfully forked",
            "title": "Resumed",
            "type": "boolean"
          },
          "message_count": {
            "default": 0,
-            "description": "Number of messages in the resumed workstream",
+            "description": "Number of messages cloned into the new workstream",
            "title": "Message Count",
            "type": "integer"
          },
@@ -2566,21 +2985,13 @@
            "type": "array"
          },
          "initial_message_status": {
-            "anyOf": [
-              {
-                "enum": [
-                  "queue_full",
-                  "refused_closed"
-                ],
-                "type": "string"
-              },
-              {
-                "type": "null"
-              }
-            ],
-            "default": null,
            "description": "Present ONLY when the workstream was created but its initial_message could not be delivered: 'queue_full' (a raced live worker's interjection queue was at capacity \u2014 resend via /send; any uploads stay staged) or 'refused_closed' (the workstream was closed mid-create). Absent whenever the message was dispatched.",
-            "title": "Initial Message Status"
+            "enum": [
+              "queue_full",
+              "refused_closed"
+            ],
+            "title": "Initial Message Status",
+            "type": "string"
          }
        },
        "required": [
@@ -2674,6 +3085,18 @@
            ],
            "default": null,
            "title": "Project Id"
+          },
+          "persistence_state": {
+            "default": "healthy",
+            "description": "Sanitized durable-history status for the loaded workstream: healthy, pending its first save, retrying automatically, or blocked by a permanent commit conflict. Older servers and unloaded rows default to healthy.",
+            "enum": [
+              "healthy",
+              "pending",
+              "retrying",
+              "conflict"
+            ],
+            "title": "Persistence State",
+            "type": "string"
          }
        },
        "required": [
@@ -2707,6 +3130,18 @@
            "$ref": "#/components/schemas/WorkstreamKind",
            "default": "interactive"
          },
+          "persistence_state": {
+            "default": "healthy",
+            "description": "Sanitized durable-history status: healthy, pending its first save, retrying automatically, or blocked by a permanent commit conflict.",
+            "enum": [
+              "healthy",
+              "pending",
+              "retrying",
+              "conflict"
+            ],
+            "title": "Persistence State",
+            "type": "string"
+          },
          "pending_approval": {
            "default": false,
            "description": "True when at least one approval cycle is live (a gate thread parked awaiting an operator approve/deny).  Mirrors the same field on ``DashboardWorkstream`` / cluster live projections so a freshly-loaded chat tab can render the inline approval gate from the detail snapshot before SSE replay arrives.",
@@ -2846,7 +3281,7 @@
            "type": "string"
          },
          "messages": {
-            "description": "Tail of the workstream's message history, projected to the canonical render shape (``role`` may be ``system`` for operator-context turns; flat tool_calls with verdict / output_assessment; top-level source / attachments / reasoning; derived denied / is_error / pending). Bounded by the ``limit`` query parameter (default 100, max 500).",
+            "description": "Requested limit-bounded tail of one authoritative total accepted conversation-row prefix, projected to the canonical render shape. Roles include ``user``, ``assistant``, ``tool``, and ``system``; compaction checkpoints project as ``role=system, source=compaction`` and cancellation-generated assistant/tool markers appear when present. The projection also carries flat tool_calls with verdict / output_assessment; top-level source / attachments / reasoning; and derived denied / is_error / pending. Bounded by the ``limit`` query parameter (default 100, max 500).",
            "items": {
              "additionalProperties": true,
              "type": "object"
@@ -2866,6 +3301,19 @@
            "default": null,
            "description": "SSE resume cursor (a ``Last-Event-ID`` value). Non-null only when the trailing turn is an executing in-flight tool batch that the live ring buffer can replay: ``messages`` then omits that turn and the client opens its initial SSE with this cursor so the existing delta replay fast-forwards the in-flight turn (tool calls, results, prompts) instead of the lossy synthetic snapshot. Null on every other read \u2014 the client connects fresh.",
            "title": "Cursor"
+          },
+          "handoff_token": {
+            "anyOf": [
+              {
+                "type": "string"
+              },
+              {
+                "type": "null"
+              }
+            ],
+            "default": null,
+            "description": "Opaque token naming the exact accepted conversation-row prefix used for this render. Present only while the workstream is loaded. A client that renders this response passes the token once as the initial event stream's ``history_token`` query parameter; the server atomically validates it while registering the listener. Admission of a later row changes the token; durable acknowledgement does not. Clients must not inspect, persist, or reuse it for later reconnects.",
+            "title": "Handoff Token"
          }
        },
        "required": [
@@ -3023,6 +3471,18 @@
            "default": null,
            "title": "Project Id"
          },
+          "persistence_state": {
+            "default": "healthy",
+            "description": "Sanitized durable-history status for this live row. Contains no storage error, commit key, retry count, or conversation content.",
+            "enum": [
+              "healthy",
+              "pending",
+              "retrying",
+              "conflict"
+            ],
+            "title": "Persistence State",
+            "type": "string"
+          },
          "pending_approval_details": {
            "description": "Inline approval payload for the coordinator children-tree UI: EVERY live approval cycle, oldest first \u2014 parallel task agents gate concurrently, so a workstream can hold several prompts at once.  Each entry carries the cycle's items + per-call_id LLM verdict cache so a coord can render approve/deny buttons + judge pill without a separate per-child round-trip; resolve each with its ``cycle_id``.  Empty when no approval is pending.  Also surfaced (verbatim) on ``GET /v1/api/cluster/ws/live`` via the ``_CLUSTER_WS_LIVE_KEYS`` projection.  Replaces 1.6's ``pending_approval_detail`` single-object field (breaking, 1.7).",
            "items": {
@@ -3534,32 +3994,42 @@
        "properties": {
          "name": {
            "description": "Memory identifier (normalized to snake_case)",
+            "maxLength": 256,
+            "minLength": 1,
            "title": "Name",
            "type": "string"
          },
          "content": {
            "description": "Memory content",
            "maxLength": 65536,
+            "minLength": 1,
            "title": "Content",
            "type": "string"
          },
          "description": {
-            "default": "",
-            "description": "Short description for relevance matching",
+            "description": "Required non-empty description used for relevance matching",
+            "minLength": 1,
            "title": "Description",
            "type": "string"
          },
          "type": {
-            "default": "general",
-            "description": "Memory type",
-            "enum": [
-              "user",
-              "general",
-              "feedback",
-              "reference"
+            "anyOf": [
+              {
+                "enum": [
+                  "user",
+                  "general",
+                  "feedback",
+                  "reference"
+                ],
+                "type": "string"
+              },
+              {
+                "type": "null"
+              }
            ],
-            "title": "Type",
-            "type": "string"
+            "default": null,
+            "description": "Memory type; omission preserves it on update and defaults on insert",
+            "title": "Type"
          },
          "scope": {
            "default": "global",
@@ -3581,7 +4051,8 @@
        },
        "required": [
          "name",
-          "content"
+          "content",
+          "description"
        ],
        "title": "SaveMemoryRequest",
        "type": "object"
@@ -3675,6 +4146,7 @@
        "properties": {
          "query": {
            "description": "Search query text",
+            "minLength": 1,
            "title": "Query",
            "type": "string"
          },
@@ -13,40 +13,6 @@
        "vitest": "^4.1"
      }
    },
-    "node_modules/@emnapi/core": {
-      "version": "1.11.1",
-      "resolved": "https://registry.npmjs.org/@emnapi/core/-/core-1.11.1.tgz",
-      "integrity": "sha512-RSvbQmHzdKzNsLYa/wHrbc3KN4sYLKAdPZxqiM2HATqv/SBk2/ENSHpvXGaLOMcsAyz0poEGqkmmKYG3OWiJEQ==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "dependencies": {
-        "@emnapi/wasi-threads": "1.2.2",
-        "tslib": "^2.4.0"
-      }
-    },
-    "node_modules/@emnapi/runtime": {
-      "version": "1.11.1",
-      "resolved": "https://registry.npmjs.org/@emnapi/runtime/-/runtime-1.11.1.tgz",
-      "integrity": "sha512-vgj7R3y3Wgx24IQaGPA/R6YFXLHVMOZ0uVEyIQPaWs+rd1AzfEMXlAC22FYwO1XkKR6NPsq7mUandH8oIRdZFw==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "dependencies": {
-        "tslib": "^2.4.0"
-      }
-    },
-    "node_modules/@emnapi/wasi-threads": {
-      "version": "1.2.2",
-      "resolved": "https://registry.npmjs.org/@emnapi/wasi-threads/-/wasi-threads-1.2.2.tgz",
-      "integrity": "sha512-c95qOXkHdydNKhscBTebqEC1CVAZpyqOfVfBzQ1qgzyl3gfeldUjIggDbIZgDKsHLgnsM+igH7TJ/eAasaVuMA==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "dependencies": {
-        "tslib": "^2.4.0"
-      }
-    },
    "node_modules/@jridgewell/sourcemap-codec": {
      "version": "1.5.5",
      "resolved": "https://registry.npmjs.org/@jridgewell/sourcemap-codec/-/sourcemap-codec-1.5.5.tgz",
@@ -54,29 +20,10 @@
      "dev": true,
      "license": "MIT"
    },
-    "node_modules/@napi-rs/wasm-runtime": {
-      "version": "1.1.6",
-      "resolved": "https://registry.npmjs.org/@napi-rs/wasm-runtime/-/wasm-runtime-1.1.6.tgz",
-      "integrity": "sha512-ZLv/JdUfkvOy9eCnnBaGfiO+XimbjebAeO+MRQqD/B+FR1tnRN0tpKSJHRbE8sFfS6aqsXZ67TQjfwfsxULVbg==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "dependencies": {
-        "@tybys/wasm-util": "^0.10.3"
-      },
-      "funding": {
-        "type": "github",
-        "url": "https://github.com/sponsors/Brooooooklyn"
-      },
-      "peerDependencies": {
-        "@emnapi/core": "^1.7.1",
-        "@emnapi/runtime": "^1.7.1"
-      }
-    },
    "node_modules/@oxc-project/types": {
-      "version": "0.138.0",
-      "resolved": "https://registry.npmjs.org/@oxc-project/types/-/types-0.138.0.tgz",
-      "integrity": "sha512-1a7ZKmrRTCoN1XMZ4L0PyyqrMnrNlLyPuOkdSX2MZg7IiIGRUyurNhAm73ptDOraoBcIordsIGKNPKUzy3ZmfA==",
+      "version": "0.143.0",
+      "resolved": "https://registry.npmjs.org/@oxc-project/types/-/types-0.143.0.tgz",
+      "integrity": "sha512-u6JZdLBTLotrNC9Vd6vPssINdzcCzleKAH6EJKImQb7GtYvX5keN2dxkoK44stCc4tffE6QQRtZTXVSzsLUlWA==",
      "dev": true,
      "license": "MIT",
      "funding": {
@@ -84,9 +31,9 @@
      }
    },
    "node_modules/@rolldown/binding-android-arm64": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/@rolldown/binding-android-arm64/-/binding-android-arm64-1.1.4.tgz",
-      "integrity": "sha512-EZLpf/8y7GXkkra90ML47kzik/GMP3EMcE9bPyHmRfxLC6z9+aW5A8poCsoxjrT5GfEcNAAvWwUHjvP1pUQkfw==",
+      "version": "1.2.3",
+      "resolved": "https://registry.npmjs.org/@rolldown/binding-android-arm64/-/binding-android-arm64-1.2.3.tgz",
+      "integrity": "sha512-zrJtHDcaZJ1Fp7xf4hNl+7seH9Cn/N5TwLYkhgXREtBwAd/jaqW3uqeHxpDugJLVICWg4eW44kOQEGJ1r6jCGw==",
      "cpu": [
        "arm64"
      ],
@@ -101,9 +48,9 @@
      }
    },
    "node_modules/@rolldown/binding-darwin-arm64": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/@rolldown/binding-darwin-arm64/-/binding-darwin-arm64-1.1.4.tgz",
-      "integrity": "sha512-aUi+HBvmYb7j8krl1+qJgkG8C17fO79gk3c+jPw4S8glRFc1DTija9S3EyaTSQUm5GJXYKDAsugBEhFHH2vYiQ==",
+      "version": "1.2.3",
+      "resolved": "https://registry.npmjs.org/@rolldown/binding-darwin-arm64/-/binding-darwin-arm64-1.2.3.tgz",
+      "integrity": "sha512-ieIiibVCp0tX7TLu2cafoNPv8wJyYi01ekXpbf8q2j7F4rGAhhXb/eQh7ge9DRBY78GwmRQtvjZDux7EDbA8kA==",
      "cpu": [
        "arm64"
      ],
@@ -118,9 +65,9 @@
      }
    },
    "node_modules/@rolldown/binding-darwin-x64": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/@rolldown/binding-darwin-x64/-/binding-darwin-x64-1.1.4.tgz",
-      "integrity": "sha512-F7hHC3gwY11+vByKPRWqwGbeXWVgKmL+pTGCinaEhdihzBV2aQ0fvZOch9cXYUOKuKKq429HeYXOqQLc7wFCEg==",
+      "version": "1.2.3",
+      "resolved": "https://registry.npmjs.org/@rolldown/binding-darwin-x64/-/binding-darwin-x64-1.2.3.tgz",
