The format is based on Keep a Changelog,

• Add context command support for chat turns (#1124)

• [slack]Allow operator-approved api base hosts (#1164)
• [web]Show dates for older sessions (#1162)
• [ci]Keep local validation on pinned tools

No notable changes.

• [slack]Challenge unknown allowlist DMs with OTP (#1163)

• [slack]Support configurable API base URL (#1159)

• [web]Support ACP-only chat setup (#1157)

No notable changes.

No notable changes.

• [web]Guard command palette input events

• [ci]Disable release biome cache

• [ci]Install biome from npm lockfile

• Add Kimi K3 provider support (#1156)

• Improve agent and sandbox status feedback (#1155)

• [web]Show direct mode when sandbox is unavailable (#1154)

• [external-agents]Auto-detect ACP agents (#1149)
• [providers]Add MiniMax M3 model support (#1151)

• [httpd]Remove unused hashmap import

• [sandbox]Harden apple container handling
• [e2e]Stabilize autoscroll indicator test
• Handle cron null fields and biome migration
• [cli]Support services without systemd (#1153)
• [providers]Derive context windows from capabilities (#1150)
• [providers]Derive openai-codex token expiry from JWT exp claim (#1152)
• [ci]Align biome version with repo config
• [e2e]Wait for onboarding summary readiness

• [ci]Align workflow rust nightly pin

• [e2e]Reset autoscroll mode per test

• [ci]Reclaim disk before e2e tests

• [ci]Run required jobs on hosted runners

• [copilot]Satisfy provider lint gates

• [providers]Enable provider features by default

• [copilot]Keep OAuth provider enabled for registry rebuild
• [copilot]Include provider in packaged builds
• [ci]Install codecov cli from pypi

• [ci]Clear profiled llama build caches

• [ci]Restore macos and fireworks release jobs

• Add GPT-5.6 model support (#1146)

• [mcp]Use direct fetch for resource_metadata URL from WWW-Authenticate (#1120)
• [caldav]Avoid panic on non-ASCII datetime in normalise_datetime (#1145)
• [gateway]Don't force-enable matrix-sdk via the metrics feature (#1139)
• [agents]Coerce stringified scalar tool args before validation (#1136)
• [browser]Tolerate null optional params in browser tool calls (#1098)
• [validation]Restore local release checks

• [whatsapp]Bump whatsapp-rust 0.5 -> 0.6 with LID-native addressing (#1144)

• Drop VOLUME declarations that shadow the home bind mount (#1122)

• [github-copilot]Improve enterprise diagnostics
• [github-copilot]Refresh auth after misdirected responses
• [github-copilot]Prefetch API endpoint metadata

No notable changes.

• [providers]Restore Gemini tool signature replay

• [providers]Default Alibaba Coding live test endpoint

• [providers]Harden live gate checks

No notable changes.

• Add NEAR AI Cloud provider (#1031)

• [providers]Split OpenAI Codex catalog
• [providers]Use explicit OpenAI capabilities (#1090)

• [providers]Propagate registry test errors

• [providers]Harden OpenAI-compatible request metadata

• [discord]Log silent voice message drops (#1081)

• [cron]Preserve host execution target (#1079)
• [web]Include clicked response in message forks (#1080)
• [providers]Strip MiniMax user names (#1078)
• [ci]Enforce provider release gates
• [ci]Isolate qmd install
• [skills]Track bundled skill disables individually (#1084)
• [skills]Propagate bundled skill test errors

No notable changes.

• [ci]Stabilize release checks

• [tools]Support nonblocking spawn agents (#1067)
• [agents]Support per-turn tool controls (#1069)

• [ci]Use valid release trigger condition
• Fix Docker build failures (#1073)
• [web]Stabilize command palette focus

• [agents]Support per-agent runtime limits (#1066)
• Agents as capability boundaries (MCP, sandbox, skills) (#1049)

• Remove file size allowlist
• [web]Remove token bar race assertion

• [web]Restore light mode syntax highlighting (#1047)
• [vault]Initialize existing password vaults (#1050)
• [gateway]Register config-declared hooks (#1048)
• [onboarding]Show configured LLMs in recommendations (#1058)
• [web]Prevent chat toolbar horizontal overflow (#1062)
• [sessions]Surface auto-title generation failures (#1064)
• [web]Show long model names in picker (#1060)
• [mcp]Hide stdio env values from status (#1063)
• [providers]Validate OpenAI-compatible endpoint URLs (#1061)
• [sandbox]Quiet image prebuild logs (#1065)

• [openai-codex]Add reasoning effort support (#1005)

• [gateway]Remove needless return

• [vault]Keep auth password changes in sync (#1026)
• [sandbox]Auto-detect docker host data mounts (#1035)
• [e2e]Wait for new chat sessions to settle
• [telephony]Dispatch Twilio gather speech (#1034)
• Fix sandbox media file reads in Docker (#1040)
• [gateway]Use mp3 for chat voice generation (#1041)
• [voice]Return wav metadata for piper audio (#1043)

• [web]Avoid false websocket disconnect timeouts (#1023)
• [sandbox]Reap docker sandbox zombies (#1025)
• [ci]Avoid slow session cleanup in e2e

• [ci]Drain fake claude stdin in coverage test
• [config]Preserve explicit defaults on startup (#1015)
• [agents]Honor BeforeLLMCall hook modifications (#1018)
• [providers]Parse thought reasoning tags (#1016)
• [agents]Dispatch before agent start hooks (#1017)
• [tools]Ignore heredoc bodies in dangerous command scan (#1019)
• [ci]Allowlist oversized files for decomposition
• [sandbox]Update slacrawl go module path (#1021)

• [ci]Stabilize external agent e2e binding

• [ci]Stabilize release e2e specs

• [tls]Support public IP SAN for generated certs (#1000)
• [mcp]Support OAuth client secrets (#1001)
• [remote-access]Add NetBird and Cloudflare Tunnel support (#1002)
• [external-agents]Add persistent agent sessions (#566)
• Add NetBird and Cloudflare Tunnel to onboarding (#1008)

• [providers]Retry mistral rate limits
• [ci]Stabilize e2e module shims
• [ci]Wait for autoscroll test position
• [sandbox]Update discrawl module path (#989)
• [release]Classify plain merge titles
• [ci]Repair autoscroll fixture overflow
• [providers]Normalize nullable tool arguments
• [providers]Harden live integration checks
• [install]Avoid Proxmox Docker prompt failure (#992)
• [ci]Stabilize token bar e2e assertion
• [install]Tolerate missing Proxmox CA cert (#997)
• [web]Prevent chat composer horizontal overflow (#998)
• [ci]Restore docs and app validation
• [ci]Stabilize chat autoscroll e2e
• [ci]Correct codex tracing skip parameter

• [release]Avoid large blob arguments
• [ci]Stabilize qmd and chat e2e checks
• [ci]Harden qmd and token bar tests
• [ci]Stabilize autoscroll overflow fixture
• [ci]Wait for chat clear before api dedupe check

• [voice]Add whisper-local STT provider, show all voice providers by default (#981)
• [voice]Surface OpenAI realtime model guidance (#984)

• [sandbox]Map home file tools to mounted path
• [agents]Preserve tool argument diagnostics (#983)
• [sandbox]Map apple container workspace paths
• [install]Keep proxmox updates on fork
• [ci]Update voice provider e2e expectations

• [nodes]Ed25519 challenge-response node identity (TOFU) (#979)
• [telephony]Add phone call support via Twilio (#920)

• [release]Remove gpg signing confirmation prompt

• [website]Hide unpublished releases from changelog
• [website]Publish 20260507.04 changelog
• [release]Avoid signing build tools
• [release]Read gpg signing confirmation from tty
• [release]Log gpg signing confirmation input
• [release]Compare gpg confirmation literally
• [browser]Resolve sandbox profile host mounts (#980)
• [release]Remove macos app cache from release workflow
• [ci]Stabilize e2e navigation and nodes setup

