7096819f2b
fix(acpx): retry queue-owner repair without resume-session (thanks @obviyus)
2026-04-01 18:30:38 +05:30
fc745db76d
ci: remove bun workflow
2026-04-01 21:58:46 +09:00
131f6dac37
refactor: unify failover signal classification
2026-04-01 21:50:58 +09:00
ed482b1ce7
fix: repair queue owner session recovery ( #58669 ) (thanks @neeravmakwana)
...
* fix(acpx): repair queue owner session recovery
* fix(acpx): avoid duplicate queue owner recovery
* fix: repair queue owner session recovery (#58669 ) (thanks @neeravmakwana)
---------
Co-authored-by: Ayaan Zaidi <hi@obviy.us>
2026-04-01 18:20:23 +05:30
cd07ebef99
fix: correct flows docs to tasks ( #58690 ) (thanks @neeravmakwana)
...
* Docs: fix stale flows command references
* Docs: address flows review comments
* docs: remove stale flows subtree from cli index
* fix: correct flows docs to tasks (#58690 ) (thanks @neeravmakwana)
---------
Co-authored-by: Ayaan Zaidi <hi@obviy.us>
2026-04-01 18:14:31 +05:30
25e2934809
refactor: route session target matching through plugin parsers
2026-04-01 13:42:57 +01:00
6433e923d4
fix: add ZAI GLM-5.1 and GLM-5V Turbo support ( #58793 ) (thanks @tomsun28)
...
* provider(zai): add GLM-5.1 and GLM-5V Turbo models
* feat(zai): extract model definition builder for glm-5 forward compat
* test(zai): cover persisted glm-5 dynamic metadata
* fix: add ZAI GLM-5.1 and GLM-5V Turbo support (#58793 ) (thanks @tomsun28)
* fix: preserve ZAI dynamic model transport config (#58793 ) (thanks @tomsun28)
---------
Co-authored-by: gongchao <chao.gong@aminer.cn>
Co-authored-by: Ayaan Zaidi <hi@obviy.us>
2026-04-01 18:09:02 +05:30
33fbd9b770
refactor: dedupe auth profile store normalization
2026-04-01 13:37:37 +01:00
ab3c646bb1
fix: preserve telegram exec approval topic routing
2026-04-01 13:34:50 +01:00
f55b6b1acf
fix: add changelog for auth profile store load crash ( #58923 ) (thanks @openperf)
2026-04-01 21:33:10 +09:00
ac68321d4d
fix(auth-profiles ): ensure credential key and token are strings to prevent crash
...
Fixes #58861
2026-04-01 21:33:10 +09:00
cd4b03c568
fix: unify structured provider failover classification ( #58856 ) (thanks @aaron-he-zhu)
2026-04-01 21:31:18 +09:00
c96ee42300
fix(agents): normalize provider errors for better failover
...
Add provider-specific error patterns for AWS Bedrock, Ollama, Mistral,
Cohere, DeepSeek, Together AI, and Cloudflare Workers AI. These providers
return errors in non-standard formats that the generic classifiers miss,
causing incorrect failover behavior (e.g., context overflow misclassified
as format error, ThrottlingException not recognized as rate limit).
Wire provider patterns into isContextOverflowError() and
classifyFailoverReason() as catch-all layers after generic classifiers.
2026-04-01 21:31:18 +09:00
74b9f22a42
fix: add Telegram error suppression controls ( #51914 ) (thanks @chinar-amrutkar)
...
* feat(telegram): add error policy for suppressing repetitive error messages
Introduces per-account error policy configuration that can suppress
repetitive error messages (e.g., 429 rate limit, ECONNRESET) to
prevent noisy error floods in Telegram channels.
Closes #34498
* fix(telegram): track error cooldown per message
* fix(telegram): prune expired error cooldowns
* fix: add Telegram error suppression controls (#51914 ) (thanks @chinar-amrutkar)
---------
Co-authored-by: chinar-amrutkar <chinar-amrutkar@users.noreply.github.com>
Co-authored-by: Ayaan Zaidi <hi@obviy.us>
2026-04-01 17:52:28 +05:30
ef286987e7
test: fix context pruning runtime fixtures
2026-04-01 16:48:00 +05:30
f70ad924a6
fix: align cache-ttl pruning with thinking replay sanitization
2026-04-01 16:33:57 +05:30
c7510e0f1a
fix(hooks): skip full gate for docs-only commits
2026-04-01 20:02:54 +09:00
c65e152b39
fix: preserve anthropic thinking replay ( #58916 )
...
