Since 5734193fdf ("fix(plugins): keep metadata snapshot memo fresh",
first shipped in v2026.5.18), the in-process plugin metadata snapshot
memo stores derived-registry results under a key recomputed from the
freshly built snapshot.index, while lookups key off the persisted-index
registry state. On installs where the registry resolves as "derived"
(persisted index absent or not covering the running checkout), the two
keys never match. Worse, the lookup-side adoption loop returns the most
recently stored registryState for the context, so two alternating call
shapes (e.g. the model-catalog build mixing workspace-scoped and global
lookups) each adopt the other shape's state, compute a key that was
never stored, and re-run the full plugin manifest scan - on every call,
forever. Chat /models (and each subsequent provider/model pick) pays
multiple full manifest scans plus all downstream snapshot-identity cache
invalidation per step, pinning a CPU core for seconds on every
interaction, in every chat channel.
Fix: store the memo under the exact memoKey/registryState the call
looked up by, instead of re-deriving a second key from snapshot.index.
Freshness is unchanged - the lookup context hash and the plugin metadata
lifecycle clears (install/reload/doctor) still own invalidation. The
now-unused index parameter of resolvePersistedRegistryMemoState is
removed.
Measured on the author's VPS (real plugin discovery, identical catalog
output of 263 entries on both sides): the full-discovery model catalog
build behind chat /models dropped from ~6.3s to ~0.3s (~21x), with
repeat snapshot lookups going from full rescans to memo hits.
Regression test: alternating derived call shapes must not re-scan
(red on main: 4 scans; green with this fix: 2).
Limit plugin metadata snapshots to the channel, provider, and startup surfaces that need them, while preserving unscoped fallback for incomplete index data and provider runtime resolution.
Refs #70533.
Refs #84628.
Co-authored-by: IWhatsskill <IWhatsskill@users.noreply.github.com>
Extract shared normalization/coercion helpers into private @openclaw/normalization-core workspace package while preserving existing plugin SDK helper subpaths.\n\nAlso keeps direct normalization-core imports internal, wires UI/build/loader resolution, and replaces the slow PR network CodeQL lane with a fast added-line boundary scan while retaining full CodeQL for scheduled/manual runs.\n\nVerification: local moved tests, plugin SDK boundary tests, extension loader tests, agents-support shard, UI build/test, build artifacts, lint, workflow guards, autoreview, and GitHub CI passed on PR head 963d893715.
Route internal model catalog imports to the extracted @openclaw/model-catalog-core package and delete obsolete internal facades.
Keep public SDK declarations self-contained by wrapping core helpers at public boundaries instead of leaking private package imports.
Verification:
- pnpm test src/plugins/contracts/model-catalog-core-imports.test.ts src/plugins/sdk-alias.test.ts packages/model-catalog-core/src/configured-model-refs.test.ts packages/model-catalog-core/src/provider-model-id-normalize.test.ts packages/model-catalog-core/src/provider-model-id-normalization.test.ts src/config/config.model-ref-validation.test.ts src/agents/model-selection.test.ts src/plugin-sdk/provider-model-shared.test.ts -- --reporter=verbose
- pnpm check:test-types
- pnpm test:extensions:package-boundary:compile
- pnpm build
- rg "@openclaw/model-catalog-core" dist/plugin-sdk packages/plugin-sdk/dist -n --glob '*.d.ts' || true
- git diff --check
- autoreview clean after fix
CI note: merged with admin override because checks-node-agentic-commands-doctor and checks-node-core-runtime-infra-state failed twice with exit 143/no-output watchdog termination after prior passing test output, while relevant local proof and the rest of CI were green.
Refactor OpenAI provider identity so OpenAI remains the canonical provider for API-key and OAuth-backed flows while legacy openai-codex state is doctor/migration-only.
Keeps OpenAI Codex Responses as an API/transport class rather than a provider identity, moves auth aliases through providerAuthAliases, updates doctor repair sequencing for old auth/profile state, and refreshes tests/docs around the canonical OpenAI behavior.
* perf(plugins): thread metadata snapshot and discovery through hot paths
With the snapshot memo now actually hitting, route the snapshot's
manifestRegistry and discovery through the helper chains that already
had fast paths for them. Eliminates redundant per-call rebuilds at
two big amplifiers.
- Provider resolve paths (resolvePluginProviders /
isPluginProvidersLoadInFlight / resolveOwningPluginIdsForProvider /
resolveExternalAuthProfilesWithPlugins) self-service a snapshot once
at the public entry, then thread it as a separate required arg
through resolvePluginProviderLoadBase,
resolveExplicitProviderOwnerPluginIds, and the setup/runtime load
state helpers. Inner reads change from
'params.pluginMetadataSnapshot?.x' to 'snapshot.x', no more
enrichedParams clone. loadPluginManifestRegistryForInstalledIndex
fires drop ~685 -> ~10 per cold start.
- Bundled-channel / auto-enable chain accepts an optional
PluginDiscoveryResult. discoverOpenClawPlugins is fired once during
snapshot building (resolveInstalledPluginIndexRegistry already
produced it internally; now bubbled up through
loadInstalledPluginIndexWithDiscovery, PluginRegistrySnapshotResult,
and onto PluginMetadataSnapshot.discovery). load-context reads
metadataSnapshot.discovery and passes it through
applyPluginAutoEnable, so the bundled-channel cascade
(collectConfiguredChannelIds, listBundledChannelIdsWith*,
listPotentialConfiguredChannelPresenceSignals) short-circuits
instead of each leaf re-firing discovery. Persisted-cache path is
unchanged: no discovery on the snapshot, downstream chain handles
its own fallback (pre-PR behavior on that path).