• [e2e]Align sandbox settings tests with tabs

• [sandbox]Clean up linux firecracker checks

• [sandbox]Compile firecracker backend with all features

• [sandbox]Remote & multi-backend sandbox support (Vercel, Daytona, Firecracker) (#942)

• [e2e]Consolidate sendRpcFromPage into shared helpers

• Remove unused sendRpcFromPage import from sessions.spec.js
• [ci]Drop e2e runtime state before cache save
• [e2e]Remove retry-hidden flakes
• [ci]Remove temporary websocket diagnostics

• [e2e]Revert tee log capture (breaks exec on CI), reduce RPC timeout to 5s
• [e2e]Reduce sendRpc retries from 40 to 10 — 40×5s=200s per test was hanging CI
• [gateway]Derive Default for ClientRegistryInner to satisfy clippy
• [gateway]Complete ClientRegistry migration — remove duplicate fields from GatewayInner
• [e2e]Add warmup RPC to waitForWsConnected to verify end-to-end WS
• [ws]Spawn RPC dispatch to prevent read-loop head-of-line blocking
• [gateway]Make chat() lock-free — move chat_override to std::sync::RwLock
• [ws]Increase sendRpc timeout from 5s to 30s for heavy RPCs
• [ci]Unblock chat send rpc
• [ci]Bound chat send prompt setup
• [ci]Avoid prompt waits on sandbox image builds
• [ci]Keep sandbox prebuilds off request paths
• [ci]Fail faster on e2e stalls
• [ci]Cache e2e build prerequisites
• [ci]Keep release e2e cache-free
• [ci]Cap e2e startup stalls
• [ci]Shard e2e across isolated gateways
• [ci]Isolate e2e workers per gateway
• [ci]Start one e2e gateway per project process
• [e2e]Refresh share notice and split settings specs
• [e2e]Avoid domcontentloaded navigation flake
• [ci]Stream e2e shard progress
• [ci]Keep e2e progress lines atomic
• [e2e]Make autoscroll fixture deterministic
• [e2e]Harden autoscroll setup against late renders
• [e2e]Use real voice settings save path
• [auth]Respect X-Forwarded-Proto for cookie Secure attribute (#970)
• [sandbox]Serialize container startup (#971)
• [providers]Replay DeepSeek reasoning content (#961)
• [matrix]Add debug logging for OIDC registration and deduplicate redirect normalization (#957)

No notable changes.

• [portable]Add data import/export for config, databases, and sessions (#951)
• [i18n]Add zh-TW Traditional Chinese locale support (#339)

• [discord]Register slash command arguments and add all thinking levels (#950)
• [web-ui]Prevent horizontal overflow in chat messages container (#952)
• [portable]Allow unwrap in test modules to satisfy workspace clippy lint
• [msteams]Collapse nested if to satisfy clippy collapsible_if lint
• [discord]Use is_none_or and allow expect in tests for clippy
• [e2e]Add smartScrollToBottom to chat-ui shim, fix command palette focus test
• [e2e]Reduce appended element height below isChatAtBottom threshold in autoscroll test
• [e2e]Add "Import Your Data" heading to auth spec onboarding matcher
• [i18n]Add missing zh-TW reasoning keys and fix biome formatting
• [telegram]Upgrade teloxide 0.13→0.17 to fix multipart ThreadId panic (#954)
• [terminal]Prevent spurious "window does not exist" error on tab creation (#955)
• [e2e]Update reasoning toggle tests for 6 effort levels (added Minimal and Extra High)
• [e2e]Use poll-based scroll assertion in sequential messages autoscroll test

• [providers]Add Zen (opencode.ai) multi-protocol provider (#944)
• [web-ui]Hide voice buttons when stt/tts disabled in config (#943)

• [web-ui]Add message action bar to assistant responses (#932)
• [gateway]Auto-generate session titles from conversation (#933)
• Add DeepInfra provider, sandbox GPU passthrough, strict model selection (#934)
• Detect Proxmox VE and offer LXC container install
• [skills]Add per-skill usage telemetry to /insights and web UI (#935)
• [sandbox]Add GitHub CLI (gh) to default sandbox packages
• [web-ui]Show progress indicator during sandbox image build

• Replace completion-based model probe with lightweight catalog check (#931)
• [deploy]Correct subcommand in systemd service file
• [web-ui]Clipboard copy button broken on insecure contexts (#936)
• [deploy]Use bare moltis command in systemd service
• [web-ui]System-notice text overflows excessively rounded container (#941)
• [httpd]Handle SIGTERM for graceful Docker shutdown (#940)
• [sandbox]Wait for image build before launching containers
• [e2e]Fix clipboard, provider order, and reasoning toggle tests
• [httpd]Gate msteams-only imports behind feature flag
• [e2e]Fix settings-nav, voice, onboarding, and command palette tests

• Add /btw, /fast, /insights, /steer, /queue commands and auxiliary model config (#926)

• [mcp]Preserve auth_state on expired OAuth token so re-auth button shows (#930)
• [web-ui]Remove scroll-hijacking ResizeObserver in chat (#922) (#925)

• [voice]Add voice personas for deterministic TTS identity (#916)

• [sandbox]Prevent sandbox escape via RestrictedHostSandbox and FailoverSandbox (#924)

• [update]Add /update command and web UI update button (#911)
• [browser]Add Obscura as lightweight sidecar browser backend (#869)
• [web]Add command palette (Cmd+K / Ctrl+K) (#904)
• [import]Add multi-source import from Claude Code, Claude Desktop, and Hermes (#914)
• [import]Add Claude Code and Hermes import to web UI (#917)

• [config]Unify provider name validation into single source of truth (#912)
• [agents]Split model.rs into chat, convert, and stream modules

• [e2e]Remove dead tests for deleted UI features
• Remove CLAUDE.local.md from repo and add to .gitignore
• Remove unused PathBuf imports in hermes-import
• Remove unused imports in httpd gateway

• [e2e]Stabilize rename test and skip broken Matrix ownership test
• [docs]Use $http_host in nginx proxy examples to preserve port (#907)
• [e2e]Fix rename assertion, unskip Matrix ownership test
• [gateway]Prevent USER.md from overriding saved user_name
• [e2e]Unskip clear-all and channel-rename tests, fix for new UI
• [release]Scope changelog entries to version deltas (#909)
• [e2e]Rewrite stop action test for thinking indicator UI
• [tests]Update comment preservation assertion for merge behavior
• [providers]Sanitize and strip user name field for channel messages (#915)
• [skills]Add missing origin metadata to data-sync, birdclaw, discrawl, wacrawl, slacrawl
• Collapse nested if-let in MCP import
• Collapse nested if-let in claude-import MCP servers
• Use vec![] macro instead of push-after-init in import commands
• [e2e]Use Ctrl+K instead of Meta+K in command palette tests
• Restore HashMap and ChannelPlugin imports in httpd gateway
• [web]Command palette keyboard navigation in headless Chromium
• [web]Use capture-phase listener for command palette keyboard nav
• [web]Eliminate rAF race in session rename input
• [web]Prevent session switch from stealing focus during rename
• [build]Align production feature flags
• [web]Show /mode none hint in slash command popup
• [web]Only guard chatInput focus against text inputs
• [e2e]Fix flaky command palette, reasoning toggle, and projects tests

• [e2e]Use .chat-session-name selector for rename assertion
• [ci]Copy scripts/ into Docker build and improve download retries

No notable changes.