* test: add anthropic thinking replay regressions
* fix: preserve anthropic thinking blocks on replay
* fix: preserve anthropic thinking replay (#58916 )
* fix: move anthropic replay changelog entry (#58916 )
2026-04-01 16:23:47 +05:30
00a49fe8b4
docs: add gateway.webchat.chatHistoryMaxChars config reference
2026-04-01 19:25:17 +09:00
32ae841098
feat(web-search): add SearXNG as bundled web search provider plugin ( #57317 )
...
* feat(web-search): add bundled searxng plugin
* test(web-search): cover searxng config wiring
* test(web-search): include searxng in bundled provider inventory
* test(web-search): keep searxng ordering aligned
* fix(web-search): sanitize searxng result rows
---------
Co-authored-by: Vincent Koc <vincentkoc@ieee.org>
2026-04-01 19:24:33 +09:00
3f67581e50
fix: retry safe wrapped Telegram send failures ( #51895 ) (thanks @chinar-amrutkar)
...
* fix(telegram): traverse error .cause chain in formatErrorMessage and match grammY HttpError
grammY wraps network failures in HttpError with message
'Network request for ... failed!' and the original error in .cause.
formatErrorMessage only checked err.message, so shouldRetry never
fired for the most common transient failure class.
Changes:
- formatErrorMessage now traverses .cause chain, appending nested
error messages (with cycle protection)
- Added 'Network request' to TELEGRAM_RETRY_RE as belt-and-suspenders
- Added tests for .cause traversal, circular references, and grammY
HttpError retry behavior
Fixes #51525
* style: fix oxfmt formatting in retry-policy.ts
* fix: add braces to satisfy oxlint requirement
* fix(telegram): keep send retries strict
* test(telegram): cover wrapped retry paths
* fix(telegram): retry rate-limited sends safely
* fix: retry safe wrapped Telegram send failures (#51895 ) (thanks @chinar-amrutkar)
* fix: preserve wrapped Telegram rate-limit retries (#51895 ) (thanks @chinar-amrutkar)
---------
Co-authored-by: chinar-amrutkar <chinar-amrutkar@users.noreply.github.com>
Co-authored-by: Ayaan Zaidi <hi@obviy.us>
2026-04-01 15:54:29 +05:30
5c8d9da749
docs: add SearXNG web search provider page and navigation
2026-04-01 19:18:15 +09:00
f1595f59b4
fix(ci): allow plugin npm preview without publish token ( #58929 )
2026-04-01 19:16:46 +09:00
d766bfc6b2
fix(memory): preserve session indexing during full reindex ( #39732 )
...
Merged via squash.
Prepared head SHA: 0dbaf5fffb
2026-04-01 13:12:30 +03:00
1654c3a851
feat(gateway): make chat history max chars configurable ( #58900 )
...
* feat(gateway): make chat history max chars configurable
* fix(gateway): address review feedback
* docs(changelog): note configurable chat history limits
2026-04-01 21:08:37 +11:00
b2bb129e2c
docs(changelog): note chat error fallback fix
2026-04-01 20:53:35 +11:00
78b48735fa
test: restore fetch stubs in embedding suites
2026-04-01 20:53:16 +11:00
101c31f5e1
test: harden ci-sensitive unit suites
2026-04-01 20:53:16 +11:00
4e63dc0b1c
fix: hide raw provider errors from chat replies
2026-04-01 20:53:16 +11:00
fbe3ca4d7d
fix(plugins): pass dangerouslyForceUnsafeInstall through archive and … ( #58879 )
...
Merged via squash.
Prepared head SHA: 87eb27d902
2026-04-01 02:52:01 -07:00
72af92ba4e
qqbot: require explicit allowlist for /bot-logs to prevent info disclosure ( #58895 )
...
* qqbot: harden /bot-logs authorization fallback
* fix(qqbot): harden bot logs allowlist guard
* fix(qqbot): normalize bot logs allowlist entries
2026-04-01 18:40:46 +09:00
07c60ae461
fix(agent): treat webchat exec approvals as native UI ( #58904 )
...
* fix(agent): treat webchat exec approvals as native UI
* docs(changelog): note webchat exec approval UI fix
* test(agent): cover webchat native approval guidance
2026-04-01 18:36:21 +09:00
8b2d24b62b
docs(security): clarify node pairing trust boundary
2026-04-01 18:27:23 +09:00
29784af1e2
style(gateway): normalize node reconnect formatting
2026-04-01 18:27:06 +09:00
f6317fb747
fix(gateway): stop pinning node commands to pairing state
2026-04-01 18:27:06 +09:00
fe57ee513f
test: drop stale task boundary allowlist entries
2026-04-01 10:17:47 +01:00
d005cc8b42
test: align cron abort regression with #58833
2026-04-01 10:17:47 +01:00
92f1772e93
test: allow boundary test on main
2026-04-01 10:17:47 +01:00
f559ea126d
fix: land slash command metadata parsing ( #58725 ) (thanks @Mlightsnow)
2026-04-01 10:17:47 +01:00
3b1f8e3461
fix: strip inbound metadata before slash command detection ( #58674 )
...
