Ayaan Zaidi

fix(gateway): classify chat error kind via canonical failover reasons · improve(ci): shorten compact PR test critical path · refactor(agents): converge run staleness policy onto diagnostic activity owner · refactor(agents): converge CLI tool terminal reason onto run-termination

fix(telegram): scope reply-fence abort authority to pre-adoption dispatch · fix(agents): stop warning about recovered exec failures after a successful reply · tool-error warning: recovered shell failure still emits trailing '⚠️ 🛠️ Bash failed' after a successful reply (codex app-server) · test: incomplete logging/console vi.mock factories poison co-sharded tests (unit-fast flake)

chore: remove stale repository artifacts · feat(telegram): open the dashboard in a Telegram Mini App · fix(infra): converge legacy state migrations on archive collisions · [Bug]: startup legacy-state migration never converges when .migrated archive already exists; gateway permanently refuses to start

fix(agents): run-lifecycle reliability — bounded release, evidence-based liveness, watchdog semantics · fix(codex): ungate app-server terminal liveness and fan out poisoned-client retirement · fix(agents): keep finite LLM idle watchdog under unlimited run budgets · fix(agents): make run liveness evidence-based

fix(auto-reply): bound terminal reply-run settle and reclaim stale lanes · fix(slack): unbounded thread pagination, process-wide write serialization, serialized inbound lookups · Design: run lifecycle reliability — bounded liveness, abort⇒release, crash resume, completion drive · [Bug]: Slack replies ~2s slower than other channels — dominant cost is Socket Mode event delivery; channel-side stalls fixed in #101888

fix(gateway): commit runtime snapshot on noop config reloads · feat(gateway): add persisted crash-loop breaker and fatal-config exit contract · fix(gateway): commit runtime snapshot on noop config reloads (#100586) · fix: hot-reload visible reply mode changes

fix(cron): job edits no longer break CLI-backend runs by stripping the default toolsAllow marker · fix(agents): queued CLI reasoning bridge + thinking token progress for claude-cli's withheld-text wire · fix(agents): preserve spill-file pointers through elision and spill truncated web_fetch output · Truncation discards recoverable output: history elision drops exec spill-file pointers; web_fetch never spills

Truncation discards recoverable output: history elision drops exec spill-file pointers; web_fetch never spills · fix(agents): scale aggregate tool-result budget with context window and compact on pressure · [Bug]: Window-relative compaction never triggers while the absolute aggregate tool-result cap silently zeroes fresh tool results (large configured contextWindow) · fix(telegram): classify inbound events from the canonical mention decision so direct mentions stop lurking

CLI session binding churns on mention-state toggles: ExplicitlyMentionedBot leaks into extraSystemPromptStatic · fix: CLI agent session resume churns when owner and non-owner alternate in a group · CLI session binding churns in multi-sender groups: promptToolNamesHash flips with senderIsOwner · fix(agents): keep cli session binding facts session-stable

Persist observed room events as bare transcript rows for durable ambient group awareness · fix(auto-reply): single canonical group history and deduped turn metadata · Per-turn agent input payload duplicates group history and restates sender/delivery facts up to 5x · fix(agents): claude-cli lifecycle cleanup — loopback fail-loud, exit-0 failover, bounded reseed, image sweep, one prepare cleanup owner

fix(telegram): final replies no longer drop on rejected rich entities, captions, quotes, or long flood waits · [Bug]: Telegram durable/final sends lack the fallbacks the streaming path has, so the final reply is the one that drops · [Bug]: Telegram webhook mode acks 200 before persistence — updates lost forever on crash/restart · fix(telegram): back off, dead-letter, and tombstone spooled updates so poison messages cannot block or duplicate

fix(provider-transport-fetch): raise SSE sanitize buffer cap to 16 MiB — unbreak gpt-5.5 on chatgpt-responses (main regression from #96989, not in 6.11) · fix(gateway): reject agent RPC runs for deleted-agent sessions · [codex] Fix Telegram album sibling media context · fix(config): fail closed when config is unreadable before a full-file write