• Make telegram channel optional (#891)
• Make discord and msteams channels optional (#899)

• [gateway]Split skills service helpers into separate module
• [gateway]Move skills impl out of mod.rs into service.rs

• [web-ui]Restore session name and inline rename in chat toolbar (#892)
• Use ApplicationType::Web for non-loopback Matrix OIDC redirect URIs (#893)
• Add TSX source to Tailwind content scanning
• Show RPC error messages in skills UI and auto-trust on install (#897)
• [config]Add mcp field to agent preset in schema map
• [e2e]Update tests for identity→profile route rename
• [e2e]Complete identity→profile rename in remaining tests
• [e2e]Stabilize session rename test against onBlur race
• [config]Use atomic writes to prevent moltis.toml corruption
• [config]Use unique temp file for atomic writes
• [config]Move tempfile from dev-dependencies to dependencies

• [web-ui]Show chat status badges in visible toolbar row (#886)
• [local-llm]On-demand model loading/unloading with idle timeout (#884)

• Remove needless return in skills toggle_bundled_skill

• [gateway]Handle bundled skill disable/enable via config (#877)
• [gateway]Derive bundled skill enabled state from config (#878)
• [tests]Increase watcher test timeout to accommodate debounce + FSEvents latency
• [web-ui]Stop auto-enabling all skills on repository import (#882)
• [skills]Store per-skill relative paths for marketplace repos (#883)
• [tests]Mark flaky FSEvents watcher test as ignored
• [tests]Use PollWatcher in watcher tests for reliable cross-platform behavior
• Use if-let instead of single-arm match in poll watcher test
• [security]Store voice API keys in credential store, not moltis.toml (#885)
• [e2e]Update session tests for inline action buttons (#886 follow-up)
• [tests]Add missing LocalLlmService trait methods to MockLocalLlm
• Collapse nested if-let in key_store timestamp check
• Derive Default for ModelLifecycleManager

• [ci]Add --timestamp and strip get-task-allow for macOS notarization

No notable changes.

• Resolve clippy lints for nightly-2026-04-24 and matrix-sdk compat
• [mcp]Prefer native MCP tools over mcporter (#874)
• [ci]Use keychain-profile for notarization and log failures

• Derive nightly toolchain from rust-toolchain.toml everywhere

• [ci]Use correct security list-keychains syntax for macOS codesign
• [ci]Align macOS certificate import with working arbor pattern

• [ci]Skip libnccl-dev to work around llama-cpp-sys-2 linking bug
• [ci]Remove pre-installed libnccl-dev from CUDA container
• [providers]Link NCCL when llama-cpp-sys-2 compiles with GGML_USE_NCCL

• [ci]Match NCCL dev headers to pre-installed runtime version

• [ci]Pin NCCL version to match CUDA 12.4 container

• [cron]Add heartbeat wake cooldown to prevent exec re-fire loop (#871)

• [config]Extract initialize_config() from discover_and_load()

• [tests]Stabilize flaky memory_config_get test

• [e2e]Stabilize flaky gemini, autoscroll, and settings-nav tests

• [ci]Allow held NCCL packages to be upgraded

• [ci]Add NCCL to build deps and fix flaky tests

• Replace expect() with let-else in splitter tests
• [benchmarks]Update code_index bench to use free-function chunker API

No notable changes.

• Replace unwrap() with expect() in splitter tests

• Remove needless Ok wrapper in refresh_qmd_index

No notable changes.

• [splitter]Extract moltis-splitter crate for AST-aware chunking (#791)

• [ci]Prevent releases.json from updating before GitHub release exists
• Revert releases.json to latest published release (20260421.05)

• [web]Add push-to-talk, VAD continuous listening, and voice settings (#303)
• [config]Layered config with defaults.toml and override-only user config (#864)
• [discord]Channel name pattern filtering and per-pattern overrides (#865)
• [config]Add default sub-agent presets (#844)
• [config]Add config compact command and auto-compact on startup
• [skills]MCP server management skill and post-install recipes (#840)

• [config]Split loader tests into submodules (file-size limit)
• [gateway]Extract NoopSkillsService into services/skills.rs (file-size limit)

• [config]Remove unused imports from test submodule

• [ci]Explicitly set draft: false on release upload to prevent orphaned drafts
• [agents]Add missing mode field to PromptRuntimeContext in tests
• [gateway]Collapse nested if for clippy
• [ctl]Clippy unnecessary_lazy_evaluations
• [graphql]Add missing SkillsService trait methods to MockSkills
• [gateway]Collapse nested if for clippy

• [home-assistant]Native Home Assistant integration crate (#827)
• [web-ui]Smart auto-scroll for chat messages (#846)
• [projects]Add code_index_enabled toggle to project settings (#837)

• [ui]Wire up project combo dropdown in chat header (#847)

• [web]Align signal channel E2E test with actual modal text
• [web]Stabilize flaky E2E tests
• Enforce correct date in silent memory turn daily log filenames (#859)
• [macos]Configure release signing and notarization (#842)
• [web]Add vault-sealed banner to main app (#839)
• [skills]Materialize bundled skill scripts to disk for execution (#861)
• [web]Biome formatting and i18n parity for projects page
• [config]Derive Default for HomeAssistantConfig and collapse nested if
• [config]Allow unwrap in code_index tests
• [home-assistant]Clippy collapsible-if and field-reassign-with-default
• [benchmarks]Add missing deps for code_index bench
• [code-index]Replace redundant closure with PathBuf::from
• Repair code-index compilation errors in gateway and benchmarks
• Resolve remaining clippy errors from code-index feature
• [providers]Strip null from enum arrays for Fireworks AI (#862)
• [web]Repair chat-autoscroll and projects E2E tests
• [benchmarks]Wire qmd feature to moltis-qmd dep for --all-features CI

• [signal]Add signal-cli channel (#841)

• [providers]Split openai_compat tests into submodules

• Remove set -e from codesign-debug recipe that breaks just test
• [web]Remove stale accountId assertion from signal channel E2E test

• [mcp]Show re-auth button when OAuth server needs re-authentication (#852)
• [sandbox]Skip sysfs tmpfs mounts for missing paths (ARM/Raspberry Pi) (#853)
• [prompt]Move datetime from system message to user content for KV cache stability (#855)
• [providers]Deep-merge properties in schema union collapse, strip redundant boolean enum (#856)
• [channels]Update ALL variant count for Signal channel type

• [ci]Move release test job to self-hosted runner
• [skills]Repair embedded bundled skill discovery in release builds

• [skills]Add bundled skill category management to onboarding and settings (#829)

• [tools]Block exec approval bypass via env-var prefix injection (#822)
• [browser]Add diagnostic logging for container readiness failures (#820)
• [gateway]Downgrade broadcast log from debug to trace (#830)
• [chat]Correct push notification click-through URL (#831)
• [tools]Skip sysfs tmpfs mounts on WSL2 (#835)
• [chat]Preserve Gemini tool call metadata (#836)
• [providers]Normalize non-strict tool schema unions (#833)
• [providers]Apply Kimi router overrides in Fireworks integration tests (#832)
• [config]Resolve ${VAR} placeholders against [env] section and DB env vars (#834)
• [web]Repair gemini tool-signature and reasoning-toggle E2E tests

• [code-index]Use discover_opts with permissive trust for gix
• [web]Render inline markdown in table cells
• [code-index]Set required_trust to Reduced for gix discovery
• [code-index]Use struct init for clippy field_reassign_with_default
• [code-index]Use open_opts with forced Trust::Full for gix
• [code-index]Skip git-dependent tests when .git is absent
• [telegram]Infer MIME from filename when Telegram sends octet-stream (#819)
• [ci]Install git before checkout in CUDA containers

• [ci]Make safe.directory step non-fatal when git is absent

• [ci]Restore original step order, fix gix open, revert npm flag

• [ci]Gix safe.directory, read_ops file_path regression, npm --ignore-scripts
• [tools]Empty file_path falls through to read_primary, whitespace rejected

