1. run.ts: remove didSendViaMessagingTool guard from incomplete turn
detection — this boolean is too coarse and blocks cross-target sends
(e.g. agent posts to slack but should still reply in originating
channel). Same-origin dedup is handled downstream by
buildReplyPayloads()/shouldSuppressMessagingToolReplies.
2. agent-runner-execution.ts: exclude isReasoning payloads from
hasNonErrorContent check — reasoning-only payloads are dropped
during delivery, so they should not prevent 429/overload error
surfacing. Also remove didSendViaMessagingTool guard for same
cross-target reason as run.ts.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
agent-runner-execution.ts: instead of returning kind:"final" which
bypasses buildReplyPayloads() filtering (streaming dedup, message_send
suppression) and post-run session bookkeeping (usage/model/provider
metadata updates), inject the error payload into runResult.payloads
and let it flow through the normal kind:"success" path.
Also adds !runResult.didSendViaMessagingTool guard to prevent
duplicate error messages when the reply was already delivered via
messaging tool.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
1. run.ts: call maybeMarkAuthProfileFailure before early return in
incomplete turn detection, so the exhausted credential enters cooldown
and multi-profile setups rotate to a healthy profile on the next turn
2. run.ts: check attempt.toolMetas for mutating tools and warn users
about potential side-effects when tools already executed before the
turn was interrupted, preventing blind retries of mutating actions
Addresses third round of review feedback on PR #50930.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
1. run.ts: exclude suppressed/recoverable tool error turns from
incomplete turn detection via !attempt.lastToolError guard — prevents
false-positive rate-limit message when buildEmbeddedRunPayloads
intentionally suppresses tool warnings
2. run.ts: use generic error message ("Agent couldn't generate a
response") instead of attributing to rate limit, since the detection
cannot distinguish mid-turn 429 from other empty-payload causes
3. agent-runner-execution.ts: remove "after tool calls completed" from
error message — this secondary check can also trigger on first-call
429 (before any tool execution), so the message should be accurate
for both pre-tool and mid-turn failures
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
1. run.ts: differentiate error message by stopReason — use rate-limit
message only for "toolUse", generic message for "error" stop reason
2. run.ts: exclude deterministic approval-prompt turns from incomplete
turn detection via !attempt.didSendDeterministicApprovalPrompt guard
3. agent-runner-execution.ts: prioritize metaErrorMsg (raw upstream error)
over errorPayloadText to avoid self-matching on pre-formatted "⚠️"
messages from run.ts
4. agent-runner-execution.ts: skip already-formatted payloads (startsWith
"⚠️") so tool-specific 429 errors are preserved rather than overwritten
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Fix legacy.ts: use registerContextEngineForOwner with core owner to
bypass public-sdk protection on default slot
- Add incomplete turn detection in run.ts: surface error when prompt()
resolves prematurely during mid-turn 429 retry producing empty payloads
- Fix TS2367: use correct StopReason union members (toolUse|error)
instead of non-existent end_turn|max_tokens
Fixes issues introduced by PR #47046 (5e293da)
* feat(minimax): add image generation and TTS providers, trim TUI model list
Register MiniMax image-01 and speech-2.8 models as plugin providers for
the image_generate and TTS tools. Both resolve CN/global base URLs from
the configured model endpoint origin.
- Image generation: base64 response, aspect-ratio support, image-to-image
via subject_reference, registered for minimax and minimax-portal
- TTS: speech-2.8-turbo (default) and speech-2.8-hd, hex-encoded audio,
voice listing via get_voice API, telephony PCM support
- Add MiniMax to TTS auto-detection cascade (after ElevenLabs, before
Microsoft) and TTS config section
- Remove MiniMax-VL-01, M2, M2.1, M2.5 and variants from TUI picker;
keep M2.7 and M2.7-highspeed only (backend routing unchanged)
* feat(minimax): trim legacy model catalog to M2.7 only
Cherry-picked from temp/feat/minimax-trim-legacy-models (949ed28).
Removes MiniMax-VL-01, M2, M2.1, M2.5 and variants from the model
catalog, model order, modern model matchers, OAuth config, docs, and
tests. Keeps only M2.7 and M2.7-highspeed.
Conflicts resolved:
- provider-catalog.ts: removed MINIMAX_TUI_MODELS filter (no longer
needed since source array is now M2.7-only)
- index.ts: kept image generation + speech provider registrations
(added by this branch), moved media understanding registrations
earlier (as intended by the cherry-picked commit)
* fix(minimax): update discovery contract test to reflect M2.7-only catalog
Cherry-picked from temp/feat/minimax-trim-legacy-models (2c750cb).
* feat(minimax): add web search provider and register in plugin entry
* fix(minimax): resolve OAuth credentials for TTS speech provider
* MiniMax: remove web search and TTS providers
* fix(minimax): throw on empty images array after generation failure
* feat(minimax): add image generation provider and trim catalog to M2.7 (#54487) (thanks @liyuan97)
---------
Co-authored-by: tars90percent <tars@minimaxi.com>
Co-authored-by: George Zhang <georgezhangtj97@gmail.com>
* fix(plugins): resolve sdk alias from import.meta.url for external plugins
When a plugin is installed outside the openclaw package (e.g.
~/.openclaw/extensions/), resolveLoaderPluginSdkPackageRoot() fails to
locate the openclaw root via cwd or argv1 hints, resulting in an empty
alias map. Jiti then cannot resolve openclaw/plugin-sdk/* imports and
the plugin fails to load with "Cannot find module".
Since sdk-alias.ts is always compiled into the openclaw package itself,
import.meta.url reliably points inside the installation directory. Add it
as an unconditional fallback in resolveLoaderPluginSdkPackageRoot() so
external plugins can always resolve the plugin SDK.