+      "integrity": "sha512-Zh9tCon19eDXJoihx0rqKhMUlMYqzwj3aPsSuHmI4RWZh62dWUL+DJN4C5YQya5TcQBJU/Fe8+rY0jhXTQITqA==",
      "cpu": [
        "x64"
      ],
@@ -135,9 +82,9 @@
      }
    },
    "node_modules/@rolldown/binding-freebsd-x64": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/@rolldown/binding-freebsd-x64/-/binding-freebsd-x64-1.1.4.tgz",
-      "integrity": "sha512-sI5yw+7s92SK6odiEhD5lKCBlWcpjHS5qyqpVQbZAJ0fIzEUXrmbl3DH2ybR3PZogulNJF+COLtmA8hUfvkCCQ==",
+      "version": "1.2.3",
+      "resolved": "https://registry.npmjs.org/@rolldown/binding-freebsd-x64/-/binding-freebsd-x64-1.2.3.tgz",
+      "integrity": "sha512-nGbJWewA1wrXXZiQhjAT5rhibGfns5ZNkDVqxsO6zJ3f3YvpoDNNmGMSbbhLuXKjNScaBJVOAboztAWVespQMg==",
      "cpu": [
        "x64"
      ],
@@ -152,9 +99,9 @@
      }
    },
    "node_modules/@rolldown/binding-linux-arm-gnueabihf": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/@rolldown/binding-linux-arm-gnueabihf/-/binding-linux-arm-gnueabihf-1.1.4.tgz",
-      "integrity": "sha512-mCi0OKgEieFircrtVYmQAFGszRtMnZ6fpZAXrxanXAu7lqZcsK1E1RAaZNG0uKAnxox3B1f4EyQNnoyMfN1vAA==",
+      "version": "1.2.3",
+      "resolved": "https://registry.npmjs.org/@rolldown/binding-linux-arm-gnueabihf/-/binding-linux-arm-gnueabihf-1.2.3.tgz",
+      "integrity": "sha512-QNniJr5Kml0kDEB98jiDOJjXNroxIIi0IXIbdYzY26Xt1pVbeP62+KnoIZLwirOymX/0jDk/2gI/bNUv7A7OIw==",
      "cpu": [
        "arm"
      ],
@@ -169,9 +116,9 @@
      }
    },
    "node_modules/@rolldown/binding-linux-arm64-gnu": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/@rolldown/binding-linux-arm64-gnu/-/binding-linux-arm64-gnu-1.1.4.tgz",
-      "integrity": "sha512-B9Ial3Kv5sh0SHnB1g/QWcUQCEvCF6QKGAl4zXypYj65mVI+B4AhFBwPtSN7pDrJeIx8Z7zdy4ntx+wQABom7w==",
+      "version": "1.2.3",
+      "resolved": "https://registry.npmjs.org/@rolldown/binding-linux-arm64-gnu/-/binding-linux-arm64-gnu-1.2.3.tgz",
+      "integrity": "sha512-TkqEAcmmvH3I/q4114NB4RVt6241Dao48pF45uLcFGrwAaIn0iITgTAKP/dLjbN0R4buJjGb91+UHSoFmpgIWw==",
      "cpu": [
        "arm64"
      ],
@@ -189,9 +136,9 @@
      }
    },
    "node_modules/@rolldown/binding-linux-arm64-musl": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/@rolldown/binding-linux-arm64-musl/-/binding-linux-arm64-musl-1.1.4.tgz",
-      "integrity": "sha512-lZVym0PuHE1KZ22gmFTC15lAkrg9iTszR617oYRB/iPY1A56ywoJzVKOJBKaot5RiikCObmur6pogpse3gRcng==",
+      "version": "1.2.3",
+      "resolved": "https://registry.npmjs.org/@rolldown/binding-linux-arm64-musl/-/binding-linux-arm64-musl-1.2.3.tgz",
+      "integrity": "sha512-NHqjnxpsndf4MPymxteFAWHHfkTL8HjWh1KB7z23ofZ6QO2euONuxDXjat69dKZRALnGypg8k8SsK8vZJoXv1Q==",
      "cpu": [
        "arm64"
      ],
@@ -209,9 +156,9 @@
      }
    },
    "node_modules/@rolldown/binding-linux-ppc64-gnu": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/@rolldown/binding-linux-ppc64-gnu/-/binding-linux-ppc64-gnu-1.1.4.tgz",
-      "integrity": "sha512-t2DNiLJWNTbnEHyUzTumldML6ET4/g16467LZoDDJ3tSxGvguL5/NyC2lCsNKuyRycg9XeDQF5SSv+TNOhQEXg==",
+      "version": "1.2.3",
+      "resolved": "https://registry.npmjs.org/@rolldown/binding-linux-ppc64-gnu/-/binding-linux-ppc64-gnu-1.2.3.tgz",
+      "integrity": "sha512-6tbrbwfz5GB9DQ4Jwo6hy9v+vR31xZlvzZ6n5Xut6Hhx5PvrA9q/HsK8KMaYQp063iqZGXwNvZtYNLD7EM/x0w==",
      "cpu": [
        "ppc64"
      ],
@@ -229,9 +176,9 @@
      }
    },
    "node_modules/@rolldown/binding-linux-s390x-gnu": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/@rolldown/binding-linux-s390x-gnu/-/binding-linux-s390x-gnu-1.1.4.tgz",
-      "integrity": "sha512-0WIRnL1Uw4BvTZRLQt+PVgo6ZKTJadlC2btP+/EOXv2f/DWbY0rEgl+y834mIVwP1FkTlWVTrGGJXf12lru7EQ==",
+      "version": "1.2.3",
+      "resolved": "https://registry.npmjs.org/@rolldown/binding-linux-s390x-gnu/-/binding-linux-s390x-gnu-1.2.3.tgz",
+      "integrity": "sha512-oyuXxXmoZHjXC917IAPFAAv4wWAa0cM9afk8nx1+9/jNNOX1uPf8yDA6p7G0RypOfw/X0PQt5IfoquY1um+zSg==",
      "cpu": [
        "s390x"
      ],
@@ -249,9 +196,9 @@
      }
    },
    "node_modules/@rolldown/binding-linux-x64-gnu": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/@rolldown/binding-linux-x64-gnu/-/binding-linux-x64-gnu-1.1.4.tgz",
-      "integrity": "sha512-JWtGshGfX+oENAKonoNkqEJX+7hC8yfhi9GUyPX1VX4mdh1y5r+ZiJLR5XzAB0aoP6s/PcILsGjKq8O0mm24bw==",
+      "version": "1.2.3",
+      "resolved": "https://registry.npmjs.org/@rolldown/binding-linux-x64-gnu/-/binding-linux-x64-gnu-1.2.3.tgz",
+      "integrity": "sha512-TytMwF2KVGqP2tgd0I1OY0PAv78dZRAYcF5ssDzjM34SUXCED3uXvSd5+lHoC0bTD6eEdFz7LdQNCO1y0oVk9w==",
      "cpu": [
        "x64"
      ],
@@ -269,9 +216,9 @@
      }
    },
    "node_modules/@rolldown/binding-linux-x64-musl": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/@rolldown/binding-linux-x64-musl/-/binding-linux-x64-musl-1.1.4.tgz",
-      "integrity": "sha512-rT6yQcxUuXs4CnbofqwHRRV0iem349rLMYpTjkgQGLjrY4ado/eDzwPZPTCgTOlF6Nkp8NEv70yLMTn6qkWxsQ==",
+      "version": "1.2.3",
+      "resolved": "https://registry.npmjs.org/@rolldown/binding-linux-x64-musl/-/binding-linux-x64-musl-1.2.3.tgz",
+      "integrity": "sha512-/E9m3qstrJFVPoULV25mVQblSNExY2+kBsYe4sy0Tn0yOOgJ8wZbZt3KnRbF/XeU2Gl1STKUQnDNTqhIE5MD4A==",
      "cpu": [
        "x64"
      ],
@@ -289,9 +236,9 @@
      }
    },
    "node_modules/@rolldown/binding-openharmony-arm64": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/@rolldown/binding-openharmony-arm64/-/binding-openharmony-arm64-1.1.4.tgz",
-      "integrity": "sha512-KXMGoboq5cyaCQjDA4GLuRiOwBQ0EyFnJoVViLeZ45/3rFItRODEr+NdsBcVpll40hhNArlm/speWGRvj08LzA==",
+      "version": "1.2.3",
+      "resolved": "https://registry.npmjs.org/@rolldown/binding-openharmony-arm64/-/binding-openharmony-arm64-1.2.3.tgz",
+      "integrity": "sha512-Kr0OcsoQI816i6HOl3vFHpd1K0eZyh76zgfj4c1nTyaTsd5r2Mj1lwM4R90y/qaCfmTn9eHy0SKwi98eitRxug==",
      "cpu": [
        "arm64"
      ],
@@ -305,29 +252,10 @@
        "node": "^20.19.0 || >=22.12.0"
      }
    },
-    "node_modules/@rolldown/binding-wasm32-wasi": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/@rolldown/binding-wasm32-wasi/-/binding-wasm32-wasi-1.1.4.tgz",
-      "integrity": "sha512-5K83rb36oJiY7BCyE9zLZtGcPV4g5wvq+xwdO0XPIwDVZI8cyB/AUjkNXGb92/rnmezEkjMOpgY61rtwjQtFwg==",
-      "cpu": [
-        "wasm32"
-      ],
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "dependencies": {
-        "@emnapi/core": "1.11.1",
-        "@emnapi/runtime": "1.11.1",
-        "@napi-rs/wasm-runtime": "^1.1.6"
-      },
-      "engines": {
-        "node": "^20.19.0 || >=22.12.0"
-      }
-    },
    "node_modules/@rolldown/binding-win32-arm64-msvc": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/@rolldown/binding-win32-arm64-msvc/-/binding-win32-arm64-msvc-1.1.4.tgz",
-      "integrity": "sha512-PnWBtw3TV5KOg69HQQDR0mnQuyCmSGR2pAB4DC1rPF808fgKeTUMj2EOEyKATpgiuxuR5APQmiDO7PDgEjTFSA==",
+      "version": "1.2.3",
+      "resolved": "https://registry.npmjs.org/@rolldown/binding-win32-arm64-msvc/-/binding-win32-arm64-msvc-1.2.3.tgz",
+      "integrity": "sha512-hOtMwTqnME+/gJcH/PCZ0wn0zPUjiWOgkHpxbSJpfGKMezHltx1S7/k1SitzVa7Ww2cqrDDaFbZEhcJZO8o+Jw==",
      "cpu": [
        "arm64"
      ],
@@ -342,9 +270,9 @@
      }
    },
    "node_modules/@rolldown/binding-win32-x64-msvc": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/@rolldown/binding-win32-x64-msvc/-/binding-win32-x64-msvc-1.1.4.tgz",
-      "integrity": "sha512-M1lpniBePobTfsa7Ks9a199e1akxsXn+GYBUKsEzv3YFzOm1HJAMNwKI3qr0Zq+mxwx9gOZoTdP1yXRYsZUocQ==",
+      "version": "1.2.3",
+      "resolved": "https://registry.npmjs.org/@rolldown/binding-win32-x64-msvc/-/binding-win32-x64-msvc-1.2.3.tgz",
+      "integrity": "sha512-ekcqMMkI2PlhYnfzQnB/cEdYUVVJViWvoUyLrbzgDoi3Snfc1mVBwdnc306ufA5ejy8JSPjT2RlW1nQSjW7efg==",
      "cpu": [
        "x64"
      ],
@@ -372,17 +300,6 @@
      "dev": true,
      "license": "MIT"
    },
-    "node_modules/@tybys/wasm-util": {
-      "version": "0.10.3",
-      "resolved": "https://registry.npmjs.org/@tybys/wasm-util/-/wasm-util-0.10.3.tgz",
-      "integrity": "sha512-F3fo1MYrRJYL3zER0OUOmkutjr1Vp23m7OsSgp7nq4SP6OqX6C/56XFIPAl5bt3zaBRjmW7SGz3u/6LwFpYcOg==",
-      "dev": true,
-      "license": "MIT",
-      "optional": true,
-      "dependencies": {
-        "tslib": "^2.4.0"
-      }
-    },
    "node_modules/@types/chai": {
      "version": "5.2.3",
      "resolved": "https://registry.npmjs.org/@types/chai/-/chai-5.2.3.tgz",
@@ -899,9 +816,9 @@
      }
    },
    "node_modules/es-module-lexer": {
-      "version": "2.3.0",
-      "resolved": "https://registry.npmjs.org/es-module-lexer/-/es-module-lexer-2.3.0.tgz",
-      "integrity": "sha512-KLdwQm2NvGLDkQDCGvmiQrhkd0JbMzXthwQAUgWjQuQdBLFa3eiBP5arXZyA+f8x+x7OXgud6bq2rxjGtHV2tw==",
+      "version": "2.3.1",
+      "resolved": "https://registry.npmjs.org/es-module-lexer/-/es-module-lexer-2.3.1.tgz",
+      "integrity": "sha512-shc1dbU90Yl/xq1QrC7QRtfcwURZuVRfPhZbDoldJ1cn1gzDvBaBWlv0eFolj5+0znnPJz5TXLxsN77X/12KTA==",
      "dev": true,
      "license": "MIT"
    },
@@ -959,9 +876,9 @@
      }
    },
    "node_modules/lightningcss": {
-      "version": "1.32.0",
-      "resolved": "https://registry.npmjs.org/lightningcss/-/lightningcss-1.32.0.tgz",
-      "integrity": "sha512-NXYBzinNrblfraPGyrbPoD19C1h9lfI/1mzgWYvXUTe414Gz/X1FD2XBZSZM7rRTrMA8JL3OtAaGifrIKhQ5yQ==",
+      "version": "1.33.0",
+      "resolved": "https://registry.npmjs.org/lightningcss/-/lightningcss-1.33.0.tgz",
+      "integrity": "sha512-WkUDrojuJs0xkgGf2udWxa3yGBRxPtxUkB79i6aCZLRgc7PM8fZe9TosfPDcvEpQZbuFASnHYmRLBLUbmLOIIA==",
      "dev": true,
      "license": "MPL-2.0",
      "dependencies": {
@@ -975,23 +892,23 @@
        "url": "https://opencollective.com/parcel"
      },
      "optionalDependencies": {
-        "lightningcss-android-arm64": "1.32.0",
-        "lightningcss-darwin-arm64": "1.32.0",
-        "lightningcss-darwin-x64": "1.32.0",
-        "lightningcss-freebsd-x64": "1.32.0",
-        "lightningcss-linux-arm-gnueabihf": "1.32.0",
-        "lightningcss-linux-arm64-gnu": "1.32.0",
-        "lightningcss-linux-arm64-musl": "1.32.0",
-        "lightningcss-linux-x64-gnu": "1.32.0",
-        "lightningcss-linux-x64-musl": "1.32.0",
-        "lightningcss-win32-arm64-msvc": "1.32.0",
-        "lightningcss-win32-x64-msvc": "1.32.0"
+        "lightningcss-android-arm64": "1.33.0",
+        "lightningcss-darwin-arm64": "1.33.0",
+        "lightningcss-darwin-x64": "1.33.0",
+        "lightningcss-freebsd-x64": "1.33.0",
+        "lightningcss-linux-arm-gnueabihf": "1.33.0",
+        "lightningcss-linux-arm64-gnu": "1.33.0",
+        "lightningcss-linux-arm64-musl": "1.33.0",
+        "lightningcss-linux-x64-gnu": "1.33.0",
+        "lightningcss-linux-x64-musl": "1.33.0",
+        "lightningcss-win32-arm64-msvc": "1.33.0",
+        "lightningcss-win32-x64-msvc": "1.33.0"
      }
    },
    "node_modules/lightningcss-android-arm64": {
-      "version": "1.32.0",
-      "resolved": "https://registry.npmjs.org/lightningcss-android-arm64/-/lightningcss-android-arm64-1.32.0.tgz",
-      "integrity": "sha512-YK7/ClTt4kAK0vo6w3X+Pnm0D2cf2vPHbhOXdoNti1Ga0al1P4TBZhwjATvjNwLEBCnKvjJc2jQgHXH0NEwlAg==",
+      "version": "1.33.0",
+      "resolved": "https://registry.npmjs.org/lightningcss-android-arm64/-/lightningcss-android-arm64-1.33.0.tgz",
+      "integrity": "sha512-gEpRTalKdosp4Bb8qWtc2iOgE5SeIHlpS1up9bFq2wAyYhl1UdTObYiHe98zEM9SQvSoqQZ1IQD0JNpg3Ml5pg==",
      "cpu": [
        "arm64"
      ],
@@ -1010,9 +927,9 @@
      }
    },
    "node_modules/lightningcss-darwin-arm64": {
-      "version": "1.32.0",
-      "resolved": "https://registry.npmjs.org/lightningcss-darwin-arm64/-/lightningcss-darwin-arm64-1.32.0.tgz",
-      "integrity": "sha512-RzeG9Ju5bag2Bv1/lwlVJvBE3q6TtXskdZLLCyfg5pt+HLz9BqlICO7LZM7VHNTTn/5PRhHFBSjk5lc4cmscPQ==",
+      "version": "1.33.0",
+      "resolved": "https://registry.npmjs.org/lightningcss-darwin-arm64/-/lightningcss-darwin-arm64-1.33.0.tgz",
+      "integrity": "sha512-Sciaz8eenNTKn9b3t7+xr0ipTp9YxKQY4npwQ3mrRuL0BAVHBLyZxofhaKBAVtzmtRZ/zTyo0/to4B1uWG/Djg==",
      "cpu": [
        "arm64"
      ],
@@ -1031,9 +948,9 @@
      }
    },
    "node_modules/lightningcss-darwin-x64": {
-      "version": "1.32.0",
-      "resolved": "https://registry.npmjs.org/lightningcss-darwin-x64/-/lightningcss-darwin-x64-1.32.0.tgz",
-      "integrity": "sha512-U+QsBp2m/s2wqpUYT/6wnlagdZbtZdndSmut/NJqlCcMLTWp5muCrID+K5UJ6jqD2BFshejCYXniPDbNh73V8w==",
+      "version": "1.33.0",
+      "resolved": "https://registry.npmjs.org/lightningcss-darwin-x64/-/lightningcss-darwin-x64-1.33.0.tgz",
+      "integrity": "sha512-Z5UPAxzrjlWNNyGy6i65cJzzvgJ5D3T6wMvs+gWpY9d7qRhANrxqAp6LhxIgZhWEw18RfJTGcRxjuLIBr+m8XQ==",
      "cpu": [
        "x64"
      ],
@@ -1052,9 +969,9 @@
      }
    },
    "node_modules/lightningcss-freebsd-x64": {
-      "version": "1.32.0",
-      "resolved": "https://registry.npmjs.org/lightningcss-freebsd-x64/-/lightningcss-freebsd-x64-1.32.0.tgz",
-      "integrity": "sha512-JCTigedEksZk3tHTTthnMdVfGf61Fky8Ji2E4YjUTEQX14xiy/lTzXnu1vwiZe3bYe0q+SpsSH/CTeDXK6WHig==",
+      "version": "1.33.0",
+      "resolved": "https://registry.npmjs.org/lightningcss-freebsd-x64/-/lightningcss-freebsd-x64-1.33.0.tgz",
+      "integrity": "sha512-QQM/Ti/hQajJwCY+RiWuCZ9sdtI/XQk7nDK5vC8kkdwixezOlDgvDx7+RT+QjK6FcFT4MpsuoBnHIo/O3StRRg==",
      "cpu": [
        "x64"
      ],
@@ -1073,9 +990,9 @@
      }
    },
    "node_modules/lightningcss-linux-arm-gnueabihf": {