Fix Telegram chunked reply delivery loss · fix(telegram): preserve literal reasoning tags

fix(telegram): preserve long streamed reply chunks · Fix Telegram chunked reply delivery loss · fix(telegram): preserve literal reasoning tags · test(telegram): cover retained preview chunk gaps

Fix Codex synthetic usage in status without local OpenAI profiles · fix(anthropic): surface reasoning and pre-tool commentary on Discord

fix(auto-reply): serialize reply session initialization · fix(media): strip internal UUID suffix from outbound media filenames · fix(telegram): topic replies stall after session conflicts · fix(telegram): keep draft preview chunks surrogate-safe

fix(telegram): keep OAuth emails out of rich entity detection · fix(v2026-6-9): telegram richMessages breaks paragraph breaks and table rendering · fix(telegram): sanitize outbound tool traces · fix(cron): drop unenforceable default toolsAllow cap on CLI-backed runs

fix #86957: drain worker-spooled Telegram updates immediately · fix #95248: OpenClaw release_lane is a no-op when claim is held by a live worker; stuck Telegram inbound events block agent response until gateway restart · fix(claude-cli): also disable native background Bash and Monitor in --print runs · test(telegram): keep live polling leases protected

fix #95378: https://github.com/openclaw/openclaw/issues/95378 · fix(telegram): skip mirror write when primary reply already exists (#94930) · fix(telegram): honor outbound reaction directives · fix(telegram): stop clearing registered webhook on channel restart

fix(telegram): send progress previews as html text · fix(channels): preserve command progress detail

[codex] Fix Telegram rich local Markdown link hrefs · fix(telegram): set richMessages default to false explicitly in schema · [codex] Fix Telegram rich local Markdown link hrefs (#94118) · fix(telegram): set richMessages default to false explicitly in schema (#94048)

Fix Telegram package output mount · Fix package Telegram temp root · Fix package Telegram RTT evidence · fix(telegram): surface rich-message disabled state

Fix Telegram rich progress detail updates · Fix Telegram rich progress detail updates (#93698)

Fix Telegram rich progress command output · feat(telegram): send rich messages as rich html · fix(telegram): recover pid-reused ingress claims · fix(telegram): preserve rich markdown line breaks

Forward suppressed-source progress for message-tool channel replies · fix(auto-reply): strip delivery hints from leading metadata · fix(plugin-sdk): expose delivery hints without utility imports · fix(auto-reply): share message-tool delivery hints

feat(telegram): send rich message text · feat(usage): ship built-in /usage full footer · fix(gateway): mirror hidden commentary-phase assistant events · fix(OpenAI Responses): disable item id replay for storeless providers

fix(telegram): classify streaming preview edit failures instead of killing the draft · fix(telegram): survive getUpdates conflicts in isolated polling ingress · fix(gateway): mirror hidden commentary-phase assistant events

fix(auto-reply): stop dropping claude-cli narration when commentary lane is off · feat(auto-reply): durable inter-tool commentary via verbose standalone progress (supersedes #89850/#89890) · refactor(agents): derive CLI commentary classification from consumer presence · refactor(auto-reply): distill verbose commentary lane wiring

fix(telegram): survive getUpdates conflicts in isolated polling ingress · fix(telegram): classify streaming preview edit failures instead of killing the draft · fix(telegram): audit follow-ups — block-mode chunk config, dedupe bucket cleanup, grammy contract trust · fix(telegram): use SDK dispatch dedupe

fix(agents): deliver native /compact replies through source suppression · fix(telegram): restore /compact on generic message ingress · fix(agents): keep compaction notices additive

fix(reply-queue): remove the drained item by reference instead of front index · fix(telegram): route account-scoped topic agents · fix(cli): bridge inter-tool commentary events to channel progress · feat(cli): emit commentary progress events from Claude CLI parser

fix(android): align provider readiness with available models · fix(sessions): reconcile stale terminal main transcripts · test(gateway): avoid future session fixture timestamps · test(telegram): align transcript append mock

fix(acp): default parent commentary in progress mode · fix(message-tool): stabilize send idempotency keys · fix(acp): re-add opt-in parent commentary progress · fix: stabilize Anthropic cache marker through tool loops

fix(auto-reply): surface fatal channel errors · Share channel progress draft compositor · fix(android): improve companion-first shell UX · Suppress internal agent failure traces before channel delivery

feat(agents): generalized native compaction ownership for CLI backends · feat(android): add installed apps node command · refactor(agents): trim native compaction ownership follow-up · fix(android): classify updated system apps