• [channels]Centralized command registry for all channels (#794)
• Gemini thought_signature round-tripping and schema validation fixes (#795)
• Self-improving agent loop — skills, memory lifecycle, deployment (#803)
• [web]Add GitHub Issues and Discussions links with count badges (#806)
• [sandbox]Install Node.js 22 LTS via NodeSource (#807)
• [web]Render markdown as HTML in chat messages (#808)
• [skills]Bundle 101 default skills with category UI and format fallback (#797)

• [tools]Split skill_tools.rs into submodules

• [ci]Add safe.directory for container jobs (gix ownership check)
• [slack]Strip leading slash from commands before gateway dispatch (#804)
• [mcp]Implement legacy SSE transport for endpoint discovery (#805)
• [ci]Biome formatting and file size allowlist
• [sandbox]Verify image in Podman store after BuildKit build (#811)
• [providers]Disable strict tools for Fireworks Kimi router (#812)
• [ci]Npm --ignore-scripts and git install ordering
• [ci]Pass safe.directory to gix via env vars in rust-ci container

• [ci]Address main rust failures

• [ci]Handle .ts locale files in i18n-check on Node <22

• [code-index]Add code indexing crate with builtin SQLite+FTS5 backend (#771)
• [config]Add server.external_url for reverse proxy WebAuthn (#785)

• Add thiserror Error types to 8 library crates (#792)

• [task_list]List all tasks by default and add list_lists action (#779)
• [sandbox]Add observability and prevent repeated package provisioning (#784)

• [nostr]Add NIP-59 Gift Wrap support for private DMs (#763)
• [matrix]Add OIDC authentication via Matrix Authentication Service (#730)

• [tests]Remove hardcoded secret fixtures (#768)

• [providers]Strip $schema recursively and downgrade fallback log level (#762)
• [matrix]Retry sync loop on transient connection failures (#761)
• [e2e]Stabilize archived-sessions and nostr channel tests
• [sandbox]Skip sysfs tmpfs overlays on Podman (#765)
• [slack]Register /commands HTTP endpoint for slash commands (#767)
• [matrix]Restore main CI after OIDC changes

• [web]Add reasoning effort toggle to chat toolbar (#750)

• [providers]Sanitize MCP tool schemas regardless of JSON Schema draft (#746)
• [node-host]Install rustls CryptoProvider before wss:// connections (#749)
• [providers]Prune orphaned required entries from tool schemas (#751)
• [web]Satisfy biome and format gates
• [testing]Avoid CUDA features in Darwin test recipes
• [web]Stabilize session delete flow
• [ci]Harden linux dependency installer

• [providers]Configurable context windows + oldest-first compaction (#737)

• [tests]Split oversized Rust test modules

• [ci]Repair MCP e2e and iOS chat event handling
• [providers]Detect Grok 3/4 as reasoning-capable models (#741)
• [nostr]Replace tokio RwLock with std RwLock to prevent panic (#742)
• [providers]Restore type annotations stripped by schema canonicalization (#740)
• [ci]Retry Linux dependency installs and align Darwin validation
• [ci]Install openssh client in Linux build container

• Allow GUIDELINES.md file override for hardcoded tool guidelines
• [channels]Add WhatsApp to default offered channels
• [whatsapp]Auto-approve owner, set device name, wrap buttons
• [whatsapp]Handle PairSuccess event for instant QR dismissal
• [whatsapp]Make Account ID optional, default to "main"
• [web]Add Slack channel to onboarding wizard

• Clean up provider tests and onboarding UI
• Move implementation code out of mod.rs and lib.rs (#731)

• [web]Remove duplicate assets warn import

• [graphql]Make sessionKey required for all chat operations
• [graphql]Address PR review — harden subscription filter and tests
• [onboarding]Show version in server footer
• [web]Prefer embedded assets over stale share dir
• [chat]Preserve sender names in user messages
• [providers]Address sender-name review feedback
• Move return inside empty-guard in append_guidelines_section (PR 714 review)
• [providers]Skip strict tool schemas for Google/Gemini via OpenRouter (#716)
• [providers]Also detect googleapis.com URLs for non-strict tools
• [providers]Collapse type arrays in strict mode patching (#716)
• [whatsapp]Upgrade whatsapp-rust ecosystem 0.2 → 0.5
• [whatsapp]Use time crate for timestamps, fix DashMap sync key ordering
• [whatsapp]Add QR code polling fallback for pairing
• [whatsapp]Skip duplicate account start, add QR polling e2e test
• [whatsapp]Use res.payload not res.result for QR polling
• [whatsapp]Render QR code as SVG in polling response
• [whatsapp]Skip history sync on pairing
• [whatsapp]Detect pairing completion via polling
• [channels]Add missing useRef import in page-channels.js
• [whatsapp]Set OS name to Moltis, add no-sync hints
• [whatsapp]Fix mangled no-sync hints in onboarding view
• [whatsapp]Auto-approve both PN and LID JIDs, add debug logging
• [whatsapp]Fix QR code not displaying when Account ID is empty
• [whatsapp]Instant pairing feedback and owner auto-approve
• [whatsapp]Subscribe to channel events in onboarding WS
• [whatsapp]Only auto-approve owner PN JID, not LID
• [whatsapp]Use PN JID for self-chat replies instead of LID
• [whatsapp]Show phone number in self-chat footer instead of LID
• [channels]Don't drop channel_user WS broadcast
• [whatsapp]Use bare phone number as self-chat session ID
• [agents]Strip trailing stop tokens leaked by LLMs as content
• [agents]Address PR review comments
• [providers]Repair openai compat CI regressions (#722)
• [agents]Collapse nested if per clippy collapsible_if lint
• [web]Restore setup onboarding alias
• [providers]Add null to enum arrays for nullable optional props (#724)
• [mcp]Treat any HTTP response as alive in health check (#732) (#733)
• [chat]Broadcast user_message event so API-sent messages appear in web UI (#734)

• [slack]Bump slack-morphism to 2.20 to enable TLS for socket mode
• Regenerate Cargo.lock for reqwest 0.13 and add query feature
• [web]Restore clear button for main session modal (#671)

• [node-exec]Merge node exec types into owners
• [web]Remove prompt memory toolbar from chat header

• [web]Harden flaky e2e tests for cron delete and identity autosave
• [sandbox]Mask /proc and /sys host metadata in Docker/Podman containers
• [sandbox]Address PR review — strengthen test assertions
• [channels]Use rand::RngExt for random_range after rand 0.10 upgrade
• Update rand 0.10 imports across workspace
• [onboarding]Surface local-llm and LM Studio in recommended providers
• [auth]Update password minimum length strings from 8 to 12 characters

• [web]Gate openai live e2e server on api key

• [voice]Add local whisper web setup
• [chat]Show cached input tokens

• [providers]Normalize openai schemas with schema crates
• [providers]Move schema normalization out of mod.rs
• [chat]Unify usage propagation paths

• [httpd]Remove unused route import
• [httpd]Drop unused get import

• [providers]Sanitize openai tool schemas
• [providers]Address PR review feedback
• [providers]Keep openai compat module under file limit
• [providers]Avoid expect in schema tests
• [web]Restore nostr channel icons
• [voice]Address PR review feedback
• [chat]Satisfy clippy in cached token tests
• [agents]Preserve streamed cache token usage
• [openai]Add live e2e coverage and schema guard
• [sessions]Wire archived sessions through gateway and web ui
• [sessions]Address archived review feedback
• [sessions]Allow archiving non-current channel chats
• [sessions]Preserve orphaned search hits
• [sessions]Allow unarchiving active channel chats

• Split oversized rust modules
• Split remaining oversized rust modules
• Remove replaced module entrypoints