-      "version": "1.32.0",
-      "resolved": "https://registry.npmjs.org/lightningcss-linux-arm-gnueabihf/-/lightningcss-linux-arm-gnueabihf-1.32.0.tgz",
-      "integrity": "sha512-x6rnnpRa2GL0zQOkt6rts3YDPzduLpWvwAF6EMhXFVZXD4tPrBkEFqzGowzCsIWsPjqSK+tyNEODUBXeeVHSkw==",
+      "version": "1.33.0",
+      "resolved": "https://registry.npmjs.org/lightningcss-linux-arm-gnueabihf/-/lightningcss-linux-arm-gnueabihf-1.33.0.tgz",
+      "integrity": "sha512-N7FVBe6iS24MlM6R/4RBTxGhQheZGs7tiQ9U32UtF75NzP5Q7xWPRqLBCKxlRQRk3rY1jCIPLzx7WzOhuUIRLQ==",
      "cpu": [
        "arm"
      ],
@@ -1094,9 +1011,9 @@
      }
    },
    "node_modules/lightningcss-linux-arm64-gnu": {
-      "version": "1.32.0",
-      "resolved": "https://registry.npmjs.org/lightningcss-linux-arm64-gnu/-/lightningcss-linux-arm64-gnu-1.32.0.tgz",
-      "integrity": "sha512-0nnMyoyOLRJXfbMOilaSRcLH3Jw5z9HDNGfT/gwCPgaDjnx0i8w7vBzFLFR1f6CMLKF8gVbebmkUN3fa/kQJpQ==",
+      "version": "1.33.0",
+      "resolved": "https://registry.npmjs.org/lightningcss-linux-arm64-gnu/-/lightningcss-linux-arm64-gnu-1.33.0.tgz",
+      "integrity": "sha512-j2v/itmy4HlNxlc6voKXYgBqNi0Ng2LShg4z7GufpEgs05P+2suBVyi9I6YHq5uoVFx9ETin3eCEhLVyXGQnKg==",
      "cpu": [
        "arm64"
      ],
@@ -1118,9 +1035,9 @@
      }
    },
    "node_modules/lightningcss-linux-arm64-musl": {
-      "version": "1.32.0",
-      "resolved": "https://registry.npmjs.org/lightningcss-linux-arm64-musl/-/lightningcss-linux-arm64-musl-1.32.0.tgz",
-      "integrity": "sha512-UpQkoenr4UJEzgVIYpI80lDFvRmPVg6oqboNHfoH4CQIfNA+HOrZ7Mo7KZP02dC6LjghPQJeBsvXhJod/wnIBg==",
+      "version": "1.33.0",
+      "resolved": "https://registry.npmjs.org/lightningcss-linux-arm64-musl/-/lightningcss-linux-arm64-musl-1.33.0.tgz",
+      "integrity": "sha512-yiO5ROMuYQgXbC60yjZU5CYSFZGKXL0HFATXt9mHJn1+zW55oCtMI9NfcVhYLMFDL7gV7oBPon/EmMMGg2OvtQ==",
      "cpu": [
        "arm64"
      ],
@@ -1142,9 +1059,9 @@
      }
    },
    "node_modules/lightningcss-linux-x64-gnu": {
-      "version": "1.32.0",
-      "resolved": "https://registry.npmjs.org/lightningcss-linux-x64-gnu/-/lightningcss-linux-x64-gnu-1.32.0.tgz",
-      "integrity": "sha512-V7Qr52IhZmdKPVr+Vtw8o+WLsQJYCTd8loIfpDaMRWGUZfBOYEJeyJIkqGIDMZPwPx24pUMfwSxxI8phr/MbOA==",
+      "version": "1.33.0",
+      "resolved": "https://registry.npmjs.org/lightningcss-linux-x64-gnu/-/lightningcss-linux-x64-gnu-1.33.0.tgz",
+      "integrity": "sha512-ar+Ju7LmcN0Jo4FpL4hpFybwNG9/3A/Br5KW2n2jyODg3MEZXaDYADdemoNS+BDNfMgKvylJLj4S5tyRActuAg==",
      "cpu": [
        "x64"
      ],
@@ -1166,9 +1083,9 @@
      }
    },
    "node_modules/lightningcss-linux-x64-musl": {
-      "version": "1.32.0",
-      "resolved": "https://registry.npmjs.org/lightningcss-linux-x64-musl/-/lightningcss-linux-x64-musl-1.32.0.tgz",
-      "integrity": "sha512-bYcLp+Vb0awsiXg/80uCRezCYHNg1/l3mt0gzHnWV9XP1W5sKa5/TCdGWaR/zBM2PeF/HbsQv/j2URNOiVuxWg==",
+      "version": "1.33.0",
+      "resolved": "https://registry.npmjs.org/lightningcss-linux-x64-musl/-/lightningcss-linux-x64-musl-1.33.0.tgz",
+      "integrity": "sha512-RYiYbkokw0trfKqqzfF55lginwEPrD3OJDfTuJzFs1MK6iFnDenaz1fqLLtX4ITG3OktJQXOeTaw1awrBAlZPw==",
      "cpu": [
        "x64"
      ],
@@ -1190,9 +1107,9 @@
      }
    },
    "node_modules/lightningcss-win32-arm64-msvc": {
-      "version": "1.32.0",
-      "resolved": "https://registry.npmjs.org/lightningcss-win32-arm64-msvc/-/lightningcss-win32-arm64-msvc-1.32.0.tgz",
-      "integrity": "sha512-8SbC8BR40pS6baCM8sbtYDSwEVQd4JlFTOlaD3gWGHfThTcABnNDBda6eTZeqbofalIJhFx0qKzgHJmcPTnGdw==",
+      "version": "1.33.0",
+      "resolved": "https://registry.npmjs.org/lightningcss-win32-arm64-msvc/-/lightningcss-win32-arm64-msvc-1.33.0.tgz",
+      "integrity": "sha512-1K+MPfLSFVpphzpdbfkhlWk6wBrTObBzS2T6db10PNOZgR9GoVsAWzwNyuhUYYbTp23j+4RrncfujZ4uAzXvwA==",
      "cpu": [
        "arm64"
      ],
@@ -1211,9 +1128,9 @@
      }
    },
    "node_modules/lightningcss-win32-x64-msvc": {
-      "version": "1.32.0",
-      "resolved": "https://registry.npmjs.org/lightningcss-win32-x64-msvc/-/lightningcss-win32-x64-msvc-1.32.0.tgz",
-      "integrity": "sha512-Amq9B/SoZYdDi1kFrojnoqPLxYhQ4Wo5XiL8EVJrVsB8ARoC1PWW6VGtT0WKCemjy8aC+louJnjS7U18x3b06Q==",
+      "version": "1.33.0",
+      "resolved": "https://registry.npmjs.org/lightningcss-win32-x64-msvc/-/lightningcss-win32-x64-msvc-1.33.0.tgz",
+      "integrity": "sha512-OlEICDx/Xl0FqSp4bry8zFnCvGpig3Gl4gCquvYwHuqJKEC1+n9NgDniFvqHGmMv1ZkqDJrDqKKSykTDX+ehuA==",
      "cpu": [
        "x64"
      ],
@@ -1242,9 +1159,9 @@
      }
    },
    "node_modules/nanoid": {
-      "version": "3.3.15",
-      "resolved": "https://registry.npmjs.org/nanoid/-/nanoid-3.3.15.tgz",
-      "integrity": "sha512-y7Wygv/7mEOvxTuEQDB8StXdMRBWf1kR/tlhAzBRUFkB2jfcLOAxO/SHmOO2zgz1pVgK29/kyupn059/bCHdjA==",
+      "version": "3.3.18",
+      "resolved": "https://registry.npmjs.org/nanoid/-/nanoid-3.3.18.tgz",
+      "integrity": "sha512-DTg4MJbGMWkfi6VZFdNt2/caMbQy4Ou+Op/hJQvGEWcnVfoA1QA+xzRKAzw9jD6+GVOOeYr/mIcuDSdug6F6+w==",
      "dev": true,
      "funding": [
        {
@@ -1261,9 +1178,9 @@
      }
    },
    "node_modules/obug": {
-      "version": "2.1.3",
-      "resolved": "https://registry.npmjs.org/obug/-/obug-2.1.3.tgz",
-      "integrity": "sha512-9miFgM2OFba7hB+pRgvtV84pYTBaoTHohvmIgiRt6dRIzbwEOIaNaP+dIlGs2fNFoB0SeISs0Jz5WFVRid6Xyg==",
+      "version": "2.1.4",
+      "resolved": "https://registry.npmjs.org/obug/-/obug-2.1.4.tgz",
+      "integrity": "sha512-4a+OsYv9UktOJKE+l1A4OufDgdRF9PifWj+tJnHURo/P+WOxpG4GzUFL9qCalmWauao6ogiG+QvnCovwPoyAWA==",
      "dev": true,
      "funding": [
        "https://github.com/sponsors/sxzz",
@@ -1302,9 +1219,9 @@
      }
    },
    "node_modules/postcss": {
-      "version": "8.5.16",
-      "resolved": "https://registry.npmjs.org/postcss/-/postcss-8.5.16.tgz",
-      "integrity": "sha512-vuwillviilfKZsg0VGj5R/YwwcHx4SLsIOI/7K6mQkWx+l5cUHTjj5g0AasTBcyXsbfTgrwsUNmVUb5xVwyPwg==",
+      "version": "8.5.26",
+      "resolved": "https://registry.npmjs.org/postcss/-/postcss-8.5.26.tgz",
+      "integrity": "sha512-u82N74LFzG8ca+dD8puPnplTXoGH4fTPpVGuIbt36G3qvNlkvfD0lEAZSxaly3KX8TS/L1A1gsCEmvKmBcVbkQ==",
      "dev": true,
      "funding": [
        {
@@ -1322,7 +1239,7 @@
      ],
      "license": "MIT",
      "dependencies": {
-        "nanoid": "^3.3.12",
+        "nanoid": "^3.3.17",
        "picocolors": "^1.1.1",
        "source-map-js": "^1.2.1"
      },
@@ -1331,13 +1248,13 @@
      }
    },
    "node_modules/rolldown": {
-      "version": "1.1.4",
-      "resolved": "https://registry.npmjs.org/rolldown/-/rolldown-1.1.4.tgz",
-      "integrity": "sha512-IjZYiLxZwpnhwhdBH2ugdTGVSdhCQUmLxLoqyjiL0JxYjyRst+5a0P3xfrTxJ5F638j4Mvvw5FAX5XE6eHpXbA==",
+      "version": "1.2.3",
+      "resolved": "https://registry.npmjs.org/rolldown/-/rolldown-1.2.3.tgz",
+      "integrity": "sha512-rn9wpmxplLf7NLNyCk9FyWh3FM43DbY8jOzCdEPzH7uflhTftRbCEpqi6Ly2osgoU8OwObtmavMbWLaWy4LX7A==",
      "dev": true,
      "license": "MIT",
      "dependencies": {
-        "@oxc-project/types": "=0.138.0",
+        "@oxc-project/types": "=0.143.0",
        "@rolldown/pluginutils": "^1.0.0"
      },
      "bin": {
@@ -1347,21 +1264,20 @@
        "node": "^20.19.0 || >=22.12.0"
      },
      "optionalDependencies": {
-        "@rolldown/binding-android-arm64": "1.1.4",
-        "@rolldown/binding-darwin-arm64": "1.1.4",
-        "@rolldown/binding-darwin-x64": "1.1.4",
-        "@rolldown/binding-freebsd-x64": "1.1.4",
-        "@rolldown/binding-linux-arm-gnueabihf": "1.1.4",
-        "@rolldown/binding-linux-arm64-gnu": "1.1.4",
-        "@rolldown/binding-linux-arm64-musl": "1.1.4",
-        "@rolldown/binding-linux-ppc64-gnu": "1.1.4",
-        "@rolldown/binding-linux-s390x-gnu": "1.1.4",
-        "@rolldown/binding-linux-x64-gnu": "1.1.4",
-        "@rolldown/binding-linux-x64-musl": "1.1.4",
-        "@rolldown/binding-openharmony-arm64": "1.1.4",
-        "@rolldown/binding-wasm32-wasi": "1.1.4",
-        "@rolldown/binding-win32-arm64-msvc": "1.1.4",
-        "@rolldown/binding-win32-x64-msvc": "1.1.4"
+        "@rolldown/binding-android-arm64": "1.2.3",
+        "@rolldown/binding-darwin-arm64": "1.2.3",
+        "@rolldown/binding-darwin-x64": "1.2.3",
+        "@rolldown/binding-freebsd-x64": "1.2.3",
+        "@rolldown/binding-linux-arm-gnueabihf": "1.2.3",
+        "@rolldown/binding-linux-arm64-gnu": "1.2.3",
+        "@rolldown/binding-linux-arm64-musl": "1.2.3",
+        "@rolldown/binding-linux-ppc64-gnu": "1.2.3",
+        "@rolldown/binding-linux-s390x-gnu": "1.2.3",
+        "@rolldown/binding-linux-x64-gnu": "1.2.3",
+        "@rolldown/binding-linux-x64-musl": "1.2.3",
+        "@rolldown/binding-openharmony-arm64": "1.2.3",
+        "@rolldown/binding-win32-arm64-msvc": "1.2.3",
+        "@rolldown/binding-win32-x64-msvc": "1.2.3"
      }
    },
    "node_modules/siginfo": {
@@ -1389,9 +1305,9 @@
      "license": "MIT"
    },
    "node_modules/std-env": {
-      "version": "4.1.0",
-      "resolved": "https://registry.npmjs.org/std-env/-/std-env-4.1.0.tgz",
-      "integrity": "sha512-Rq7ybcX2RuC55r9oaPVEW7/xu3tj8u4GeBYHBWCychFtzMIr86A7e3PPEBPT37sHStKX3+TiX/Fr/ACmJLVlLQ==",
+      "version": "4.2.0",
+      "resolved": "https://registry.npmjs.org/std-env/-/std-env-4.2.0.tgz",
+      "integrity": "sha512-oCUKSupKTHX53EyjDtuZQ64pjLJ6yYCtpmEw0goYxtjG9KpbRe8KAsl2tBUGU9DyMcJ0RwJ8GqJAFzMXcXW1Rw==",
      "dev": true,
      "license": "MIT"
    },
@@ -1403,9 +1319,9 @@
      "license": "MIT"
    },
    "node_modules/tinyexec": {
-      "version": "1.2.4",
-      "resolved": "https://registry.npmjs.org/tinyexec/-/tinyexec-1.2.4.tgz",
-      "integrity": "sha512-SHf/r48b7vOrjve9PxJo3MN5v5yuyjHvdUcrQffT3WXMUfnGmHDVbC4k3sHJaJTgZCwpUplIaAo5ANtMyp3YHg==",
+      "version": "1.3.0",
+      "resolved": "https://registry.npmjs.org/tinyexec/-/tinyexec-1.3.0.tgz",
+      "integrity": "sha512-QKAl9m8gWWGHV8jZcPeym6j+XULi6tOf1mT83WYJ4Lk2ytW/uwAWkrP0uFsdoYMdueVJ0qs26wZ+23xeB4ibNQ==",
      "dev": true,
      "license": "MIT",
      "engines": {
@@ -1430,23 +1346,15 @@
      }
    },
    "node_modules/tinyrainbow": {
-      "version": "3.1.0",
-      "resolved": "https://registry.npmjs.org/tinyrainbow/-/tinyrainbow-3.1.0.tgz",
-      "integrity": "sha512-Bf+ILmBgretUrdJxzXM0SgXLZ3XfiaUuOj/IKQHuTXip+05Xn+uyEYdVg0kYDipTBcLrCVyUzAPz7QmArb0mmw==",
+      "version": "3.1.1",
+      "resolved": "https://registry.npmjs.org/tinyrainbow/-/tinyrainbow-3.1.1.tgz",
+      "integrity": "sha512-yau8yJdTt989Mm0Bd/236QnzEiPf2xLLTqUZRUJOo/3CB078LSwzei343DgtJVmfJKJE3TMINY1u42SQsP6mXw==",
      "dev": true,
      "license": "MIT",
      "engines": {
        "node": ">=14.0.0"
      }
    },
-    "node_modules/tslib": {
-      "version": "2.8.1",
-      "resolved": "https://registry.npmjs.org/tslib/-/tslib-2.8.1.tgz",
-      "integrity": "sha512-oJFu94HQb+KVduSUQL7wnpmqnfmLsOA/nAh6b6EH0wCEoK0/mPeXU6c3wKDV83MkOuHPRHtSXKKU99IBazS/2w==",
-      "dev": true,
-      "license": "0BSD",
-      "optional": true
-    },
    "node_modules/typescript": {
      "version": "7.0.2",
      "resolved": "https://registry.npmjs.org/typescript/-/typescript-7.0.2.tgz",
@@ -1483,16 +1391,16 @@
      }
    },
    "node_modules/vite": {
-      "version": "8.1.3",
-      "resolved": "https://registry.npmjs.org/vite/-/vite-8.1.3.tgz",
-      "integrity": "sha512-Ds+gBRbj0lwRO2Y5hwnUBdxSwlAve9LeRyU4sNnAr0ewW0gWF0n5bgXgUzbgZ49MV9BVUAQUFYVcDUcilUExMA==",
+      "version": "8.2.1",
+      "resolved": "https://registry.npmjs.org/vite/-/vite-8.2.1.tgz",
+      "integrity": "sha512-EU/eS7BH3XROHh2YnBefjM6DBKA6ZeMZEYQbj7NLWg5wHYlhB8B/Mayd5XsgWq+NFYccDOTemRpdETWR6Ka/lw==",
      "dev": true,
      "license": "MIT",
      "dependencies": {
-        "lightningcss": "^1.32.0",
-        "picomatch": "^4.0.4",
-        "postcss": "^8.5.16",
-        "rolldown": "~1.1.3",
+        "lightningcss": "^1.33.0",
+        "picomatch": "^4.0.5",
+        "postcss": "^8.5.25",
+        "rolldown": "~1.2.1",
        "tinyglobby": "^0.2.17"
      },
      "bin": {
@@ -1509,7 +1417,7 @@
      },
      "peerDependencies": {
        "@types/node": "^20.19.0 || >=22.12.0",
-        "@vitejs/devtools": "^0.3.0",
+        "@vitejs/devtools": "^0.4.0",
        "esbuild": "^0.27.0 || ^0.28.0",
        "jiti": ">=1.21.0",
        "less": "^4.0.0",
@@ -18,8 +18,6 @@ import type {
  ConsoleCreateWsRequest,
  ConsoleCreateWsResponse,
  ConsoleHealthResponse,
-  CreateWorkstreamRequest,
-  CreateWorkstreamResponse,
  ListAttachmentsResponse,
  CreateMcpServerRequest,
  CreatePolicyOptions,
@@ -39,6 +37,9 @@ import type {
  McpServerDetail,
  RegistryInstallRequest,
  RegistrySearchResponse,
+  RouteCreateRequest,
+  RouteCreateResponse,
+  RouteLiveResponse,
  SkillDiscoverResponse,
  SkillInfo,
  SkillInstallRequest,
@@ -154,10 +155,8 @@ export class TurnstoneConsole extends BaseClient {
   * owning node directly.
   */
  async routeCreateWorkstream(
-    opts?: CreateWorkstreamRequest & { target_node?: string },
-  ): Promise<
-    CreateWorkstreamResponse & { node_url?: string; node_id?: string }
-  > {
+    opts?: RouteCreateRequest,
+  ): Promise<RouteCreateResponse> {
    const attachments = opts?.attachments;
    if (attachments && attachments.length > 0) {
      // The console's multipart route_create routes by `?ws_id=` only —
@@ -192,6 +191,13 @@ export class TurnstoneConsole extends BaseClient {
    });
  }

