* fix(openai-codex): honor providerConfig.baseUrl in dynamic-model synthesis fallback
The synthesis fallback in resolveCodexForwardCompatModel hardcoded
OPENAI_CODEX_BASE_URL when the model registry had no template row to
clone, which meant openai-codex providers configured with a custom
baseUrl (e.g. a local proxy that forwards Codex traffic) silently
fell back to api.openai.com / chatgpt.com - bypassing the proxy and
typically failing the auth contract.
Synthesis now reads ctx.providerConfig.baseUrl when present, with the
existing OPENAI_CODEX_BASE_URL constant as the fallback. No effect on
template-clone or registry-find paths, which already inherit the
configured baseUrl through the cloned template.
* docs(changelog): add Unreleased Fixes entry for #76428 codex synthesis baseUrl honor
* fix(gateway): read-only fast path for models.list catalog loading
The gateway model catalog refresh calls loadModelCatalog without
readOnly, triggering ensureOpenClawModelsJson (60-70s), full PI SDK
registry instantiation, auth storage discovery, and live provider
plugin augmentation on every Control UI list/refresh. None of this
is needed for a read-only UI listing.
Three changes:
1. Gateway catalog refresh now passes readOnly: true to loadModelCatalog.
2. In readOnly mode, skip augmentModelCatalogWithProviderPlugins — live
provider discovery is explicit admin/background work, not a UI list
operation.
3. Add a persisted models.json fast path: when readOnly is true, first
try reading the existing models.json directly and converting
providers.<provider>.models[] to catalog rows. Falls back to the
full PI registry path if the file is missing or unreadable.
Observed improvement on a production install:
loadGatewayModelCatalog: 967 entries / 4651ms → 89 entries / 8ms
Live models.list during startup: ~18s → ~2s
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix(gateway): preserve full model catalog view
* fix(agents): preserve read-only catalog defaults
* fix(agents): preserve provider catalog defaults
---------
Co-authored-by: Marvinthebored <peter@lindsey.jp>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: Vincent Koc <vincentkoc@ieee.org>
Resolve bare official external plugin IDs through the official catalog before generic npm fallback, preserving explicit npm semantics and catalog integrity through the hook-pack fallback.\n\nFixes #76373.\n\nThanks @bek91 and @vincentkoc.
sessions.list calls buildGatewaySessionRow for every visible session,
running transcript usage fallback, display model inference, cost/context
recomputation, thinking level enumeration, agent runtime metadata, and
plugin extension projection per row. On installs with 30-50+ sessions
this blocks the event loop for 20-80+ seconds, starving Discord
heartbeats and Control UI RPCs.
Add skipTranscriptUsageFallback and lightweightListRow flags to
buildGatewaySessionRow. In lightweight mode, skip transcript usage
fallback, display model inference, cost/context recomputation, thinking
level options, agent runtime metadata, and plugin extension projection.
Use persisted entry fields directly for cost, tokens, and model identity.
listSessionsFromStoreAsync now passes both flags for bulk list rows.
Detail endpoints and single-row loads are unaffected.
Observed improvement on a production install (33 sessions):
sessions.list row construction dropped from ~82s to ~6s.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
(cherry picked from commit 803879ec54)
Large transcript JSONL records (multi-MB tool results, file content)
block the event loop via JSON.parse before truncation logic can skip
them. Add a 256 KiB line-size guard to parseTailTranscriptRecord and
extractUsageSnapshotFromTranscriptLine, and replace the full transcript
index scan in readSessionTitleFieldsFromTranscriptAsync with the
existing bounded sync reader.
Observed improvement on a production install (33 sessions):
chat.history dropped from 13-16s to ~1.2s, event loop utilization
from 0.999 to normal, steady-state CPU from ~100% to 0.2-0.3%.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The 2026.4.30 changelog section maps to a beta-only tag with no public
release page, so the file-transfer highlight was invisible on the
releases page. Moving to Unreleased so it lands in the next published
release.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Fixes#76415.
- Explains compacted history boundaries in WebChat.
- Adds an Open checkpoints action for pre-compaction recovery.
- Updates WebChat docs and changelog with Thanks @BunsDev.
- Validated targeted UI tests, formatting/diff checks, Testbox changed gate, and exact-head CI.
Security: UI/docs/tests/styles-only change that reuses existing checkpoint APIs; no new dependencies, filesystem reads, workflow changes, or secret handling.
Preserve the ClawHub CLI dry-run preflight while making the printed publish preview include CLAWHUB_WORKDIR. Add regression coverage that stubs the ClawHub CLI and verifies --dry-run is forwarded through the publish script.
Summary:
- The PR adds a Feishu-local media-aware dedupe-key helper, wires it through inbound receive/debounce/persistent/broadcast dedupe paths, adds Feishu regression coverage, and adds an Unreleased changelog entry.
- Reproducibility: yes. Source inspection shows a high-confidence path: two Feishu audio events for the same a ... _key` values collide in current-main receive and persistent dedupe before media parsing distinguishes them.
Automerge notes:
- No ClawSweeper repair was needed after automerge opt-in.
Validation:
- ClawSweeper review passed for head c0229f8a48.
- Required merge gates passed before the squash merge.
Prepared head SHA: c0229f8a48
Review: https://github.com/openclaw/openclaw/pull/76408#issuecomment-4365292618
Co-authored-by: clawsweeper <274271284+clawsweeper[bot]@users.noreply.github.com>
Co-authored-by: SymbolStar <24540119+SymbolStar@users.noreply.github.com>
Summary:
- The PR removes session-file repair's trailing-assistant disk trim, updates regression coverage, clarifies transcript hygiene docs, and adds a changelog entry for the Telegram/WebChat history loss fix.
- Reproducibility: yes. Current main has a clear source path: a normal trailing assistant JSONL record is popped by `repairSessionFileIfNeeded`, and the main-branch test suite asserts that deletion.
Automerge notes:
- PR branch already contained follow-up commit before automerge: docs(agents): clarify session repair preservation
- PR branch already contained follow-up commit before automerge: fix(agents): preserve delivered assistant replies in session repair
Validation:
- ClawSweeper review passed for head 66c187fd76.
- Required merge gates passed before the squash merge.
Prepared head SHA: 66c187fd76
Review: https://github.com/openclaw/openclaw/pull/76420#issuecomment-4365323320
Co-authored-by: Ayaan Zaidi <hi@obviy.us>
Co-authored-by: clawsweeper <274271284+clawsweeper[bot]@users.noreply.github.com>
* fix(discord): surface stalled transport health
* fix(discord): surface stalled transport health
* fix(discord): surface stalled transport health
---------
Co-authored-by: clawsweeper <274271284+clawsweeper[bot]@users.noreply.github.com>
Summary:
- The PR broadens memory-core's embedding retry classifier for socket/network errors, adds focused classifier tests, and adds an Unreleased changelog fix.
- Reproducibility: yes. Current main's retry classifier rejects the socket/fetch samples, and the reindex embedding path delegates to that classifier; a read-only regex probe confirmed the PR accepts the target samples.
Automerge notes:
- PR branch already contained follow-up commit before automerge: fix(memory): retry reindex on socket errors
Validation:
- ClawSweeper review passed for head b4618c4532.
- Required merge gates passed before the squash merge.
Prepared head SHA: b4618c4532
Review: https://github.com/openclaw/openclaw/pull/76311#issuecomment-4364956064
Co-authored-by: buyitsydney <buyitsydney@users.noreply.github.com>
Co-authored-by: clawsweeper <274271284+clawsweeper[bot]@users.noreply.github.com>