Share commentary progress drafts with Telegram · fix(android): add notification app picker

Retry stale CLI sessions inside runner lifecycle · Fix silent internal aborts after tool-use turns · fix(agents): skip below-target CLI compaction failures · refactor(agents): simplify stale cli retry cleanup

Fix iMessage startup watch replay · Share commentary progress drafts with Telegram · perf(cli): reduce room-event Claude cache churn · fix(gateway): recover channels after reload stop timeout

fix(agent): preserve media task success on delivery miss · Reuse CLI sessions for room events · fix(trajectory): keep latest runtime capture within cap · fix(command): retry claude-cli transcript probe to close flush race

fix(telegram): avoid replaying long streamed replies · Alias Docker main tags to latest release · fix(telegram): clamp partial draft overflow · Publish Docker browser release images

fix(worker): retry opaque AWS RunInstances failures · fix(worker): retry opaque aws runinstances failures

ci(mantis): pass crabbox capacity regions

fix(telegram): preserve inbound text entities · fix(mantis): release telegram user leases on startup failure · fix(worker): preserve active AWS SSH ingress · fix(android): harden play media permission removal

Fix Telegram legacy cache migration · [codex] Store Telegram caches in plugin state · style(android): sharpen voice mode surfaces · style(android): refine list surface spacing

feat(android): overhaul mobile experience · fix(android): address overhaul review findings · refactor(android): make overhaul UI canonical · test(android): update gateway hello callback fixtures

feat(xai): add device code OAuth login · fix(mantis): finish interrupted telegram proof sessions · fix(xai): use public plugin test runtime (#84005) (thanks @fuller-stack-dev) · fix(xai): add device code changelog (#84005) (thanks @fuller-stack-dev)

Add Telegram provider setup flow · fix: harden release stability recovery and auth fallback · [codex] Add gated Telegram native DM tool-progress drafts · fix(telegram): avoid mirroring progress into active transcripts

fix(telegram): preserve bot reply-chain context · fix(android): prompt on changed TLS thumbprint · fix(telegram): warn when a media group silently drops failed photos [AI-assisted] · ci(mantis): add telegram proof label trigger

Discord ambient room events · Fix Telegram truncated progress finals · Fix Telegram context loss after restart replay · docs(changelog): note telegram final reply fix

fix(qa): emit Discord trigger timing artifacts · fix(mantis): publish evidence to r2 · Add Discord RTT tracking · data(rtt): record 2026.5.12 retest

fix(mantis): publish evidence to r2 · fix(telegram): keep plugin slash commands on native path · Fix Telegram cron HTML announce formatting · Backport oauthRef runtime auth fix to release 2026.5.12

Treat ambient group chatter as room events · Bootstrap agent sessions before send · Fix Telegram polling stall liveness · fix(agents): escalate LLM idle timeout to model fallback after profile rotation

fix(matrix): default markdown tables to bullets across Matrix clients · fix(agents): use run workspace for post-compaction context · fix(active-memory): classify threaded Telegram DMs · fix(agents): recover read offsets beyond EOF

Fix Telegram reply turn context shape · Render Telegram reply context inline · fix(browser): read Chrome MCP screenshot extension · fix(telegram): honor force document for videos

fix(context): persist Codex run context maps · fix Telegram stale reply context · Add Telegram real-user Crabbox proof · Detect Docker Playwright Chromium