+  async routeWorkstreamLive(wsId: string): Promise<RouteLiveResponse> {
+    return this.request(
+      "GET",
+      `/v1/api/route/workstreams/${encodeURIComponent(wsId)}/live`,
+    );
+  }
+
  async routeUploadAttachment(
    wsId: string,
    file: AttachmentUpload,
@@ -1,4 +1,8 @@
-import type { ClusterOverviewResponse, ClusterSnapshotNode } from "./types.js";
+import type {
+  ClusterOverviewResponse,
+  ClusterSnapshotNode,
+  ConversationPersistenceState,
+} from "./types.js";

 // ---------------------------------------------------------------------------
 // Server SSE events
@@ -35,6 +39,37 @@ export interface HistoryEvent {
  messages: Array<Record<string, unknown>>;
 }

+/**
+ * The REST history rendered by the caller no longer names the live accepted
+ * row prefix. Stop this stream, refetch and render history, then open a new
+ * stream with its cursor and one-shot token. The SDK does not do this
+ * automatically.
+ */
+export interface HistoryResyncEvent {
+  type: "history_resync";
+  /** Present on registration-time handoff mismatches; implied by a scoped stream. */
+  ws_id?: string;
+  reason: string;
+}
+
+/** One accepted user row, projected live to every workstream consumer. */
+export interface UserTurnEvent {
+  type: "user_turn";
+  ws_id?: string;
+  content: string;
+  attachments?: Array<{
+    attachment_id: string;
+    kind: string;
+    filename: string;
+    mime_type: string;
+  }>;
+  sender?: string;
+  source?: string;
+  /** Optimistic-browser correlation only; not delivery idempotency. */
+  client_send_ids: string[];
+  _event_id?: number;
+}
+
 export interface ThinkingStartEvent {
  type: "thinking_start";
 }
@@ -112,6 +147,12 @@ export interface ToolResultEvent {
  name: string;
  output: string;
  is_error?: boolean;
+  preview?: Record<string, unknown>;
+  /** True only for the final guarded row accepted into conversation history. */
+  accepted?: boolean;
+  effect_status?: string;
+  /** Monotonic accepted-row identity; present for projection-capable clients. */
+  _event_id?: number;
 }