• [agents]Remove duplicate tool arg match arm

• [providers]Preserve native tool arg types
• [gateway]Repair split module blockers
• Restore post-merge validation
• Repair local lint regressions
• [chat]Clean up split validation fallout
• [gateway]Remove split warning debt
• [httpd]Restore split module boundaries
• [httpd]Drop stale server re-export
• [httpd]Repair test and bridge lint fallout
• [httpd]Restore vault test imports
• [telegram]Mark byte-truncated documents
• [httpd]Repair ngrok server split
• [provider-setup]Restore ollama validation payload
• [gateway]Drop duplicate broadcaster default impl
• [gateway]Remove stale mcp service doc comment
• [httpd]Move runtime-only route import
• [tools]Scope apple sandbox helpers to macos
• [tools]Restore sandbox test imports
• [ci]Restore main build stability

• [gateway]Resolve release CI regressions

• [voice]Add base_url config for OpenAI TTS and Whisper STT providers

• [browser]Type browserless API version as enum
• [nodes]Extract core types and constants to dedicated crate
• [nodes]Move pure ssh and env helper functions to node-exec-types
• [mcp]Extract config parsing functions to moltis-mcp
• [mcp]Extract MCP-agent tool bridge to dedicated crate
• Group lock-free broadcast state into Broadcaster struct
• [tools]Unify cron schema field helper

• [telegram]Persist inbound documents
• [review]Make inbound document replay robust
• [browser]Add Browserless v2 websocket fallback support
• [browser]Change log level from warn to debug for sandboxed browser websocket connection failures
• Complete MCP bridge extraction and address Greptile review
• [broadcast]Make seq private, add #[must_use] to next_seq, add tests
• [chat]Serialize data dir override tests
• [voice]Address PR review feedback
• [tools]Hide wasm internals and relax cron schema
• [tools]Simplify cron schema unions
• [agents]Preserve legacy tool-call compatibility

• [nodes]Define NodeInfoProvider trait for decoupling

No notable changes.

• [discord]Handle inbound voice and image attachments
• [hooks]Include channel provenance in payloads
• [web]Add Projects section to Settings sidebar navigation
• [tools]Native filesystem tools (Read, Write, Edit, MultiEdit, Glob, Grep)
• [tools]Typed error taxonomy for Read (not_found / permission_denied / too_large / not_regular_file)
• [tools]Phase 1 polish (byte cap, session key, fs-tools feature, contract tests)
• [tools]CRLF-tolerant Edit recovery for fs tools
• [tools]Per-session FsState with must-read-before-write + re-read loop detection
• [tools][tools.fs] config + path allow/deny policy (phase 4)
• [tools]Checkpoint_before_mutation + binary base64 + respect_gitignore + docs page
• [tools]Phase 2 sandbox bridge for Read/Write/Edit/MultiEdit/Glob
• [tools]Phase 2b — Grep sandbox routing
• [tools]Phase 3c — adaptive Read paging coupled to context window
• [chat]Deterministic compaction with budget discipline
• [chat]Pluggable compaction modes with config + docs
• [chat]Implement recency_preserving compaction mode
• [chat]Implement structured compaction mode
• [chat]Surface mode + token usage in compaction broadcasts
• [chat,web]Surface compaction mode + tokens in UI and channels
• [compaction]Add chat.compaction.show_settings_hint opt-out
• [tools]Claude Code compat — BOM strip, binary extensions, smart quotes, mtime tracking
• [tools]PDF text extraction in Read tool
• [tools]Image dispatch in Read + Grep context alias + Edit param aliases
• [memory]Add prompt memory styles
• [memory]Extend config surfaces
• [provider-setup]Remove automatic model probe, add manual Test button
• [tools]Auto-page reads and serialize fs mutations
• [tools]Expose read continuation offsets
• [tools]Surface sandbox scan truncation
• [tools]Wire layered tool policy into runtime with per-provider, per-agent, per-channel, and per-sender support
• [tools]Add sandbox tools_policy as layer 6 in policy resolution
• [channels]Add Nostr DM channel support
• [channels]Add Nostr web UI, E2E tests, and documentation
• [channels]Add Nostr to onboarding flow
• [nostr]Add metrics, NIP-44 decryption, and integration tests
• [website]Add slack, matrix, and nostr channels
• [website]Add Nostr channel SVG icon
• [chat]Add summary budget discipline for compaction
• [website]Add community quote from discussion #680 and make quotes horizontally scrollable
• [web]Add option to disable terminal in Web UI
• [auth]Add brute-force protection with IP ban and account lockout

• [chat]Align multimodal rewrite updates
• [telegram]Extract STT setup hint constant
• [discord]Reuse inbound downloader per handler
• [chat]Simplify deterministic compaction module
• [chat]Split compaction_run into per-strategy submodules
• [tools]Use ripgrep crates for fs grep
• [tools]Split read.rs into read/ module with pdf.rs and image.rs
• [tools]Centralize sandbox filesystem access
• [media]Sniff mime and expose sandbox file ops
• [tools]Add native host sandbox file ops
• [tools]Add container-aware fs transports
• [tools]Stream OCI file transfers
• [tools]Stream OCI reads from cp

• [agents]Dispatch ToolResultPersist hooks
• [agents]Sanitize ToolResultPersist tool names
• [agents]Handle Z.AI text tool calls and dedupe providers
• [gateway]Address PR comment follow-ups
• [hooks]Honor MessageReceived actions
• [chat]Harden MessageReceived hook handling
• [telegram]Avoid placeholder voice fallbacks
• [telegram]Preserve caption when STT unavailable
• [discord]Address review feedback
• [discord]Log stt-unavailable voice notes
• [common]Block mapped ipv6 ssrf bypass
• [hooks]Address review feedback
• [chat]Warn on invalid channel bindings
• [hooks]Address remaining provenance feedback
• [agents]Warn on invalid hook channel context
• [plugins]Refresh hook docs and logger timestamps
• [common]Update message received fixtures
• [web]Store container ref in teardownProjects, remove unused settings_projects route
• [tools]Fs tools require absolute paths, add workspace_root for Glob/Grep
• [agents]Warn on tool name collision in ToolRegistry::register*
• [compaction]Address code review issues #1-7
• [compaction]Address Greptile PR #653 review comments
• [chat]Correct recent_messages_preserved flag in compaction
• [compaction]Strip preamble and directive from re-compaction extraction
• [compaction]Invert candidate order so bullets outlive plain lines
• [compaction]Protect summary tags from budget-pressure dropping
• [compaction]Address PR #653 review findings
• [chat]Memory-file summary lookup across all compaction modes
• [compaction]Extract_summary_body picks newest summary on iterative re-compaction
• [chat]Wire chat.compaction.threshold_percent into auto-compact trigger
• [compaction]Address Greptile round-2 review on #653 (P1 + P2)
• [web]Track compacting status message per-session to avoid removing compact card
• [compaction]Restore 0.95 default threshold + remove dead code
• [chat]Close compact/store race and tag auto_compact broadcast paths
• [compaction]Address Greptile round-5 P2 findings on #653
• [tools]Enforce exec allowlist when approval_mode is off
• [tools]Deny dangerous commands in off mode instead of hanging
• [tools]Warn when safe-bin bypasses explicit allowlist in off mode
• [config]List tools.policy.profile in preset silent-policy warning
• [tools]Address filesystem review comments
• [tools]Address Greptile round-2 findings (max_read_bytes wiring, sandbox note_fs_mutation, truncated semantics)
• [tools]Address Greptile round-3 findings (Grep policy, must-read sandbox, Glob root deny)
• [tools]Address Greptile round-4 findings (binary read tracking, sandbox Write new-file)
• [tools]Sandbox Grep post-filters results through path policy
• [tools]PDF/image dispatch now enforces path policy, sandbox guard, and FsState recording
• [memory]Tighten qmd and rpc validation
• [chat]Honor [skills] enabled=false at runtime
• [chat]Replace test Mutex<()> with Semaphore
• [chat]Reuse existing config in context skill discovery
• [httpd]Redirect remote setup traffic to onboarding wizard
• [agents]Detect and break tool-call reflex loops (#658)
• [agents]Address Greptile review feedback on #658
• [agents]Loop detector handles mixed-outcome batches correctly (#658)
• [agents]Treat success=false without error field as failure (#658)
• [e2e]Wait for Preact render flush in matrix senders test
• [web]Show Clear button for main session in modal
• [memory]Add missing runtime module
• [web]Remove unused VALIDATION_HINT_RUNNING_TEXT, clear stale test results
• [gateway]Unify config-override test lock to prevent flaky test
• [tools]Smart-quote recovery preserves file content + sandbox grep uses PCRE
• [tools]Scope tar helper to tests
• [tools]Preserve loop warnings for auto-paged reads
• [chat]Populate sender_id in channel binding and runtime context
• [chat]Read sandbox state from runtime context in PolicyContext
• [agents]Add missing channel_sender_id field in test
• [tools]Use struct init instead of field reassign in test
• [tools]Expand profile field in provider, sender, and sandbox policy layers
• [httpd]Start stored channels on vault unseal
• [httpd]Add missing continue in unsupported channel type guard
• [config]Revert unintended matrix named field promotion
• [nostr]Fix integration test DM round-trip reliability
• [nostr]Prevent panic on UTF-8 boundary when truncating large DMs
• [website]Use official Nostr protocol logo (CC0, mbarulli/nostr-logo)
• [web]Preserve Nostr OTP settings on edit modal save
• [nostr]Resolve clippy collapsible_match and len_without_is_empty
• [nostr]Implement OTP challenge initiation and DM delivery
• [nostr]Implement OTP verification path for code replies
• [nostr]Use std::sync::RwLock for accounts map to avoid blocking panic
• [web]Remove channel-error class from conditionally-rendered error divs
• [chat]Correct budget accounting bugs in compress_summary
• [chat]Address greptile review feedback (greploop iteration 1)
• [channels]Finish discussion 425 follow-ups
• [channels]Address review feedback
• [channels]Harden channel command authorization and session scoping
• [gateway]Truncate approval previews safely
• [channels]Chunk unicode safely
• [chat]Isolate sqlite memory tests
• [web]Stabilize mocked channel refresh
• [channels]Address greptile review feedback
• [channels]Truncate command in approve/deny confirmation messages
• [auth]Second-pass security hardening
• [web]Address PR review comments
• [config]Preserve TOML section order on web UI save
• Apply local fixes

