4118f31d89
fix(llama-cpp): make endpoint auth transitions reproducible ( #126498 )
2026-08-19 18:37:28 -07:00
0135046830
refactor(llama-cpp): use one provider for managed and existing servers ( #126434 )
...
* refactor(llama-cpp): unify server ownership modes
* test(llama-cpp): preserve shared discovery limits
* fix(plugin-sdk): retain provider auth removal export
2026-08-19 13:57:33 -07:00
7c1779e6e3
refactor(llama-cpp): adopt canonical runtime owners ( #126371 )
2026-08-19 10:17:59 -07:00
c2de3206d4
feat(llama-cpp): support external llama-server
...
* feat(llama-cpp): add external server provider
* feat(llama-cpp): document external server setup
* refactor(llama-cpp): harden external provider boundaries
* fix(llama-cpp): support external structured output
* fix(llama-cpp): isolate replacement endpoint credentials
* test(llama-cpp): register external live shard
* fix(llama-cpp): preserve explicit endpoint authorization
* fix(llama-cpp): clear disabled inline credentials
* fix(llama-cpp): preserve external local service configs
* test(llama-cpp): cover retained external configs
* test(llama-cpp): cover authorization precedence
2026-08-19 17:32:00 +03:00
72c227c7cc
fix(llama): support embedding-only managed servers ( #125383 )
...
* fix(llama): support embedding-only managed servers
* docs(llama): describe embedding-only setup
* fix(llama): remove unused model reference export
* fix(llama-cpp): keep guided setup chat-capable
---------
Co-authored-by: Patrick Erichsen <patrick.a.erichsen@gmail.com >
2026-08-17 23:29:43 -07:00
5f1bbed42d
fix(doctor): report missing managed local embedding setup ( #123575 )
...
* fix(gateway): expose startup blockers before cutover
* fix(gateway): include session blockers in preflight
* fix(gateway): keep preflight finding type private
* fix(gateway): preflight startup auth blockers
* fix(gateway): complete startup preflight readiness
* fix(llama-cpp): keep preflight remediation private
* fix(gateway): keep preflight passive and activation-aware
* fix(gateway): apply startup guard in preflight
* fix(gateway): align auth mode preflight
* fix(gateway): keep preflight state reads isolated
Share the read-only inspection snapshot scope across duplicated runtime chunks so blocked gateway preflight remains non-mutating when bundled provider artifacts read canonical state.
* fix(gateway): keep startup preflight passive
* fix(gateway): ignore inactive embedding owner shadows
* fix(gateway): preserve startup preflight parity
* fix(llama-cpp): keep cache inspection types private
* fix(gateway): close startup preflight parity gaps
* fix(gateway): handle uninitialized memory databases
* test(gateway): observe shell fallback portably
* fix(llama-cpp): normalize embedding model paths
* refactor(gateway): drop broad startup preflight surface
* fix(doctor): report missing managed local embedding setup
* style(memory): simplify setup enablement check
* fix(memory): keep diagnostic result type private
* fix(memory): inspect local setup with remote secret refs
* fix(memory): keep doctor index inspection immutable
* fix(memory): make readiness inspection owner-aware
* fix(doctor): mirror memory slot allowlist policy
* test(doctor): use canonical memory slot id
* fix(doctor): normalize memory provider ids
* fix(doctor): resolve external embedding readiness owner
* fix(plugins): keep embedding inspection result internal
* fix(doctor): isolate plugin state during lint
* fix(doctor): route lint metadata through snapshot
* fix(cli): keep doctor lint startup source-only
* fix(cli): keep doctor lint compile-cache free
* fix(doctor): keep local embedding readiness opt-in
* test(doctor): preserve plugin artifact roots during lint
* fix(doctor): refresh memory readiness registration
* test(doctor): type nullable provider policy mock
* fix(doctor): scope lint state snapshot to provider check
* fix(doctor): isolate selected plugin state checks
* test(doctor): restore only scoped environment
* fix(doctor): defer readiness state inspection
* fix(doctor): keep deferred config reads isolated
* fix(doctor): keep plugin state mode internal
* fix(config): preserve default plugin validation
2026-08-15 20:15:06 -06:00
1306ec805d
fix(llama-cpp): fully persist verified downloads ( #123738 )
2026-08-14 09:59:25 -07:00
f65a6f81de
feat(llama-cpp): raise default context size to 64K ( #123701 )
...
The managed llama-server default ctx-size was 8192, but the full OpenClaw
agent system prompt alone is ~31K tokens, so the first agent turn overflowed
the context window and forced immediate compaction (observed live on the Mac
app local-model onboarding). Raise the default to 65536 so a fresh local-model
install can run a real agent turn out of the box.