 export interface ToolOutputChunkEvent {
@@ -157,6 +198,49 @@ export interface CancelledEvent {
  type: "cancelled";
 }

+/**
+ * Context-compaction lifecycle. `start` carries `trigger` ("manual"/"auto";
+ * auto adds `where` + `pct`); `progress` carries chunked-summarization
+ * `part`/`total`/`depth` (or `retry_in`/`error` for a retry wait); `end`
+ * carries `ok` plus either `before_tokens`/`after_tokens`/`summary` or the
+ * failure `reason`/`message`. The successful end's summary also replays from
+ * `/history` as a `role: "system"`, `source: "compaction"` entry.
+ */
+export interface CompactionEvent {
+  type: "compaction";
+  phase: "start" | "progress" | "end";
+  /** Correlates every event of one compaction run (0 from legacy emitters). */
+  compaction_id?: number;
+  /**
+   * End events only: true marks a force-abandoned compaction retiring
+   * after a successor generation took over — skip failure notices for
+   * those (an OK end's result still stands; the history swap happened).
+   */
+  superseded?: boolean;
+  /**
+   * Failed ends only: the emitter-computed display verdict — show
+   * `message` only when true, instead of re-deriving suppression from
+   * reason/trigger/superseded client-side.
+   */
+  notice?: boolean;
+  /** Present on start and on every end (ok or failed). */
+  trigger?: "manual" | "auto";
+  where?: string;
+  pct?: number;
+  part?: number;
+  total?: number;
+  depth?: number;
+  retry_in?: number;
+  error?: string;
+  warning?: string;
+  ok?: boolean;
+  reason?: string;
+  message?: string;
+  before_tokens?: number;
+  after_tokens?: number;
+  summary?: string;
+}
+
 // Global events

 export interface WsStateEvent {
@@ -167,6 +251,8 @@ export interface WsStateEvent {
  context_ratio: number;
  activity: string;
  activity_state: string;
+  /** Defaults to `healthy` when omitted by an older node. */
+  persistence_state?: ConversationPersistenceState;
  /** Full assistant response text — populated on idle transitions only. */
  content?: string;
 }
@@ -194,6 +280,8 @@ export interface WsClosedEvent {
 export type ServerEvent =
  | ConnectedEvent
  | HistoryEvent
+  | HistoryResyncEvent
+  | UserTurnEvent
  | ThinkingStartEvent
  | ThinkingStopEvent
  | ContentEvent
@@ -212,6 +300,7 @@ export type ServerEvent =
  | BusyErrorEvent
  | ClearUiEvent
  | CancelledEvent
+  | CompactionEvent
  | WsStateEvent
  | WsActivityEvent
  | WsRenameEvent
@@ -240,6 +329,8 @@ export interface ClusterStateEvent {
  context_ratio: number;
  activity: string;
  activity_state: string;
+  /** Defaults to `healthy` when omitted by an older node. */
+  persistence_state?: ConversationPersistenceState;
 }

 export interface ClusterWsCreatedEvent {
@@ -247,6 +338,8 @@ export interface ClusterWsCreatedEvent {
  ws_id: string;
  node_id: string;
  name: string;
+  /** Defaults to `healthy` when omitted by an older node. */
+  persistence_state?: ConversationPersistenceState;
 }

 export interface ClusterWsClosedEvent {
@@ -330,3 +423,13 @@ export function isApprovalResolvedEvent(
 export function isCancelledEvent(e: ServerEvent): e is CancelledEvent {
  return e.type === "cancelled";
 }
+
+export function isHistoryResyncEvent(
+  e: ServerEvent,
+): e is HistoryResyncEvent {
+  return e.type === "history_resync";
+}
+
+export function isUserTurnEvent(e: ServerEvent): e is UserTurnEvent {
+  return e.type === "user_turn";
+}
@@ -30,6 +30,8 @@ export type {
  ClusterEvent,
  ConnectedEvent,
  HistoryEvent,
+  HistoryResyncEvent,
+  UserTurnEvent,
  ThinkingStartEvent,
  ThinkingStopEvent,
  ContentEvent,
@@ -71,19 +73,27 @@ export {
  isApproveRequestEvent,
  isApprovalResolvedEvent,
  isCancelledEvent,
+  isHistoryResyncEvent,
+  isUserTurnEvent,
 } from "./events.js";

 // Request/response types
 export type {
+  ConversationPersistenceState,
  SendRequest,
  SendResponse,
  ApproveRequest,
+  ApproveResponse,
+  CancelRequest,
+  CancelResponse,
  CommandRequest,
  CreateWorkstreamRequest,
  CreateWorkstreamResponse,
  CloseWorkstreamRequest,
  WorkstreamInfo,
  ListWorkstreamsResponse,
+  WorkstreamHistoryResponse,
+  StreamEventsOptions,
  DashboardWorkstream,
  DashboardAggregate,
  DashboardResponse,
@@ -110,6 +120,9 @@ export type {
  NodeDetailResponse,
  ConsoleCreateWsRequest,
  ConsoleCreateWsResponse,
+  RouteCreateRequest,
+  RouteCreateResponse,
+  RouteLiveResponse,
  ConsoleHealthResponse,
  CreateScheduleRequest,
  UpdateScheduleRequest,
@@ -3,9 +3,11 @@ import type { ServerEvent } from "./events.js";
 import type {
  AttachmentContent,
  AttachmentUpload,
+  ApproveResponse,
  AuthLoginResponse,
  AuthSetupResponse,
  AuthStatusResponse,
+  CancelResponse,
  CreateWorkstreamRequest,
  CreateWorkstreamResponse,
  DashboardResponse,
@@ -23,8 +25,10 @@ import type {
  SendResponse,
  SkillSummary,
  StatusResponse,
+  StreamEventsOptions,
  TurnResult,
  UploadAttachmentResponse,
+  WorkstreamHistoryResponse,
 } from "./types.js";

 function generateWsId(): string {
@@ -111,12 +115,15 @@ export class TurnstoneServer extends BaseClient {
  async send(
    message: string,
    wsId: string,
-    opts?: { attachmentIds?: string[] },
+    opts?: { attachmentIds?: string[]; clientSendId?: string },
  ): Promise<SendResponse> {
    const body: Record<string, unknown> = { message };
    if (opts?.attachmentIds !== undefined) {
      body.attachment_ids = opts.attachmentIds;
    }
+    if (opts?.clientSendId !== undefined) {
+      body.client_send_id = opts.clientSendId;
+    }
    return this.request(
      "POST",
      `/v1/api/workstreams/${encodeURIComponent(wsId)}/send`,
@@ -173,7 +180,7 @@ export class TurnstoneServer extends BaseClient {
    cycleId?: string;
    /** Alternative selector: any call_id inside the target cycle. */
    callId?: string;
-  }): Promise<StatusResponse> {
+  }): Promise<ApproveResponse> {
    return this.request(
      "POST",
      `/v1/api/workstreams/${encodeURIComponent(opts.wsId)}/approve`,
@@ -201,7 +208,7 @@ export class TurnstoneServer extends BaseClient {
  async cancel(
    wsId: string,
    opts?: { force?: boolean },
-  ): Promise<StatusResponse> {
+  ): Promise<CancelResponse> {
    const body: Record<string, unknown> = {};
    if (opts?.force) body.force = true;
    return this.request(
@@ -229,11 +236,45 @@ export class TurnstoneServer extends BaseClient {
    );
  }

+  // -- History ---------------------------------------------------------------
+
+  /**
+   * Return the requested tail of the authoritative total accepted row prefix.
+   * A 503 is non-authoritative and must not replace an existing transcript.
+   */
+  async getHistory(
+    wsId: string,
+    opts?: { limit?: number },
+  ): Promise<WorkstreamHistoryResponse> {
+    return this.request(
+      "GET",
+      `/v1/api/workstreams/${encodeURIComponent(wsId)}/history`,
+      { params: { limit: opts?.limit ?? 100 } },
+    );
+  }
+
  // -- Streaming ------------------------------------------------------------

-  async *streamEvents(wsId: string): AsyncIterableIterator<ServerEvent> {
+  /**
+   * Open one caller-managed event stream. Pass history hints only after fully
+   * rendering the corresponding `getHistory()` response. On `history_resync`,
+   * stop this iterator, refetch and render history, then open a new stream with
+   * the new hints. No automatic reconnect or transcript repair is performed.
+   */
+  async *streamEvents(
+    wsId: string,
+    opts?: StreamEventsOptions,
+  ): AsyncIterableIterator<ServerEvent> {
+    const params: Record<string, string | number> = { user_turn: 1 };
+    if (opts?.lastEventId !== undefined) {
+      params.last_event_id = opts.lastEventId;
+    }
+    if (opts?.historyToken) {
+      params.history_token = opts.historyToken;
+    }
    yield* this.streamSSE<ServerEvent>(
      `/v1/api/workstreams/${encodeURIComponent(wsId)}/events`,
+      params,
    );
  }

@@ -275,7 +316,7 @@ export class TurnstoneServer extends BaseClient {
      // Start consuming the per-workstream SSE stream first
      const events = this.streamSSE<ServerEvent>(
        `/v1/api/workstreams/${encodeURIComponent(wsId)}/events`,
-        undefined,
+        { user_turn: 1 },
        controller.signal,
      );