• [config]Warn when preset tool policies are set but tools.policy is empty
• [nostr]Address PR review comments
• [auth]Harden remote access with 9 security improvements

• [oauth]Log loopback redirect URI rewrites at debug level
• [skills]Ship native read_skill tool
• [skills]Harden read_skill with assets/, binary files, and metadata surfacing

• [oauth]Share loopback redirect normalizer and apply to provider setup
• [oauth]Eliminate dead branches in normalize_loopback_redirect

• [web]Stabilize node selector and fork delete e2e

• [gateway]Dcg-guard PATH augmentation and loud missing-dcg warning
• [gateway]Refresh stale dcg-guard files and use async subprocess
• [gateway]Dcg-guard HOME fallback and unconditional startup log
• [agents]Suppress auto-continue after substantive final answer
• [agents]Address review feedback on auto-continue fix
• [mcp]Normalize loopback redirect URIs to http for OAuth registration
• [provider-setup]Normalize pre-loaded loopback redirect URIs
• [skills]Address Greptile review feedback on read_skill
• [skills]Address greptile review feedback (greploop iteration 2)
• [skills]Cap skill body size in read_primary (greploop iteration 3)
• [skills]Per-subdir sidecar cap + data_dir-scoped discoverer
• [voice]Honor whisper.model and whisper.language in STT factory

• [hooks]Pin dcg install to tag and verify checksum

• [providers]Add Alibaba Cloud Coding Plan provider

• [ci]Avoid dynamic provider secrets

• Remove redundant http client fallback check

• [common]Ensure User-Agent survives all HTTP client fallback paths
• [test]Address PR review — clarify comment and assert no Error events
• [crons]Persist schedule field values across modal re-renders
• [crons]Read schedule fields as snake_case from server responses
• [providers]Deliver MiniMax system prompt via first user message
• [tests]Use streaming probe for OpenAI integration tests
• [providers]Bump GPT-5 probe output cap to 16 tokens
• [tests]Use Secret<T> for API keys in model discovery tests
• [providers]Handle multimodal content in MiniMax system prompt rewrite

• [msteams]Comprehensive Teams channel implementation
• [providers]Add Gemini 3.x models to catalog and update capability detection
• [providers]Add ModelCapabilities struct to ModelInfo and DiscoveredModel
• [chat]Use ModelCapabilities in API responses instead of provider lookups

• [web]Integrate Tailscale Funnel into Teams channel setup
• [web]Remove 'Requires public URL' badge from Teams card
• [web]Simplify Teams onboarding now that Remote Access step exists
• [msteams]Address PR review feedback
• [msteams]Use Graph token for reactions and thread context
• [msteams]Prevent streaming retry storm and URL injection in search
• [web]Disambiguate OAuth E2E selector for model picker
• [web]Keep new chats at top of sidebar
• Auto-allow direnv in superset worktree setup
• Load BOOT.md per-session via system prompt instead of broken hook (#594)
• Remove stale boot-md assertion from discover_hooks test
• Update model count assertion for gemini-3 reasoning variants
• [installer]Address PR review feedback
• [browser]Include Podman in container availability check
• [browser]Update stale comment to include Podman
• [gateway]Narrow skill and memory watch roots
• [gateway]Avoid blocking skill watcher refresh
• [providers]Extract MiniMax system messages to top-level field
• [providers]Warn on non-string MiniMax system message content
• [providers]Add PartialEq/Eq to ModelCapabilities, use infer() in tests
• [providers]Narrow gemma exclusion to gemma-3n- only
• [agents]Surface workspace prompt truncation
• [gateway]Address review feedback
• [gateway]Match workspace prompt normalization
• [agents]Make workspace file truncation limit configurable (#593)
• Replace magic constant with ChatConfig::default() fallback
• Eliminate double chars().count() and add zero-value validation
• Report truncation when max_chars is zero
• Remove duplicate workspace_file_max_chars key in schema map
• [tools]Wire ExecConfig timeout and max_output_bytes to ExecTool
• [tools]Make timeout schema description reflect configured default
• [providers]Resolve 404 when selecting Ollama model in web UI
• [web]Preserve ollama pull hint in humanizeProbeError
• [providers]Forward auth header in Ollama native probe fallback
• Align auth middleware tests with gateway state

• [matrix]Add slash command support
• [models]Make model detection opt-in and add stop button

• [matrix]Match help command by exact name, not prefix
• [models]Abort probe tasks on cancel, show feedback, await RPC
• [tls]Include lan bind SANs in auto-generated certs
• [tls]Address PR review feedback
• [agents]Use system message for auto-continue nudge instead of user message
• [common]Add default User-Agent header to shared HTTP client
• [common]Use MOLTIS_VERSION for default user-agent and apply headers in apply_proxy
• [agents]Keep auto-continue nudge as user message
• [provider-setup]Include lmstudio in known_providers and replace ollama name checks
• [provider-setup]Add dedicated local_only field to KnownProvider
• Harden superset setup envrc handling

• [agents]Auto-continue when model stops mid-task + max iterations UX
• [config]Make auto-continue tool-call threshold configurable

• Address PR review — translatable continue message, document tool-call threshold
• Guard auto-continue against min_tool_calls=0 usize tautology
• [minimax]Restore system prompts and null tool args
• [providers]Discover live anthropic models
• [providers]Mark anthropic recommendations globally