The default-download 16 GiB RAM floor already bounds weaker machines, and
Gemma 4 supports far more than 64K, so this only changes headroom, not the
offer gate. Docs updated to match.
2026-08-14 08:26:41 -07:00
7c5eaede1e
feat(codex): pass explicit skill selections as structured turn input ( #123441 )
...
* feat(codex): pass explicit skill selections as structured turn input
Thread OpenClaw-resolved explicit $skill selections through the reply
pipeline to the Codex app-server harness and emit them as structured
UserInput::Skill turn items when the Codex catalog (skills/list) knows the
same SKILL.md path. Codex then injects the skill natively and blocks its
text scanner from double-selecting the same name, while unmatched names
stay scannable for Codex-native-only skills and the prompt instruction
block keeps owning skills Codex cannot see. Catalog misses and RPC errors
fail open to the instruction path; the turn never fails on skill lookup.
Part of #123367
* fix(plugins): defer llama-cpp ssrf-runtime load out of the setup closure
The doctor-contract closure guard forbids the SSRF barrel in the
doctor/legacy-setup closure (cold-loads DNS, proxy state, logging); the
static import landed with the managed llama-server switch (#123105 ) and
first fails when a PR selects the guard lane. Defer to a dynamic import at
download time per the guard's stated remedy.
2026-08-13 20:37:47 -07:00
23c9c04572
fix(sessions): hide unowned host catalogs on multi-user gateways ( #123421 )
...
* fix(sessions): scope catalogs by caller identity
* fix(llama-cpp): keep doctor imports lightweight
2026-08-13 19:28:22 -07:00
1348387076
refactor(plugins): replace node-llama-cpp with managed llama-server ( #123105 )
...
Move llama.cpp chat and local embeddings onto a verified externally managed llama-server runtime. Remove the in-process native runtime, forked embedding workers, and node-llama-cpp dependency while preserving guided setup, local GGUF models, tool-capable agent runs, diagnostics, and operator docs.
2026-08-13 16:58:20 -07:00
cad77fb39c
refactor: consolidate remaining coercion helpers ( #122020 )
2026-08-11 10:22:01 -07:00
9df9cc2228
test(llama-cpp): remove duplicate progress bridge ( #121886 )
2026-08-10 23:25:51 -07:00
e67d164403
fix(llama-cpp): require full-process cleanup recovery
...
Punchcard-Session: frost-brook-timber-mx
2026-08-05 22:26:11 +08:00
51374854f0
fix(llama-cpp): keep restart error internal
...
Punchcard-Session: frost-brook-timber-mx
2026-08-05 15:53:40 +08:00
a87ee810da
fix(llama-cpp): fence native runtime handoff
...
Punchcard-Session: frost-brook-timber-mx
2026-08-05 15:33:49 +08:00
2b0e1e579a
fix(llama-cpp): report cleanup recovery to initiating request
...
Punchcard-Session: frost-brook-timber-mx
2026-08-05 14:46:39 +08:00
8b443bb34f
test(llama-cpp): split inference mapping coverage
...
Punchcard-Session: frost-brook-timber-mx
2026-08-05 02:44:19 +08:00
220445b1b7
fix(llama-cpp): type shutdown cleanup failures
...
Punchcard-Session: frost-brook-timber-mx
2026-08-05 02:43:34 +08:00
865b517235
fix(llama-cpp): record native shutdown failures
2026-08-05 02:25:52 +08:00
6dc5d00374
fix(llama-cpp): explain cleanup failure recovery
2026-08-05 02:10:27 +08:00
31d8e046bf
fix(llama-cpp): reject queued work after cleanup failure
2026-08-05 01:40:08 +08:00
622d8a048e
fix(llama-cpp): make cleanup failures terminal
2026-08-04 23:37:13 +08:00
de8afade57
refactor(llama-cpp): clarify terminal cleanup contract
2026-08-04 23:12:06 +08:00
a9d768ced6
fix(llama-cpp): detach retired model state
2026-08-04 23:03:19 +08:00
889cd316c7
fix(llama-cpp): keep failed cleanup terminal
2026-08-04 22:57:59 +08:00
48b950d9fc
fix(llama-cpp): make runtime cleanup best effort
2026-08-04 22:15:34 +08:00
b8eebb07e4
test(llama-cpp): keep lifecycle helpers private
2026-08-04 21:27:41 +08:00
76ba214b98
fix(llama-cpp): dispose runtime on plugin stop
2026-08-04 16:10:51 +08:00
1cc374b2f0
test(llama-cpp): consolidate provider fixtures ( #118426 )
2026-08-02 20:59:50 -07:00
b521e0a6dc
fix(llama-cpp): stop hijacking explicitly configured HTTP provider routes ( #118293 )
...