@@ -353,7 +394,14 @@ export class TurnstoneServer extends BaseClient {
  }

  async saveMemory(opts: SaveMemoryRequest): Promise<MemoryInfo> {
-    return this.request("POST", "/v1/api/memories", { json: opts });
+    if (typeof opts.description !== "string" || !opts.description.trim()) {
+      throw new TypeError(
+        "memory description is required and must be non-empty",
+      );
+    }
+    return this.request("POST", "/v1/api/memories", {
+      json: { ...opts, description: opts.description.trim() },
+    });
  }

  async searchMemories(
@@ -2,6 +2,13 @@
 // Shared types
 // ---------------------------------------------------------------------------

+/** Sanitized operator-visible state of accepted conversation persistence. */
+export type ConversationPersistenceState =
+  | "healthy"
+  | "pending"
+  | "retrying"
+  | "conflict";
+
 export interface ErrorResponse {
  error: string;
 }
@@ -56,6 +63,11 @@ export interface SendRequest {
   * workstream are auto-consumed; an empty list disables auto-consume.
   */
  attachment_ids?: string[];
+  /**
+   * Opaque optimistic-send correlation echoed by user_turn/history.
+   * Reusing it does not collapse or deduplicate accepted turns.
+   */
+  client_send_id?: string;
 }

 export interface SendResponse {
@@ -116,7 +128,26 @@ export interface ApproveRequest {
  approved: boolean;
  feedback?: string | null;
  always?: boolean;
-  ws_id: string;
+  /** Resolve exactly this approval cycle. */
+  cycle_id?: string | null;
+  /** Resolve the approval cycle containing this tool call. */
+  call_id?: string | null;
+}
+
+export interface ApproveResponse {
+  status: string;
+  /** The cycle resolved by the request, or null when none was pending. */
+  cycle_id: string | null;
+}
+
+export interface CancelRequest {
+  force?: boolean;
+}
+
+export interface CancelResponse {
+  status: string;
+  /** Credential-redacted snapshot of pending work affected by cancellation. */
+  dropped: Record<string, unknown>;
 }

 export interface CommandRequest {
@@ -128,7 +159,20 @@ export interface CreateWorkstreamRequest {
  name?: string;
  model?: string;
  auto_approve?: boolean;
+  /** Tool names accepted as a CSV string or array; blanks are removed server-side. */
+  auto_approve_tools?: string | string[];
+  /** Override judge model alias for this workstream. */
+  judge_model?: string;
+  /**
+   * Owner override for trusted service identities. Ordinary callers remain
+   * bound to their authenticated principal.
+   */
+  user_id?: string;
  resume_ws?: string;
+  /** Completion-notification targets as JSON text or structured target objects. */
+  notify_targets?: string | Array<Record<string, string>>;
+  /** Client surface label such as web, cli, chat, or scheduled. */
+  client_type?: string;
  skill?: string;
  /**
   * Persona name (slug) to create the workstream with. Resolved and
@@ -193,6 +237,8 @@ export interface WorkstreamInfo {
  parent_ws_id: string | null;
  user_id: string;
  project_id: string | null;
+  /** Defaults to `healthy` when omitted by an older node. */
+  persistence_state?: ConversationPersistenceState;
 }

 export interface ListWorkstreamsResponse {
@@ -208,14 +254,33 @@ export interface WorkstreamDetailResponse {
  state: string;
  user_id: string;
  kind: string;
+  /** Defaults to `healthy` when omitted by an older node. */
+  persistence_state?: ConversationPersistenceState;
 }

 export interface WorkstreamHistoryResponse {
  ws_id: string;
-  // Tail of the workstream's reconstructed message history
-  // (provider-fidelity OpenAI-like shape). Bounded by the ?limit=
-  // query param (default 100, max 500).
+  /**
+   * Requested limit-bounded tail of the authoritative total accepted
+   * conversation-row prefix.
+   * Roles include user, assistant, tool, and system; projected compaction and
+   * cancellation markers participate in the same prefix.
+   */
  messages: Record<string, unknown>[];
+  /** Initial event-ring cursor returned by the history projection, if needed. */
+  cursor: number | null;
+  /**
+   * Opaque one-shot token naming the exact live prefix used for this render.
+   * Null for a workstream that is not currently loaded.
+   */
+  handoff_token: string | null;
+}
+
+export interface StreamEventsOptions {
+  /** Initial event-ring cursor, normally copied from `getHistory()`. */
+  lastEventId?: number;
+  /** One-shot live-prefix token, copied only from the history just rendered. */
+  historyToken?: string;
 }

 export interface DashboardWorkstream {
@@ -232,6 +297,8 @@ export interface DashboardWorkstream {
  node?: string;
  model?: string;
  model_alias?: string;
+  /** Defaults to `healthy` when omitted by an older node. */
+  persistence_state?: ConversationPersistenceState;
 }

 export interface DashboardAggregate {
@@ -498,6 +565,8 @@ export interface ClusterWorkstreamInfo {
  activity?: string;
  activity_state?: string;
  tool_calls?: number;
+  /** Defaults to `healthy` when omitted by an older node. */
+  persistence_state?: ConversationPersistenceState;
 }

 export interface ClusterWorkstreamsResponse {
@@ -541,7 +610,11 @@ export interface ConsoleCreateWsRequest {
  skill?: string;
  /** Persona slug — resolved and snapshotted at creation. */
  persona?: string;
+  /** Project to attach the workstream to. */
+  project_id?: string;
  resume_ws?: string;
+  /** Override judge model alias for this workstream. */
+  judge_model?: string;
 }

 export interface ConsoleCreateWsResponse {
@@ -550,6 +623,22 @@ export interface ConsoleCreateWsResponse {
  target_node: string;
 }

+export interface RouteCreateRequest extends CreateWorkstreamRequest {
+  /** Pin placement to this node by generating a matching rendezvous key. */
+  target_node?: string;
+}
+
+export interface RouteCreateResponse extends CreateWorkstreamResponse {
+  node_url: string;
+  node_id: string;
+  routing_strategy: "rendezvous" | "target_node" | "resume";
+}
+
+export interface RouteLiveResponse {
+  ws_id: string;
+  live: boolean;
+}
+
 export interface ConsoleHealthResponse {
  status: string;
  service: string;
@@ -805,7 +894,7 @@ export interface WorkstreamsOptions {
 export interface SaveMemoryRequest {
  name: string;
  content: string;
-  description?: string;
+  description: string;
  type?: "user" | "general" | "feedback" | "reference";
  scope?: "global" | "workstream" | "user";
  scope_id?: string;
@@ -62,6 +62,58 @@ describe("TurnstoneConsole", () => {
    expect(url).toContain("page=2");
  });

+  it("createWorkstream sends the live cluster-create contract", async () => {
+    const fetchFn = mockFetch({
+      status: "ok",
+      correlation_id: "ws-new",
+      target_node: "node-a",
+    });
+    const client = new TurnstoneConsole({
+      baseUrl: "http://test",
+      fetch: fetchFn,
+    });
+    await client.createWorkstream({
+      node_id: "node-a",
+      project_id: "project-42",
+      judge_model: "judge-fast",
+    });
+
+    const [, init] = (fetchFn as ReturnType<typeof vi.fn>).mock.calls[0];
+    expect(JSON.parse(init.body)).toEqual({
+      node_id: "node-a",
+      project_id: "project-42",
+      judge_model: "judge-fast",
+    });
+  });
+
+  it("routeCreateWorkstream returns placement metadata", async () => {
+    const fetchFn = mockFetch({
+      ws_id: "ws-new",
+      name: "routed",
+      node_url: "http://node-a:8080",
+      node_id: "node-a",
+      routing_strategy: "target_node",
+    });
+    const client = new TurnstoneConsole({
+      baseUrl: "http://test",
+      fetch: fetchFn,
+    });
+    const response = await client.routeCreateWorkstream({
+      name: "routed",
+      target_node: "node-a",
+      client_type: "scheduled",
+      notify_targets: [{ channel_type: "slack", channel_id: "C123" }],
+    });
+
+    expect(response.node_id).toBe("node-a");
+    expect(response.routing_strategy).toBe("target_node");
+    const [, init] = (fetchFn as ReturnType<typeof vi.fn>).mock.calls[0];
+    expect(JSON.parse(init.body)).toMatchObject({
+      client_type: "scheduled",
+      notify_targets: [{ channel_type: "slack", channel_id: "C123" }],
+    });
+  });
+
  it("routeCreateWorkstream rejects attachments + target_node", async () => {
    const fetchFn = vi.fn().mockResolvedValue(
      new Response("{}", {
@@ -84,6 +136,21 @@ describe("TurnstoneConsole", () => {
    expect(fetchFn).not.toHaveBeenCalled();
  });

+  it("routeWorkstreamLive returns the non-mutating liveness probe", async () => {
+    const fetchFn = mockFetch({ ws_id: "saved/ws", live: true });
+    const client = new TurnstoneConsole({
+      baseUrl: "http://test",
+      fetch: fetchFn,
+    });
+
+    const response = await client.routeWorkstreamLive("saved/ws");
+
+    expect(response).toEqual({ ws_id: "saved/ws", live: true });
+    const [url, init] = (fetchFn as ReturnType<typeof vi.fn>).mock.calls[0];
+    expect(url).toContain("/v1/api/route/workstreams/saved%2Fws/live");
+    expect(init.method).toBe("GET");
+  });
+
  it("health returns parsed response", async () => {
    const fetchFn = mockFetch({
      status: "ok",
@@ -8,6 +8,8 @@ import {
  isApproveRequestEvent,
  isApprovalResolvedEvent,
  isReasoningEvent,
+  isHistoryResyncEvent,
+  isUserTurnEvent,
 } from "../src/events.js";
 import type { ServerEvent } from "../src/events.js";

@@ -44,6 +46,26 @@ describe("event type guards", () => {
    expect(isToolResultEvent(e)).toBe(true);
  });

+  it("carries accepted tool projection metadata", () => {
+    const e: ServerEvent = {
+      type: "tool_result",
+      call_id: "c-final",
+      name: "open_preview",
+      output: "guarded\nscalar",
+      is_error: true,
+      preview: { kind: "html", attachment_id: "preview-1" },
+      accepted: true,
+      effect_status: "unknown",
+      _event_id: 42,
+    };
+    expect(isToolResultEvent(e)).toBe(true);
+    if (!isToolResultEvent(e)) throw new Error("tool result type guard failed");
+    expect(e.accepted).toBe(true);
+    expect(e.preview).toEqual({ kind: "html", attachment_id: "preview-1" });
+    expect(e.effect_status).toBe("unknown");
+    expect(e._event_id).toBe(42);
+  });
+
  it("isWsStateEvent", () => {
    const e: ServerEvent = {
      type: "ws_state",
@@ -53,6 +75,7 @@ describe("event type guards", () => {
      context_ratio: 0,
      activity: "",
      activity_state: "",
+      persistence_state: "retrying",
    };
    expect(isWsStateEvent(e)).toBe(true);
  });
@@ -70,4 +93,26 @@ describe("event type guards", () => {
    };
    expect(isApprovalResolvedEvent(e)).toBe(true);
  });
+
+  it("isHistoryResyncEvent", () => {
+    const e: ServerEvent = {
+      type: "history_resync",
+      ws_id: "ws1",
+      reason: "handoff_mismatch",
+    };
+    expect(isHistoryResyncEvent(e)).toBe(true);
+    expect(isContentEvent(e)).toBe(false);
+  });
+
+  it("isUserTurnEvent", () => {
+    const e: ServerEvent = {
+      type: "user_turn",
+      content: "hello",
+      sender: "user-1",
+      client_send_ids: ["browser-send"],
+      _event_id: 17,
+    };
+    expect(isUserTurnEvent(e)).toBe(true);
+    expect(isContentEvent(e)).toBe(false);
+  });
 });
@@ -58,11 +58,58 @@ describe("TurnstoneServer", () => {
      baseUrl: "http://test",
      fetch: fetchFn,
    });
-    const resp = await client.createWorkstream({ name: "Analysis" });
+    const resp = await client.createWorkstream({
+      name: "Analysis",
+      judge_model: "judge-fast",
+      client_type: "scheduled",
+      notify_targets: [{ channel_type: "slack", channel_id: "C123" }],
+    });
    expect(resp.ws_id).toBe("ws_new");

    const [, init] = (fetchFn as ReturnType<typeof vi.fn>).mock.calls[0];
-    expect(JSON.parse(init.body)).toEqual({ name: "Analysis" });
+    expect(JSON.parse(init.body)).toEqual({
+      name: "Analysis",
+      judge_model: "judge-fast",
+      client_type: "scheduled",
+      notify_targets: [{ channel_type: "slack", channel_id: "C123" }],
+    });
+  });
+
+  it("saveMemory requires and normalizes the description", async () => {
+    const fetchFn = mockFetch({
+      memory_id: "m1",
+      name: "deployment_process",
+      description: "Production deployment workflow",
+      type: "general",
+      scope: "global",
+      scope_id: "",
+      content: "Deploy from main",
+      created: "2026-08-11T00:00:00",
+      updated: "2026-08-11T00:00:00",
+    });
+    const client = new TurnstoneServer({
+      baseUrl: "http://test",
+      fetch: fetchFn,
+    });
+
+    await client.saveMemory({
+      name: "deployment_process",
+      content: "Deploy from main",
+      description: "  Production deployment workflow  ",
+    });
+    const [, init] = (fetchFn as ReturnType<typeof vi.fn>).mock.calls[0];
+    expect(JSON.parse(init.body)).toMatchObject({
+      description: "Production deployment workflow",
+    });
+
+    await expect(
+      client.saveMemory({
+        name: "deployment_process",
+        content: "Deploy from main",
+        description: "   ",
+      }),
+    ).rejects.toThrow("description is required");
+    expect(fetchFn).toHaveBeenCalledTimes(1);
  });

  it("send posts correct payload", async () => {
@@ -78,6 +125,111 @@ describe("TurnstoneServer", () => {
    expect(JSON.parse(init.body)).toEqual({ message: "Hello" });
  });