• [webhooks]Add generic webhook ingress for triggering AI agents
• [web]Link to Hoppscotch for webhook testing
• [web]Add CORS to webhook ingress and copy-curl button
• [website]Add Webhooks to landing page features

• [web]Extract webhooks nav icon to external SVG file

• [cli]Report release version in --version output
• [providers]Propagate cache tokens in Responses API and custom providers
• [providers]Read cached_tokens from input_tokens_details in non-streaming Responses SSE
• [agents]Match provider-specific context window error strings
• [chat]Honor public sessionKey in GraphQL flows
• [chat]Address PR review — safer mock assertions and precedence test
• [web]Use globe icon for webhooks settings nav
• [web]Use ModelSelect and ComboSelect for webhook agent/model fields
• [web]Use public URL (ngrok/tailscale) for webhook endpoint display
• Improve webhook test script with verbose output and TLS support
• Use OnceLock for webhook state fields instead of Arc::get_mut
• [web]Show Hoppscotch link below webhook list, not only in empty state
• [web]Match webhooks layout to cron/sandbox and fix nav icon
• [web]Add missing space before Hoppscotch link
• [web]Constrain webhooks list width with max-w-form
• [web]Revert max-w-form, keep webhooks list full width
• [web]Add webhooks icon to settings nav via components.css
• [web]Recommend curl and Hoppscotch desktop for webhook testing
• Use HeaderValue::from_static for CORS headers, remove warning
• [web]Improve webhook test command button and footer text
• Don't dedup generic webhooks by body hash, rebuild i18n
• Address PR review comments
• [security]Redact auth secrets from webhook API responses
• Drain unprocessed webhook deliveries on worker startup
• Include 'processing' deliveries in crash-recovery drain
• Gitlab_token auth config key mismatch, add regression tests
• [security]Fail closed on non-parseable Stripe signature timestamp
• Enable foreign_keys pragma and explicit cascade delete
• [ci]Restore CHANGELOG.md to main state (changelog guard)
• Forward agent_id to chat.send_sync in webhook worker
• Enforce CIDR allowlist, set foreign_keys on pool options
• [security]Gate forwarded headers on behind_proxy, fix PagerDuty multi-sig
• Use ConnectInfo for direct IP, disable source_profile on edit
• [webhooks]Harden webhook execution and secrets
• Proactive audit — 6 issues found and fixed
• Wrap cascade deletes in transactions to prevent partial data loss
• Resolve settings nav CI regression

• Add webhooks feature documentation

• [openclaw-import]Convert non-default agents to spawn_agent presets

• [web]Allow session sidebar links to open in new tabs
• [web]Tighten session sidebar link accessibility
• [docker]Add missing default features to Dockerfile build
• [docker]Use default features instead of explicit list

• [website]Add provider/channel pills section and update branding
• [website]Add positioning, how-it-works, use cases, and community quote

• [providers]Avoid quadratic SSE buffer copies
• [providers]Align copilot stream error handling

• [providers]Route Copilot enterprise tokens via proxy endpoint (#352)
• [providers]Address PR review comments on Copilot enterprise
• [providers]Harden Copilot enterprise proxy security
• [providers]Reject bare IP addresses in Copilot proxy-ep
• [providers]Address Copilot enterprise review feedback
• [providers]Stream enterprise copilot responses
• [website]Crop MiniMax icon, grayscale raster icons, official GraphQL logo
• [website]Add Discord source links to community quote
• [website]Update LoC stats on security page
• [website]Update LoC stats in all locale files
• [website]Regenerate all locale files with new homepage sections
• [website]Update i18n titles and regenerate all locale files
• [website]Translate new homepage sections in all 9 locale files
• [website]Translate remaining English strings in all locale files
• [website]Localize injected nav tabs
• [website]Address greptile review feedback
• [website]Sync i18n builder with locale pages
• [website]Correct i18n generator keys

• [providers]Redact CopilotTokenResponse token in Debug output

• [web]Restore all-features build

• [web]Map config reload errors explicitly

• [cron]Auto-clean orphaned sessions and prune sandbox containers

• [swift-bridge]Await embedded httpd shutdown
• [cron]Use time crate for retention math and fix named-session guard
• [sandbox]Include remove_image_override in cleanup_session
• [cron]Skip pruning cycle when session key lookup fails
• [web]Reload offered channels from config
• [ci]Await channel preload in settings e2e

• Add GitHub artifact attestations to release workflow

No notable changes.

No notable changes.

• [web]Restore matrix onboarding icon

• [providers]Add zai-code provider for Z.AI Coding plan
• [tools]Add cross-session search recall
• [tools]Add automatic edit checkpoints
• [projects]Harden context loading
• [skills]Add portable bundle quarantine flow
• [exec]Add ssh remote routing
• [web]Add skills bundle ui and ssh target visibility
• [web]Clarify ssh execution targets
• [ssh]Add managed deploy keys and targets
• [nodes]Add remote exec doctor panel
• [ssh]Harden managed targets and host pinning
• [nodes]Repair active ssh host pins from doctor
• [ssh]Add actionable runtime failure hints
• [web]Add tools overview to settings
• [web]Allow renaming channel-bound sessions
• [security]Add GPG signing for release artifacts
• [security]Add release verification script
• [gateway]Add channel settings agent tool
• [providers]Collapse model lists, hide legacy models, add recommended flag
• [web]Add live remote access settings
• [remote]Improve onboarding public access flow
• [tools]Add Firecrawl integration for web scraping and search
• [web]Add recommended provider tier in onboarding and docs guide
• [config]Add upstream_proxy for application-level HTTP proxy support
• [channels]Add Matrix channel integration
• [matrix]Complete channel parity and web ui coverage
• [matrix]Add encrypted chat support and vault-backed channel secrets
• [matrix]Add account ownership mode
• [matrix]Harden ownership recovery flow
• [matrix]Add generic channel location fallback

• [ssh]Use secrecy for imported key material
• Replace vendored sqlx-sqlite with git dependency