* fix(llama-cpp): preserve configured HTTP routes
* chore: drop release-owned CHANGELOG edit from PR branch
2026-08-02 17:06:58 -07:00
bce957fe61
fix(llama-cpp): clarify local model setup
2026-08-02 18:04:30 +08:00
e98fdeefdf
refactor(errors): consolidate message formatting ( #117818 )
2026-08-01 22:05:49 -07:00
101bcaf011
fix(llama-cpp): preserve native tool, context, and reasoning lifecycles ( #116903 )
...
* fix(llama-cpp): stream native tool call lifecycle
* fix(llama-cpp): normalize native response lifecycle and limits
---------
Co-authored-by: Peter Steinberger <steipete@macos.shared >
2026-07-31 08:36:06 -07:00
0e1304d0de
fix(llama-cpp): recover plaintext tool calls ( #116736 )
...
Co-authored-by: Peter Steinberger <steipete@macos.shared >
2026-07-31 02:45:32 -07:00
2a5ad61ce9
feat(system-agent): constrain planner JSON output at generation time ( #113482 )
...
* feat(system-agent): constrain planner JSON output
* fix(ai): align response format request typing
* fix(ai): preserve response format backend contracts
2026-07-24 21:00:28 -07:00
f4607b802f
fix(llama-cpp): preserve runtime asset paths ( #110233 )
...
Co-authored-by: VACInc <3279061+VACInc@users.noreply.github.com >
2026-07-17 16:36:05 -07:00
a5237fe925
feat(llama-cpp): gate Gemma default by RAM ( #109585 )
2026-07-16 22:17:57 -07:00
658b601ee5
feat(llama-cpp): in-process local GGUF text inference provider ( #109444 )
...
* feat(llama-cpp): add in-process text inference
* test(llama-cpp): narrow setup provider fixture
* fix(llama-cpp): trim public surface and refresh docs map
* fix(llama-cpp): import Context type in inference test
2026-07-16 18:53:55 -07:00
2eb9c7ebd7
refactor(extensions): privatize small plugin internals ( #107774 )
...
* refactor(llama-cpp): privatize embedding internals
* refactor(parallel): privatize MCP response helpers
* refactor(logbook): privatize analysis parsers
* refactor(crabbox): narrow worker provider exports
* refactor(acpx): privatize process reaper internals
* chore(deadcode): refresh unused-export baseline
2026-07-14 14:15:44 -07:00
85a96409f1
feat(memory): surface llama.cpp diagnostics
2026-07-11 16:40:14 +08:00
33163d4273
refactor(plugins): localize private declarations ( #101406 )
2026-07-06 23:26:47 -07:00
d3f7f7d1fc
chore(deadcode): remove unused test-only helpers
2026-06-22 15:48:43 +08:00
94e6255666
feat(memory): apply outputDimensionality truncation to local GGUF embeddings ( fixes #58765 ) ( #93758 )
...
* feat(memory): apply outputDimensionality truncation to local GGUF embeddings
The outputDimensionality config field was passed through to the local
embedding provider but never applied. Local GGUF models (e.g.
Qwen3-Embedding-0.6B) always returned their full dimension vector.
Apply slice(0, N) after normalization so MRL-capable models can benefit
from dimension truncation — matching the behavior already supported by
Gemini embedding-2 and OpenAI providers.
Fixes #58765
* fix(memory): preserve local embedding dimensions through worker
---------
Co-authored-by: Vincent Koc <25068+vincentkoc@users.noreply.github.com >
2026-06-17 05:05:49 +08:00
44e6caff54
fix(memory): accept local default model path migration ( #92954 )
...
* fix(memory): accept local default model path migration
Treat the official local default embedding model's hf URI and downloaded GGUF path identities as equivalent so upgraded local memory indexes do not pause solely on path-format changes.
* fix(memory): satisfy local identity lint
Avoid filtered array tail access in the local model filename helper while preserving the same compatibility behavior.
* fix(memory): preserve local embedding identity aliases
---------
Co-authored-by: Vincent Koc <25068+vincentkoc@users.noreply.github.com >
2026-06-15 09:29:42 +08:00
3137110167
fix(memory): move local llama.cpp runtime to provider plugin
...
* fix(memory): move local llama.cpp runtime to provider plugin
* chore: ignore llama cpp dynamic dependency
* test: remove invalid local provider alias fixture
* chore: refresh llama cpp shrinkwrap
* chore: drop stale memory embedding defaults facade
2026-06-09 14:30:35 +08:00
Peter Steinberger