+  it("send threads the optional browser correlation token", async () => {
+    const fetchFn = mockFetch({ status: "ok" });
+    const client = new TurnstoneServer({
+      baseUrl: "http://test",
+      fetch: fetchFn,
+    });
+    await client.send("Hello", "ws1", { clientSendId: "browser-send_1" });
+
+    const [, init] = (fetchFn as ReturnType<typeof vi.fn>).mock.calls[0];
+    expect(JSON.parse(init.body)).toEqual({
+      message: "Hello",
+      client_send_id: "browser-send_1",
+    });
+  });
+
+  it("approve selects a cycle without duplicating ws_id in the body", async () => {
+    const fetchFn = mockFetch({ status: "ok", cycle_id: "cycle-1" });
+    const client = new TurnstoneServer({
+      baseUrl: "http://test",
+      fetch: fetchFn,
+    });
+    const response = await client.approve({
+      wsId: "ws1",
+      approved: false,
+      cycleId: "cycle-1",
+      callId: "call-1",
+    });
+
+    const [url, init] = (fetchFn as ReturnType<typeof vi.fn>).mock.calls[0];
+    expect(url).toBe("http://test/v1/api/workstreams/ws1/approve");
+    expect(JSON.parse(init.body)).toEqual({
+      approved: false,
+      cycle_id: "cycle-1",
+      call_id: "call-1",
+    });
+    expect(response.cycle_id).toBe("cycle-1");
+  });
+
+  it("cancel preserves the dropped-work snapshot", async () => {
+    const fetchFn = mockFetch({
+      status: "cancelled",
+      dropped: { tool_calls: ["call-1"] },
+    });
+    const client = new TurnstoneServer({
+      baseUrl: "http://test",
+      fetch: fetchFn,
+    });
+    const response = await client.cancel("ws1", { force: true });
+
+    const [, init] = (fetchFn as ReturnType<typeof vi.fn>).mock.calls[0];
+    expect(JSON.parse(init.body)).toEqual({ force: true });
+    expect(response.dropped).toEqual({ tool_calls: ["call-1"] });
+  });
+
+  it("getHistory returns the cursor and one-shot handoff token", async () => {
+    const fetchFn = mockFetch({
+      ws_id: "ws1",
+      messages: [{ role: "system", source: "compaction", content: "summary" }],
+      cursor: 0,
+      handoff_token: "epoch.7",
+    });
+    const client = new TurnstoneServer({
+      baseUrl: "http://test",
+      fetch: fetchFn,
+    });
+
+    const history = await client.getHistory("ws1", { limit: 42 });
+
+    expect(history.cursor).toBe(0);
+    expect(history.handoff_token).toBe("epoch.7");
+    const [url] = (fetchFn as ReturnType<typeof vi.fn>).mock.calls[0];
+    expect(url).toBe("http://test/v1/api/workstreams/ws1/history?limit=42");
+  });
+
+  it("streamEvents forwards caller-managed initial history hints", async () => {
+    const fetchFn = vi
+      .fn()
+      .mockResolvedValue(
+        new Response(
+          'data: {"type":"history_resync","ws_id":"ws1","reason":"handoff_mismatch"}\n\n',
+          { status: 200, headers: { "content-type": "text/event-stream" } },
+        ),
+      );
+    const client = new TurnstoneServer({
+      baseUrl: "http://test",
+      fetch: fetchFn,
+    });
+
+    const events = [];
+    for await (const event of client.streamEvents("ws1", {
+      lastEventId: 0,
+      historyToken: "epoch.7",
+    })) {
+      events.push(event);
+    }
+
+    expect(events).toEqual([
+      { type: "history_resync", ws_id: "ws1", reason: "handoff_mismatch" },
+    ]);
+    const [url] = (fetchFn as ReturnType<typeof vi.fn>).mock.calls[0];
+    expect(url).toBe(
+      "http://test/v1/api/workstreams/ws1/events?user_turn=1&last_event_id=0&history_token=epoch.7",
+    );
+  });
+
  it("injects auth header when token provided", async () => {
    const fetchFn = mockFetch({ workstreams: [] });
    const client = new TurnstoneServer({
@@ -21,6 +21,8 @@ from turnstone.console.collector import ClusterCollector
 from turnstone.console.coordinator_adapter import CoordinatorAdapter
 from turnstone.console.coordinator_ui import ConsoleCoordinatorUI
 from turnstone.core.auth import AuthResult
+from turnstone.core.model_registry import ModelConfig
+from turnstone.core.providers import ModelCapabilities
 from turnstone.core.session_manager import SessionManager

 if TYPE_CHECKING:
@@ -72,9 +74,21 @@ class _FakeConfigStore:


 def _fake_registry() -> MagicMock:
-    """MagicMock whose ``.resolve()`` succeeds so the 503 gate passes."""
+    """MagicMock whose legacy and atomic binding resolutions both succeed."""
+    client = MagicMock()
+    cfg = ModelConfig(
+        alias="default",
+        base_url="https://example.invalid/v1",
+        api_key="test",
+        model="gpt-4",
+    )
+    provider = MagicMock()
+    provider.provider_name = "openai"
+    provider.get_capabilities.return_value = ModelCapabilities()
    reg = MagicMock()
-    reg.resolve.return_value = (MagicMock(), "gpt-4", MagicMock())
+    reg.default = "default"
+    reg.resolve.return_value = (client, cfg.model, cfg, 0)
+    reg.resolve_binding.return_value = (client, cfg.model, cfg, provider, 0)
    return reg


@@ -71,6 +71,9 @@ def patch_session_storage(
    calls: list[str] = []