* test(plugins): isolate snapshot memo across tests that mock manifest registry
The snapshot memo is now process-scoped and effective (~98% hit rate).
Three test files were depending on cache misses (because the broken
cache returned them) — each test would set up its own
loadPluginManifestRegistry mock and expect a fresh derive. With the
cache fixed, an earlier test's mocked registry now leaks into later
tests in the same file.
- io.write-config.test.ts: afterEach now clears the snapshot memo so
the 'demo' plugin mocked in the first test does not survive into
'keeps shipped plugin install config records when index migration
fails', which expects an empty registry to surface the 'plugin not
found: demo' warning.
- gateway/model-pricing-cache.ts: resetGatewayModelPricingCacheForTest
also clears the memo. Tests in model-pricing-cache.test.ts assert
loadPluginManifestRegistryForInstalledIndex was called; the memo
hit otherwise skips the call.
- providers.test.ts: vi.doMock loadPluginMetadataSnapshot to wrap the
existing loadPluginManifestRegistryMock fixture. The plumbing
commit added an auto-fetch fall-through in
resolveOwningPluginIdsForProvider; without the mock, providers
tests hit real disk reads and return empty registries (which is
what surfaced as 9 unrelated-looking failures in the prior CI
run).
* fix(plugins): preserve setup.cliBackends owner matching in provider scan
resolveOwningPluginIdsForProvider now also checks plugin.setup?.cliBackends.
The pre-PR no-registry fallback used resolvePluginContributionOwners which
includes both top-level cliBackends and setup.cliBackends; the PR's manifest
scan replacement was missing the setup case.
* fix(plugins): inherit active registry workspaceDir before loading metadata snapshot
isPluginProvidersLoadInFlight and resolvePluginProviders now resolve
env and workspaceDir once at the entry point (falling back to
getActivePluginRegistryWorkspaceDir) and pass them into both
loadPluginMetadataSnapshot and resolvePluginProviderLoadBase. Pre-fix
the snapshot used params.workspaceDir raw while the load base inherited
the active workspace, so workspace-scoped provider plugins could be
absent from the snapshot manifest registry even though owner resolution
expected them.
Regression test asserts the snapshot mock receives the active
workspaceDir when the caller omits it.
* perf(gateway): thread discovery into applyPluginAutoEnable call sites
Every gateway applyPluginAutoEnable call now passes the snapshot's
PluginDiscoveryResult so the bundled-channel cascade (collectConfiguredChannelIds
→ listBundledChannelIdsWith* → listPotentialConfiguredChannelPresenceSignals)
short-circuits instead of each leaf re-firing discovery.
Startup-time sites pull discovery from the snapshot/lookup-table they already
hold:
- server-plugin-bootstrap.ts (pluginLookUpTable)
- server-startup-plugins.ts (pluginMetadataSnapshot)
- server-startup-config.ts (pluginMetadataSnapshot)
- server-plugins.ts (pluginLookUpTable, both call sites)
Per-RPC sites (server.impl getRuntimeConfig callback, server-methods/channels
status + start handlers, server-methods/send) source discovery via
getCurrentPluginMetadataSnapshot using the runtime config to validate
compatibility. Falls through to the original slow path when the snapshot is
absent or incompatible.
Keep derived plugin metadata snapshots out of the process memo/current snapshot cache so newly added plugins under derived discovery paths are found without restart.
* fix(plugins): keep metadata memo freshness
* fix(plugins): keep metadata memo freshness
* fix(plugins): resolve metadata memo review gaps
* fix(plugins): scope metadata memo watches to env
* fix(plugins): tighten metadata memo fingerprint return type
`resolvePersistedRegistryFastMemoFingerprint` was annotated `: unknown`
but always returns object literals (`{ disabled: true }` or
`{ index, npmPackageJson }`). Spreading the unknown-typed result on
line 478 (`...fastFingerprint`) was rejected by tsgo with TS2698, which
cascaded across every check that runs the project compile (build,
tsgo:prod, check:test-types, lint, all node test shards).
Tighten the return type to `Record<string, unknown>` to match the
function's actual return shapes and unblock the spread.
* test(gateway): tolerate ENOENT in sessions.list spy predicate
The `sessions.list configuredAgentsOnly hides disk-discovered
unregistered agent stores` test spies on `fsSync.readFileSync` and
predicates with `fsSync.realpathSync.native(file) === realDiskOnlyStorePath`
for every captured read. The native realpath call throws on missing
files, so any new readFileSync of a path that may not exist (e.g. the
persisted plugin install records probe added in this PR) crashes the
predicate before the assertion runs.
Wrap the predicate in ENOENT tolerance so the test stays robust against
any future readFileSync of files that may not exist on disk.
* fix(plugins): refresh memo from cached registry
* fix(plugins): use high resolution memo fingerprints
* test(plugins): stabilize memo freshness regression
* test(cli): satisfy config mutation mock hash contract
---------
Co-authored-by: Peter Steinberger <steipete@gmail.com>
Galin Iliev