Add /context map treemap · Fix Telegram stale reply context · perf(telegram): append reply-chain cache records · Add Telegram auth transition E2E

fix(telegram): share bot api throttler · Fast-path native slash commands · Harden Telegram QA live harness · Hydrate Telegram reply chains

Use grammY inline keyboard helper · refactor(telegram): simplify provider delivery and auth · telegram: avoid silent fallback after message tool send (#78685) · docs(changelog): note telegram poll cap (#78762) (thanks @obviyus)

Fix Telegram embedded final transcript gap-fill · fix(net): bound guarded fetch dispatcher cleanup · fix(gateway): fast-path qualified session-list model refs · Show current thinking level in Telegram picker

fix(embed): arm lastBlockReplyText only after block reply emission · fix: expose ollama thinking profile before activation · fix(update): stop dev updates after fetch failure · fix(agents): persist embedded session JSONL transcripts (#77823)

Clarify reply target context · Fix Telegram forum topic final delivery · fix(telegram): start polling after webhook cleanup timeout · fix(gateway): continue update runs after restart

Fix plugin-only tool and registry latency regressions · fix #76071: handle thoughtSignature-only parts to prevent Gemini stream hang · fix(agents): exempt tool-execution timeouts from model fallback (#52147) · fix(cron): retry recurring wake-now jobs on temporary busy skips

Fix Telegram reply cold-start latency · data: drop canary-only main rtt row · fix(rtt): wait between telegram samples · fix(rtt): parse telegram scenario list

Remove Telegram native draft preview transport

feat(telegram): support mini app URL buttons · Fix GitHub Copilot model picker catalog · Fix provider-scoped manifest model picker · Fix Telegram WSL2 gateway stalls

Fix native slash command replies in groups · fix(telegram): persist native command metadata to target sessions · fix(agents): seed claude-cli fallback prompts with prior-session context (#69973) · fix(bundled-runtime-deps): collapse nested cache pluginRoot to enclosing key (Fixes #72956)

Fix stale model provider API config recovery

test(e2e): cover npm onboarding runtime deps · Fix CLI transcript compaction lifecycle · fix(agents): allow large Claude live JSONL lines · feat(tts): add provider personas

fix(active-memory): isolate recall auth health · fix(config): allow per-agent contextTokens override in agents.list · fix(agents): guard Windows CLI argv against ENAMETOOLONG by moving system prompt to stdin · fix(agents): make sessions_spawn mode=session errors actionable when thread binding is unavailable (#67400)

ci: use release approval for npm Telegram E2E · ci: add manual npm Telegram beta E2E · test: add npm Telegram live smoke

Fix cron message trace channel · Measure Telegram QA reply RTT · Fix Claude CLI session continuity across OAuth refresh · fix(gateway): default restart acknowledgements

refactor(gateway): unify startup task execution · fix(gateway): write restart sentinels atomically · fix(gateway): preserve restart continuations after reboot · Fix Claude CLI session continuity after token refresh

Fix Telegram media file delivery · Fix cron chat delivery policy · fix: fix Telegram media file delivery (#69641) · fix(media): preserve outbound attachment filenames

Surface pending scope upgrades in gateway auth errors · Explain pairing scope upgrades during reconnects · Fix pairing-required recovery details · Split gateway probe capability from reachability

fix(gateway): preserve cron delivery validation semantics · fix(cron): preserve deferred heartbeat target override · fix(gateway): reject invalid cron announce delivery config · fix(cron): key recurring delivery dedupe to execution

fix(cron): keep message tool when delivery is none · fix(telegram): fence stale reply delivery after abort · fix: allow unknown properties in WakeParams schema (#68347) · Align documented bootstrap context defaults with runtime values

fix(telegram): clean up thread bindings to stale/failed ACP sessions on startup · fix(telegram): clear compaction replay after visible boundaries · fix(reply): preserve post-stream error payloads · fix(memory-core): emit one degraded sqlite-vec warning per manager