    class _Stub:
+        def get_workstream(self, ws_id: str) -> None:
+            return None
+
        def is_watch_active(self, watch_id: str) -> bool:
            calls.append(watch_id)
            if raise_on_is_active:
@@ -0,0 +1,180 @@
+"""Shared helpers for the Python-driven node harnesses that evaluate the
+``shared_static`` ES modules with script semantics."""
+
+from __future__ import annotations
+
+import re
+import shutil
+from typing import TYPE_CHECKING
+
+import pytest
+
+if TYPE_CHECKING:
+    from pathlib import Path
+
+
+def has_node() -> bool:
+    return shutil.which("node") is not None
+
+
+# Module-level ``pytestmark = node_skip`` in each harness suite — the node
+# detection lives here once, so a future change (version floor, env
+# override) cannot land in one suite and silently miss another.
+node_skip = pytest.mark.skipif(not has_node(), reason="node not available")
+
+
+def demodulize(path: Path) -> str:
+    """Strip ES-module syntax so ``vm.runInThisContext`` (script semantics)
+    can evaluate the file: imports drop (the harness loads the whole
+    dependency set into one shared context, so cross-file bindings resolve
+    as context globals, exactly like the pre-module classic scripts), and
+    ``export`` keywords peel off their declarations.
+
+    Single-sourced here for every JS harness: a new module syntax form
+    (``export default``, re-exports) must be handled once, not per suite —
+    a divergence between per-file copies surfaces as a confusing
+    ``vm.runInThisContext`` SyntaxError in whichever suite lagged.
+    """
+    src = path.read_text(encoding="utf-8")
+    src = re.sub(r"^import\s+\{[\s\S]*?\}\s+from\s+\"[^\"]+\";\s*$", "", src, flags=re.M)
+    src = re.sub(r"^import\s+[^;\n]+;\s*$", "", src, flags=re.M)
+    src = re.sub(
+        r"^export\s+(?=(?:async\s+)?(?:function|const|let|var|class)\b)", "", src, flags=re.M
+    )
+    src = re.sub(r"^export\s*\{[^}]*\};\s*$", "", src, flags=re.M)
+    return src
+
+
+def slice_braced_block(source: str, anchor: int) -> str | None:
+    """Slice the ``{ … }`` block starting at/just after ``anchor``.
+
+    THE brace walker every JS harness suite shares (the comment-AND-
+    string-aware superset of the per-suite predecessors, which disagreed
+    on comment handling and window bounds — the same source
+    reorganization could pass one suite's structural pin while breaking
+    the other's with a slice-dependent failure).  Comment awareness makes
+    it correct on raw AND pre-stripped input alike.  Returns ``None``
+    when no ``{`` opens within 200 chars of ``anchor`` (a missing brace
+    must not silently slice some later unrelated block) or the block is
+    unterminated.
+    """
+    start = source.find("{", anchor)
+    if start == -1 or start - anchor > 200:
+        return None
+    depth = 0
+    quote = ""
+    escaped = False
+    line_comment = False
+    block_comment = False
+    i = start
+    while i < len(source):
+        ch = source[i]
+        nxt = source[i + 1] if i + 1 < len(source) else ""
+        if line_comment:
+            if ch == "\n":
+                line_comment = False
+        elif block_comment:
+            if ch == "*" and nxt == "/":
+                block_comment = False
+                i += 1
+        elif quote:
+            if escaped:
+                escaped = False
+            elif ch == "\\":
+                escaped = True
+            elif ch == quote:
+                quote = ""
+        elif ch == "/" and nxt == "/":
+            line_comment = True
+            i += 1
+        elif ch == "/" and nxt == "*":
+            block_comment = True
+            i += 1
+        elif ch in {'"', "'", "`"}:
+            quote = ch
+        elif ch == "{":
+            depth += 1
+        elif ch == "}":
+            depth -= 1
+            if depth == 0:
+                return source[start : i + 1]
+        i += 1
+    return None
+
+
+def extract_braced(source: str, signature: str) -> str:
+    """Extract one JS function/method (signature included) — raising form.
+
+    ``signature`` must end at its opening ``{``.  The loud sibling of
+    :func:`slice_braced_block` for suites that treat a missing or
+    unterminated function as a hard failure rather than a skip.
+    """
+    start = source.index(signature)
+    brace = start + len(signature) - 1
+    if source[brace] != "{":
+        raise AssertionError(f"signature does not end at an opening brace: {signature}")
+    block = slice_braced_block(source, brace)
+    if block is None:
+        raise AssertionError(f"unterminated JavaScript function: {signature}")
+    return source[start:brace] + block
+
+
+def strip_js_comments(source: str) -> str:
+    """Strip ``//`` and ``/* */`` comments for source-pattern assertions —
+    the single implementation every JS harness suite shares.
+
+    STRING-AWARE and OFFSET-PRESERVING (comments become spaces, byte
+    length identical): a ``//`` inside a string literal (``"https://…"``)
+    is content, not a comment — a string-blind scanner truncates the rest
+    of the line, and pattern pins then silently assert against corrupted
+    text (a ``not in`` guard passes vacuously after the pattern it
+    polices was reintroduced).  Length preservation keeps downstream
+    offset math (brace walkers, ``.index`` comparisons) valid.  This is
+    the strict superset of every per-suite predecessor, hoisted so the
+    suites cannot diverge again.
+
+    Limitation — regex literals (``/pattern/flags``) are not detected: a
+    ``//`` inside one would be misread as a line comment.  Safe for
+    every region currently scanned; extend the tracker before scanning a
+    region with regex literals.
+    """
+    out: list[str] = []
+    n = len(source)
+    i = 0
+    in_str: str | None = None
+    while i < n:
+        ch = source[i]
+        if in_str:
+            out.append(ch)
+            if ch == "\\" and i + 1 < n:
+                out.append(source[i + 1])
+                i += 2
+                continue
+            if ch == in_str:
+                in_str = None
+            i += 1
+            continue
+        # Line comment: replace with spaces up to newline (preserve
+        # length so downstream offset math still works).
+        if ch == "/" and i + 1 < n and source[i + 1] == "/":
+            j = source.find("\n", i)
+            if j == -1:
+                j = n
+            out.append(" " * (j - i))
+            i = j
+            continue
+        # Block comment: replace with spaces up to closing */.
+        if ch == "/" and i + 1 < n and source[i + 1] == "*":
+            j = source.find("*/", i + 2)
+            if j == -1:
+                out.append(" " * (n - i))
+                i = n
+                continue
+            out.append(" " * (j + 2 - i))
+            i = j + 2
+            continue
+        if ch in ('"', "'", "`"):
+            in_str = ch
+        out.append(ch)
+        i += 1
+    return "".join(out)
@@ -0,0 +1,57 @@
+"""Shared OIDC posture builder for the model-auth / OBO test surface.
+
+One construction site for the posture the mint and write-validator suites
+read, built as a REAL (frozen) ``OIDCConfig`` so an override for a field
+the dataclass does not carry raises at the call site. Named with a leading
+underscore so pytest does not collect it.
+"""
+
+from __future__ import annotations
+
+from types import SimpleNamespace
+from typing import TYPE_CHECKING, Any
+
+from turnstone.core.oidc import OIDCConfig
+
+if TYPE_CHECKING:
+    from collections.abc import Iterator
+
+# The issuer / token-endpoint pair the mint suites route their mock
+# transports on.
+ISSUER = "https://idp.test"
+TOKEN_ENDPOINT = "https://idp.test/token"
+
+
+def make_oidc_config(**overrides: Any) -> OIDCConfig:
+    """A full, mintable OIDC posture; tests override the field under test,
+    everything else rides the dataclass defaults."""
+    defaults: dict[str, Any] = {
+        "enabled": True,
+        "issuer": ISSUER,
+        "client_id": "cid",
+        "client_secret": "csecret",
+        "token_endpoint": TOKEN_ENDPOINT,
+    }
+    defaults.update(overrides)
+    return OIDCConfig(**defaults)
+
+
+def keyed_app_state() -> SimpleNamespace:
+    """App-state stub satisfying ``ModelRegistry.reload``'s dynamic-auth key
+    guard, for suites exercising reload mechanics rather than key policy."""
+    return SimpleNamespace(mcp_token_store=object())
+
+
+def mint_warn_state_reset() -> Iterator[None]:
+    """Reset generator behind the mint suites' autouse fixtures: empties the
+    process-global mint warn/dedup/cause state before AND after each test,
+    so warn-dedup assertions are not order-dependent. Modules install it as
+    ``yield from mint_warn_state_reset()`` in an autouse fixture.
+    """
+    # Lazy import: non-mint consumers of this helper module (the write-
+    # validator suites) shouldn't pay the mcp_oauth import.
+    from turnstone.core.mcp_oauth import reset_model_mint_warn_state_for_tests
+
+    reset_model_mint_warn_state_for_tests()
+    yield
+    reset_model_mint_warn_state_for_tests()
@@ -0,0 +1,235 @@
+"""#832 replay-parity harness: scenario table + runner.
+
+The audit is controller determinism: with the plant's chunk sequence held
+fixed, the streaming phase must produce an identical UI event sequence
+and an identical committed message — modulo the RULED behavior changes
+restated in full on the transforms in ``test_832_parity.py``.  This
+module is the shared half: the scenario scripts (one row per
+chunk-field→UI translation the consumer performs) and the runner that
+drives one through the streaming seam, recording everything the turn
+observably produced.
+
+Baselines are captured from the PRE-FOLD path (``UPDATE_832_PARITY=1``,
+run at a tree where ``session.py`` is byte-identical to pre-fold main)
+into ``tests/data/parity_832/``.  The runner adapts to EITHER world by
+signature, so a recapture at an old tree records real old-world
+behavior, and capture mode refuses to write a record whose failure is
+the harness's own call shape.  Assert mode replays the same scripts
+through the current tree and compares against the baseline, applying the
+ruled transforms; a mismatch outside a ruled transform is a regression.
+
+The provider fake arms ``cancel_ref`` EAGERLY (a closeable sentinel
+appended inside ``create_streaming``, before the iterator is returned),
+mirroring every real adapter: the wrapper classifies
+creation-vs-midstream failures by that arming, so a fake that skipped it
+would exercise only the creation arm.
+"""
+
+from __future__ import annotations
+
+import inspect
+import json
+import os
+import re
+from pathlib import Path
+from typing import Any
+
+from tests._session_helpers import (
+    RecordingUI,
+    make_session,
+    replace_session_lane,
+    scripted_provider,
+)
+from turnstone.core.providers._protocol import StreamChunk, ToolCallDelta, UsageInfo
+from turnstone.core.trajectory import Turn
+
+FIXTURE_DIR = Path(__file__).parent / "data" / "parity_832"
+UPDATE = os.environ.get("UPDATE_832_PARITY") == "1"
+
+
+def _tc(index: int, call_id: str, name: str = "", args: str = "") -> ToolCallDelta:
+    return ToolCallDelta(index=index, id=call_id, name=name, arguments_delta=args)
+
+
+_USAGE_A = UsageInfo(prompt_tokens=11, completion_tokens=0, total_tokens=11)
+_USAGE_B = UsageInfo(prompt_tokens=11, completion_tokens=7, total_tokens=18)
+
+# Scenario table — the V11 grid, one script per row.  Scripts are chunk
+# LISTS; the runner re-iterates a fresh iterator per attempt.
+SCENARIOS: dict[str, list[StreamChunk]] = {
+    "content_only": [
+        StreamChunk(content_delta="Hello "),
+        StreamChunk(content_delta="world."),
+        StreamChunk(finish_reason="stop", usage=_USAGE_B),
+    ],
+    "reasoning_then_content": [
+        StreamChunk(reasoning_delta="think a", usage=_USAGE_A),
+        StreamChunk(reasoning_delta=" think b"),
+        StreamChunk(content_delta="Answer."),
+        StreamChunk(finish_reason="stop", usage=_USAGE_B),
+    ],
+    "tools_simple": [
+        StreamChunk(content_delta="Calling."),
+        StreamChunk(tool_call_deltas=[_tc(0, "call_1", "get_weather", '{"city": ')]),
+        StreamChunk(tool_call_deltas=[_tc(0, "", "", '"Paris"}')]),
+        StreamChunk(finish_reason="tool_calls", usage=_USAGE_B),
+    ],
+    "combined_content_tools_finish": [
+        StreamChunk(content_delta="Before "),
+        StreamChunk(
+            content_delta="tools",
+            tool_call_deltas=[_tc(0, "call_1", "get_weather", '{"city": "Nice"}')],
+            finish_reason="tool_calls",
+        ),
+        StreamChunk(usage=_USAGE_B),
+    ],
+    "info_prefinish": [
+        StreamChunk(info_delta="[Searching: pinniped taxonomy]"),
+        StreamChunk(content_delta="Seals are pinnipeds."),
+        StreamChunk(finish_reason="stop", usage=_USAGE_B),
+    ],
+    "info_postfinish_footer": [
+        StreamChunk(content_delta="Answer with sources."),
+        StreamChunk(finish_reason="stop", usage=_USAGE_B),
+        StreamChunk(info_delta="Sources:\n- example.com/page"),
+    ],
+    "think_tags_split_across_chunks": [
+        StreamChunk(content_delta="<thi"),
+        StreamChunk(content_delta="nk>plan</think>\n\nAnswer"),
+        StreamChunk(finish_reason="stop", usage=_USAGE_B),
+    ],
+    "blank_id_tools": [
+        StreamChunk(tool_call_deltas=[_tc(0, "", "get_weather", '{"city": "Oslo"}')]),
+        StreamChunk(finish_reason="tool_calls", usage=_USAGE_B),
+    ],
+    "length_with_tools": [
+        StreamChunk(content_delta="Partial answer"),
+        StreamChunk(tool_call_deltas=[_tc(0, "call_1", "get_weather", '{"city": "Par')]),
+        StreamChunk(finish_reason="length", usage=_USAGE_B),
+    ],
+    "content_filter": [
+        StreamChunk(content_delta="Redac"),
+        StreamChunk(finish_reason="content_filter", usage=_USAGE_B),
+    ],
+    "no_finish_clean_exhaust": [
+        StreamChunk(content_delta="Half an ans"),
+        StreamChunk(usage=_USAGE_A),
+    ],
+    "finish_only_no_content": [
+        StreamChunk(finish_reason="stop", usage=_USAGE_B),
+    ],
+    "provider_blocks_on_terminal": [
+        StreamChunk(content_delta="Blocked."),
+        StreamChunk(
+            finish_reason="stop",
+            usage=_USAGE_B,
+            provider_blocks=[{"type": "reasoning_text", "text": "captured"}],
+        ),
+    ],
+}
+
+
+_SYNTH_ID = re.compile(r"^call_[0-9a-f]{32}$")
+
+
+def _mask_synth_ids(record: dict[str, Any]) -> dict[str, Any]:
+    """Replace uuid-backfilled tool-call ids with stable placeholders.
+
+    The blank-id repair mints ``call_<uuid4hex>`` per run — real
+    nondeterminism inside the seam, but not behavior: mask ONLY that exact
+    shape (never a scripted provider id) with an index-stable token so
+    captures compare across runs.  Applied to the committed projection;
+    UI events never carry call ids in this harness.
+    """
+    result = record.get("result")
+    if not result:
+        return record
+    for i, tc in enumerate(result.get("tool_calls") or []):
+        if _SYNTH_ID.match(tc.get("id", "")):
+            tc["id"] = f"synth-id-{i}"
+    for i, block in enumerate(result.get("provider_content") or []):
+        if isinstance(block, dict) and _SYNTH_ID.match(str(block.get("id", ""))):
+            block["id"] = f"synth-id-{i}"
+    return record
+
+
+def run_scenario(name: str) -> dict[str, Any]:
+    """Drive one scenario through the streaming seam; return the record.
+
+    The record is everything the streaming phase observably produced: the
+    ordered UI events, the committed-message projection, the mid-stream
+    usage slot, and the exception class if the seam raised.  Deliberately
+    seam-level at ``_stream_response`` — full ``send()`` scenarios ride
+    the ported ladder suites instead.
+
+    Signature-adaptive so ``UPDATE_832_PARITY=1`` at a PRE-fold tree
+    records real old-world behavior: the pre-fold seam was
+    ``_stream_response(msgs, my_generation) -> dict``, the post-fold one
+    is ``_stream_response(my_generation) -> ModelTurnResult`` (wire
+    prepared inside).  A harness-shape failure must never be recorded as
+    behavior — ``write_fixture`` refuses one.
+    """
+    ui = RecordingUI()
+    session = make_session(ui=ui)
+    # Zero the ladder backoff: a scenario that reaches the mid-stream
+    # re-issue ladder (no_finish_clean_exhaust) must not sleep real
+    # exponential delays in a unit run.  The retry-notice transform in
+    # test_832_parity hardcodes the matching "0s" wording.
+    session._RETRY_BASE_DELAY = 0
+    replace_session_lane(session, provider=scripted_provider(SCENARIOS[name]))
+
+    pre_fold = "msgs" in inspect.signature(type(session)._stream_response).parameters
+    record: dict[str, Any] = {"scenario": name}
+    try:
+        if pre_fold:
+            # Splatted: the pre-fold seam took (msgs, my_generation), and a
+            # literal two-argument call reads as an arity error against the
+            # signature this tree actually has.
+            pre_fold_args: tuple[Any, ...] = ([{"role": "user", "content": "hi"}], 0)
+            msg = session._stream_response(*pre_fold_args)
+            msg.pop("_wire_msgs", None)
+            record["result"] = {
+                "content": msg.get("content", ""),
+                "tool_calls": msg.get("tool_calls"),
+                "provider_content": msg.get("_provider_content"),
+            }
+        else:
+            session.messages.append(Turn.user("hi"))
+            result = session._stream_response(0)
+            record["result"] = {
+                "content": result.content,
+                "tool_calls": result.tool_calls or None,
+                "provider_content": (
+                    [dict(b) for b in result.turn.native.blocks] if result.turn.native else None
+                ),
+            }
+        record["raised"] = None
+    except BaseException as exc:  # noqa: BLE001 — the record IS the observation
+        record["result"] = None
+        record["raised"] = type(exc).__name__
+    record["ui_events"] = [[k, d] for k, d in ui.events]
+    record["last_usage"] = session._last_usage
+    record["cancelled_partial"] = session._cancelled_partial_msg
+    return _mask_synth_ids(record)
+
+
+def fixture_path(name: str) -> Path:
+    return FIXTURE_DIR / f"{name}.json"
+
+
+def load_fixture(name: str) -> dict[str, Any]:
+    return json.loads(fixture_path(name).read_text())
+
+
+def write_fixture(name: str, record: dict[str, Any]) -> None:
+    # A TypeError before ANY UI event is the harness's own call-shape
+    # failure (run_scenario's signature adapter no longer matches this
+    # tree's seam), not old-world behavior — refuse to destroy the
+    # baseline with it.
+    if record.get("raised") == "TypeError" and not record.get("ui_events"):
+        raise AssertionError(
+            f"parity capture for {name!r} died calling the seam (TypeError before "
+            f"any UI event) — fix run_scenario's signature adapter; do not record"
+        )
+    FIXTURE_DIR.mkdir(parents=True, exist_ok=True)
+    fixture_path(name).write_text(json.dumps(record, indent=2, sort_keys=True) + "\n")
--- a/Show More
+++ b/Show More