Fixes: Error: Cannot find module 'openclaw/plugin-sdk/plugin-entry'
* fix(plugins): pass loader moduleUrl to resolve sdk alias for external plugins
The previous approach of adding import.meta.url as an unconditional
fallback inside resolveLoaderPluginSdkPackageRoot() broke test isolation:
tests that expected null from untrusted fixtures started finding the real
openclaw root. Revert that and instead thread an optional moduleUrl through
buildPluginLoaderAliasMap → resolvePluginSdkScopedAliasMap →
listPluginSdkExportedSubpaths → resolveLoaderPluginSdkPackageRoot.
loader.ts passes its own import.meta.url as the hint, which is always
inside the openclaw installation. This guarantees the sdk alias map is
built correctly even when argv1 does not resolve to the openclaw root
(e.g. single-binary distributions, custom launchers, or Docker images
where the binary wrapper is not a standard npm symlink).
Tests that call sdk-alias helpers directly without moduleUrl are
unaffected and continue to enforce the existing isolation semantics.
A new test covers the moduleUrl resolution path explicitly.
* fix(plugins): use existing fixture file for moduleUrl hint in test
The previous test pointed loaderModuleUrl to dist/plugins/loader.js
which is not created by createPluginSdkAliasFixture, causing resolution
to fall back to the real openclaw root instead of the fixture root.
Use fixture.root/openclaw.mjs (created by the bin+marker fixture) so
the moduleUrl hint reliably resolves to the fixture package root.
* fix(test): use fixture.root as cwd in external plugin alias test
When process.cwd() is mocked to the external plugin dir, the
findNearestPluginSdkPackageRoot(process.cwd()) fallback resolves to
the real openclaw repo root in the CI test runner, making the test
resolve the wrong aliases. Using fixture.root as cwd ensures all
resolution paths consistently point to the fixture.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(release): add plugin-sdk:check-exports to release:check
plugin-sdk subpath exports (e.g. openclaw/plugin-sdk/plugin-entry,
openclaw/plugin-sdk/provider-auth) were missing from the published
package.json, causing external plugins to fail at load time with
'Cannot find module openclaw/plugin-sdk/plugin-entry'.
Root cause: sync-plugin-sdk-exports.mjs syncs plugin-sdk-entrypoints.json
into package.json exports, but this sync was never validated in the
release:check pipeline. As a result, any drift between
plugin-sdk-entrypoints.json and the published package.json goes
undetected until users hit the runtime error.
Fix: add plugin-sdk:check-exports to release:check so the CI gate
fails loudly if the exports are out of sync before publishing.
* fix(test): isolate moduleUrl hint test from process.cwd() fallback
Use externalPluginRoot as cwd instead of fixture.root, so only the
moduleUrl hint can resolve the openclaw package root. Previously,
withCwd(fixture.root) allowed the process.cwd() fallback to also
resolve the fixture root, making the moduleUrl path untested.
Spotted by greptile-apps review on #54283.
* fix(test): use empty string to disable argv1 in moduleUrl hint test
Passing undefined for argv1 in buildPluginLoaderAliasMap triggers the
STARTUP_ARGV1 default (process.argv[1], the vitest runner binary inside
the openclaw repo). resolveTrustedOpenClawRootFromArgvHint then resolves
to the real openclaw root before the moduleUrl hint is checked, making
the test resolve wrong aliases.
Pass "" instead: falsy so the hint is skipped, but does not trigger the
default parameter value. Only the moduleUrl can bridge the gap.
Made-with: Cursor
* fix(plugins): thread moduleUrl through SDK alias resolution for external plugins (#54283) Thanks @xieyongliang
---------
Co-authored-by: bojsun <bojie.sun@bytedance.com>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Jerry <jerry@JerrydeMacBook-Air-2.local>
Co-authored-by: yongliang.xie <yongliang.xie@bytedance.com>
Co-authored-by: George Zhang <georgezhangtj97@gmail.com>
* fix(feishu): use message create_time instead of Date.now() for Timestamp field
When a message is sent offline and later retried by the Feishu client
upon reconnection, Date.now() captures the *delivery* time rather than
the *authoring* time. This causes downstream consumers to see a
timestamp that can be minutes or hours after the user actually composed
the message, leading to incorrect temporal semantics — for example, a
"delete this" command may target the wrong resource because the agent
believes the instruction was issued much later than it actually was.
Replace every Date.now() used for message timestamps with the original
create_time from the Feishu event payload (millisecond-epoch string),
falling back to Date.now() only when the field is absent. The
definition is also hoisted to the top of handleFeishuMessage so that
both the pending-history path and the main inbound-payload path share
the same authoritative value.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* test(feishu): verify Timestamp uses message create_time
Add two test cases:
1. When create_time is present, Timestamp must equal the parsed value
2. When create_time is absent, Timestamp falls back to Date.now()
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* chore: revert unrelated formatting change to lifecycle.test.ts
This file was inadvertently formatted in a prior commit. Reverting to
match main and keep the PR scoped to the Feishu timestamp fix only.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(feishu): use message create_time for inbound timestamps (#52809) (thanks @schumilin)
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: George Zhang <georgezhangtj97@gmail.com>
* fix(feishu): close WebSocket connections on monitor stop/abort
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* test(feishu): add WebSocket cleanup tests
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(feishu): close WebSocket connections on monitor stop (#52844) (thanks @schumilin)
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: George Zhang <georgezhangtj97@gmail.com>
Josh Lehman
chenxingzhen
Matt Van Horn
Tak Hoffman
Harold Hunt
Peter Steinberger
Jackal Xin
liyuan97
xieyongliang
Devin Robison
Lin Z
Nimrod Gutman
khhjoe