fix(stability): session skills snapshot, tool-loop guard, TUI watchdog, LM Studio preload backoff · fix(ollama): strip provider prefix from model ID in chat requests · feat(google): add Gemini TTS provider · Fix/cron announce no reply

fix(ollama): simplify cloud onboarding · fix: prune stale dist chunks after npm upgrades · fix: avoid running native pnpm binaries through node · fix(openrouter): handle reasoning_details in Qwen3 stream parsing

fix: cache external plugin catalog lookups in auto-enable · fix: forward optional params dropped at the runEmbeddedAttempt call site · fix(telegram): persist topic-name cache · fix: filter binary content from Telegram captions to prevent token explosion

feat(telegram): expose forum topic names in agent context

fix(telegram): bypass sequentializer for approval callback_queries · docs(cron): clarify day-of-month + day-of-week OR logic · doctor: warn about orphaned agent dirs · fix(agents): persist bootstrap marker after clean sessions_yield

fix(auth): stop forcing unsupported Codex OAuth scopes · fix(media): disable pinned DNS dispatcher for FormData transcription requests · fix: avoid qa scenario pack reads during packaged CLI startup · fix: scope pinDns override to multipart audio (#64766) (thanks @GodsBoy)

fix(subagents): deduplicate completion announce — prevent duplicate internal-context turns in parent session on retry · test: add Telegram QA E2E lane · fix(gateway): return real usage for OpenAI-compatible chat completions · fix(imessage): strip length-prefixed UTF-8 from imsg rpc text

fix(daemon): preserve safe env vars on gateway reinstall · fix(cron): repair nextRunAtMs=0 on non-schedule edits · fix(sessions): provider-qualified context limits (#62472) · fix(imessage): avoid DM self-chat false positives

make port optional for TLS manual connections · fix(android): auto-resume pairing approval · fix(build): prune stale bundled plugin node_modules · Agents: prefer runtime-resolved metadata for explicit codex gpt-5.4

fix(tts): use wav for Groq speech on OpenAI provider · Mistral: send reasoning_effort for mistral-small-latest (#62162) · Fix: HTTP 404 classification for model fallback chain (#62119) · Plugins: allowlist compat for capability provider fallback (#62205)

fix: remove heartbeat transcript pruning to prevent session corruption · fix: preserve explicit node routing under elevated auto exec · fix(exec): stop emitting tool updates after session is backgrounded · fix(gateway): resolve globally-installed bundled plugins in facade-runtime

fix(telegram): enable voice-note transcription in DMs · fix: implement Windows stale gateway process cleanup before restart · fix(infra): windows-task-restart fallback to startup entry when schtasks task is unregistered · fix: update embedded runner transport override

fix(daemon): preserve Windows Task Scheduler settings on reinstall and exit early on failed restart · fix(cron): suppress NO_REPLY sentinel in direct delivery path · fix(agents): keep tool_use and toolResult together when splitting messages · fix(agents): restore exec host=node routing and per-call override from auto

fix(android): smooth bootstrap approval wait · fix(pairing): mint tokens for merged device roles · fix(pairing): align mobile setup with secure endpoints · fix(channels): emit passive hooks for mention-skipped group messages

fix: normalize Kimi anthropic tool payloads · fix(agents): pin subagent gateway calls to admin scope to prevent scope-upgrade pairing failures · Android: auto-send assistant prompts · Android: add assistant entrypoints

fix(acpx): repair queue owner session recovery · fix: restore bundled runtime dependency provisioning · Docs: fix stale flows command references · fix(line): resolve dist runtime contract path

fix(telegram): allow RFC 2544 benchmark IPs in media download SSRF policy (#57452) · fix(xai): normalize image tool results for responses · Agents: fix subagent model precedence · WhatsApp: hydrate participating groups on connect

fix: add orphaned session key migration · fix(gateway): prevent OOM and socket hang up via Claim Check pattern for large attachments [media://inbound/<id> contract] · fix(hooks): default hooks.internal.enabled to true so bundled hooks load on fresh installs · fix(gateway): restore default operator scopes for pure HTTP token auth