Slash commands like /model and /new were silently ignored when the inbound
message body included metadata prefix blocks (Conversation info, Sender info,
timestamps) injected by buildInboundUserContextPrefix. The command detection
functions (hasControlCommand, isControlCommandMessage, parseSendPolicyCommand)
now call stripInboundMetadata before normalizeCommandBody so embedded slash
commands are correctly recognized.
2026-04-01 10:17:20 +01:00
fb28b02540
fix: preserve bundled channel plugin compat ( #58873 )
...
* fix: preserve bundled channel plugin compat
* fix: preserve bundled channel plugin compat (#58873 )
* fix: scope channel plugin compat to bundled plugins (#58873 )
2026-04-01 14:42:36 +05:30
2d53ffdec1
fix(exec): resolve remote approval regressions ( #58792 )
...
* fix(exec): restore remote approval policy defaults
* fix(exec): handle headless cron approval conflicts
* fix(exec): make allow-always durable
* fix(exec): persist exact-command shell trust
* fix(doctor): match host exec fallback
* fix(exec): preserve blocked and inline approval state
* Doctor: surface allow-always ask bypass
* Doctor: match effective exec policy
* Exec: match node durable command text
* Exec: tighten durable approval security
* Exec: restore owner approver fallback
* Config: refresh Slack approval metadata
---------
Co-authored-by: scoootscooob <zhentongfan@gmail.com>
2026-04-01 02:07:20 -07:00
4ceb01f9ed
docs(gateway): document node pairing repair flow
2026-04-01 18:02:56 +09:00
db0cea5689
refactor(gateway): extract node pairing reconciliation
2026-04-01 18:02:31 +09:00
4590ac31cc
fix: note exec approval continuation in changelog ( #58860 ) (thanks @Nanako0129)
2026-04-01 17:56:55 +09:00
7f53c1ca00
test(exec): cover delayed Discord approval continuation
2026-04-01 17:56:55 +09:00
63da2c7034
fix(exec): resume agent session after approval completion
2026-04-01 17:56:55 +09:00
4fa11632b4
fix: escalate to model fallback after rate-limit profile rotation cap ( #58707 )
...
* fix: escalate to model fallback after rate-limit profile rotation cap
Per-model rate limits (e.g. Anthropic Sonnet-only quotas) are not
relieved by rotating auth profiles — if all profiles share the same
model quota, cycling between them loops forever without falling back
to the next model in the configured fallbacks chain.
Apply the same rotation-cap pattern introduced for overloaded_error
(#58348 ) to rate_limit errors:
- Add `rateLimitedProfileRotations` to auth.cooldowns config (default: 1)
- After N profile rotations on a rate_limit error, throw FailoverError
to trigger cross-provider model fallback
- Add `resolveRateLimitProfileRotationLimit` helper following the same
pattern as `resolveOverloadProfileRotationLimit`
Fixes #58572
* fix: cap prompt-side rate-limit failover (#58707 ) (thanks @Forgely3D)
* fix: restore latest-main gates for #58707
---------
Co-authored-by: Ember (Forgely3D) <ember@forgely.co>
Co-authored-by: Peter Steinberger <steipete@gmail.com>
2026-04-01 17:54:10 +09:00
8fce663861
fix(subagents): harden task-registry lifecycle writes ( #58869 )
...
* fix(subagents): harden task-registry lifecycle writes
* chore(changelog): note subagent task-registry hardening
* fix(subagents): align lifecycle terminal timestamps
* fix(subagents): sanitize lifecycle warning metadata
2026-04-01 17:50:52 +09:00
1ce410a7be
fix(sandbox): use browser image for browser runtime matching ( #58759 )
...
* fix(sandbox): compare browser runtimes against sandbox browser image
* refactor(sandbox): route docker runtime matching by config label kind
* fix(sandbox): tag browser runtime removals correctly
* test(sandbox): share docker backend test config
* fix(sandbox): normalize browser runtime image matching
---------
Co-authored-by: Vincent Koc <vincentkoc@ieee.org>
2026-04-01 17:44:52 +09:00
Ayaan Zaidi
Peter Steinberger
Neerav Makwana
Tomsun28
openperf
Aaron Zhu
Chinar Amrutkar
Vincent Koc
Charles Dusek
upupc
Luke
ryanlee-gemini
HansY
nanakotsai
Forgely3D
sandpile