• [web]Remove unused gon import

• [providers]Address PR review comments for zai-code
• [vault]Allow unencrypted session history while sealed
• [vault]Address PR review comments
• [security]Address PR review feedback
• [ssh]Tighten timeout and warning handling
• [ssh]Address latest review follow-ups
• [gateway]Collapse legacy ssh node lookup
• [auth]Guard ssh key deletion race
• [httpd]Satisfy ssh route lint
• [ssh]Reject option-like targets
• [ssh]Hide import passphrases from argv
• [ssh]Quote known hosts path
• [web]Use browser location port for node join URL
• [web]Guard e2e assertion for default-port case
• [e2e]Use sidebar selector for sealed-vault session visibility test
• [web]Fall back to getRandomValues for session UUID on plain HTTP
• [web]Address PR review feedback
• [providers]Speed up model probes
• [security]Address PR review feedback for GPG signing
• [security]Prevent gpg --import grep from aborting verify script
• [security]Show GPG signer identity and failure diagnostics
• [security]Pin GPG key fingerprint to prevent TOFU attacks
• [voice]Surface elevenlabs stt failures
• [gateway]Address channel settings PR feedback
• [voice]Handle empty stt transcripts
• [web]Show unsupported model reason inline instead of tooltip-only
• [web]Show probe error inline in model selector cards
• [web]Show probe error inline in preferred models selector
• [web]Preserve server error message for model probes
• [web]Sort and collapse onboarding model selector like settings
• [web]Sort models by version number when no date available
• [httpd]Harden ngrok controller lifecycle
• [httpd]Retain ngrok controller after startup
• [ngrok]Harden loopback tunnel handling
• [ngrok]Avoid fatal startup on tunnel errors
• [ngrok]Clarify defaults and warnings
• Update local setup and ElevenLabs error logging
• [web]Localize toggle button in onboarding and fix JSDoc comment
• [chat]Use relative timestamps in created_at test
• [tools]Address firecrawl PR review feedback
• [tools]Resolve firecrawl web_search registration and timeout race
• [agents]Stabilize prompt cache and compact tool results
• [web]Stabilize send-document e2e test
• [mcp]Address PR review comments for streamable HTTP transport
• [mcp]Update template docs and log messages for streamable HTTP
• [providers]Default to vision support for unknown models (#556)
• [providers]Also exempt gpt-4-vision from denylist
• [providers]Surface real error on provider probe failure
• [web]Apply serverMessage pattern to validateProviderConnection
• [web]Allow multi-model selection during provider setup
• [web]Scope Select All to visible models, check save_models response
• [config]Wrap upstream_proxy in Secret<String> and redact credentials in logs
• [proxy]Use rfind for @ redaction, warn on parse failure, document Slack gap
• [providers]Rediscover models from /v1/models before probing
• [providers]Address PR review — move Ollama probes outside lock, use runtime env
• [providers]Check total model count in RediscoveryResult::is_empty
• [matrix]Address review feedback
• [matrix]Address latest review feedback
• [matrix]Set reply thread ids after main merge
• [matrix]Gate DM invites through dm_policy instead of room_policy
• [matrix]Gate poll responses through access control
• [voice]Use inspect_err in elevenlabs stt
• [web]Preserve @ in matrix allowlists
• [matrix]Unify otp approval flow and sender visibility
• [web]Default matrix setup to password auth
• [matrix]Improve ownership recovery UX
• [graphql]Implement retry ownership in test mock
• [e2e]Correct username assertion in matrix senders test
• [web]Satisfy biome hook and lint checks

• [telegram]Route forum-topic replies to correct thread
• [telegram]Restore raw chat_id in logs, add thread_id to tracing
• [providers]Increase model probe timeout for local LLM servers
• [providers]Address PR review feedback

• [telegram]Isolate forum-topic sessions by thread_id

• [telegram]Consolidate parse_chat_target, fix typing indicator

• [telegram]Propagate thread_id parse errors in parse_chat_target
• [providers]Skip model discovery for custom providers with explicit models
• [providers]Replace redundant test with one that pins the new guard
• [provider-setup]Skip probe for custom providers without model
• [provider-setup]Remove redundant is_chat_capable_model filter
• [chat]Use system role for compaction summary
• [chat]Use user role for compaction summary
• [providers]Restore MiniMax top-level system prompt extraction
• [telegram]Allow unwrap in topic tests to satisfy workspace clippy lints
• [provider-setup]Use Arc<AtomicBool> instead of static in test

• [website]Add local dev server with SSR partial injection
• [web]Add changelog link to header nav
• [providers]Add prompt caching for Anthropic and OpenRouter
• [telegram]Extract plaintext and markdown documents from messages
• [providers]Add Fireworks.ai as primary provider

• [website]Shared nav via SSR partial, add Changelog link
• [telegram]Use std::str::from_utf8 for UTF-8 truncation
• [telegram]Normalize MIME type once, avoid redundant UTF-8 scans

• [website]Allow nav links without data-page to navigate normally
• [website]Highlight Changelog tab and show GitHub stars on /changelog
• [website]Share GitHub stars script via nav partial, fix Changelog click
• [web]Point report issue link to template chooser
• [providers]Align indentation in stream_with_tools debug macro
• [telegram]Address PR review comments on document handling
• [telegram]Enforce char limit on all document content paths
• [telegram]Prevent U+FFFD from byte-boundary truncation of CJK text
• [install]Remove spurious -1 revision from .deb filename in installer

• [release]Build changelog HTML in prepare-release instead of CI

• [ci]Use file input for changelog blob to avoid argument list too long

• [website]Add changelog HTML page and fix RPM version override

• [install]Support date-based version tags in installer and package builds

• [gateway]Embedded web chat UI at root endpoint
• [gateway]Add services, pairing, expanded methods and auth
• [agents]Add LLM chat with streaming, multi-provider support and feature flags
• [gateway]Add Tailwind-based chat UI with dark/light theme
• [config]Add multi-format config file with provider enable/disable support
• [gateway]Add model selector and WebSocket auto-reconnect
• [oauth]Add OpenAI Codex OAuth provider and reusable OAuth infrastructure
• [tools]Add LLM code execution with agent loop, tool calling, and security layers
• [agents]Add debug logging and end-to-end exec tool test
• [agents]Text-based tool calling fallback for non-native providers
• [gateway]Log user message on chat.send
• [memory]Implement memory management system with hybrid search
• [tools]Wire approval gating into exec tool with UI
• [brew]Add Homebrew formula for tap-based installation
• [website]Add static site and roadmap for moltis features
• [website]Rewrite with Tailwind CSS, Inter/JetBrains fonts, and polish
• [packaging]Add Debian package builds for amd64 and arm64
• [packaging]Add Arch Linux package builds for x86_64 and aarch64
• [packaging]Add RPM, Flatpak, Snap, AppImage, Nix, and Homebrew packaging
• [agents]Register all Codex models in provider registry
• [gateway]Structured error handling, exec cards, and approval UI
• [gateway]Add provider management UI with multi-model support
• [gateway]Persist API keys and add session management
• [gateway]Add session sidebar UI and fix session management
• [gateway]Route chat events by session key, add unread dots and thinking restore
• [agents]Add missing LLM providers and GitHub Copilot OAuth
• [gateway]Add session search with autocomplete, scroll-to and highlight
• [gateway]Include model and provider name in chat final events
• [claude]Save plans and sessions to prompts/ via hooks
• [gateway]Multi-page SPA with nav panel, crons page and methods page
• [cron]Wire cron callbacks and register CronTool for LLM use
• [cron]Implement production-grade cron scheduling system
• [projects]Add project management with context loading and session binding
• [gateway]Searchable model selector, per-session model, chat history
• [gateway]Persist model/provider in chat history and style model footer
• [gateway]Add token usage display per-message and per-session
• [gateway]Move model selector to chat page, add providers to nav panel
• [projects,sessions]Migrate project and session metadata storage to SQLite
• [gateway]Reorganize navigation, move providers to dedicated page
• [sandbox]Per-session sandbox toggle with sandbox-on-by-default
• [gateway]Show LLM thinking text, persist token usage, fix chat scroll
• [gateway]Add slash commands (/clear, /compact, /context) with autocomplete
• [context]Display sandbox details in /context command
• [providers]Add Kimi Code OAuth (device flow) provider
• [gateway]Add Channels navigation page
• [gateway]Move enabled toggle to last column in cron job table
• [telegram]Add username allowlist matching and message log store
• [compact]Auto-compact on context limit, use session provider, show summary card
• [worktree]Implement workspace worktree lifecycle features
• [ui]Project selector combo in chat header with session filtering
• [telegram]Per-channel sessions, slash commands, and default model config
• [gateway]Add logs/forensic page with real-time streaming and persistence
• [telegram]Command autocomplete, /new session, /sessions with inline keyboard
• [gateway]Live session updates, channel icons, and active session indicator
• [gateway]Add amber ping dot indicator for active sessions
• [gateway]UI improvements and provider/context enhancements
• [skills]Add agent skills system crate with discovery, registry, and CLI
• [gateway]Add Skills navigation page to web UI
• [skills]Repository-based skills with per-skill enable/disable
• [skills]Accept GitHub URLs in skill source input
• [gateway]Add native HTTPS/TLS support behind tls cargo feature
• [gateway]Hybrid asset serving with filesystem dev and embedded release
• [sandbox]Configurable images, on-demand caching, Apple Container auto-detection
• [onboarding]Replace wizard with inline identity editing in Settings
• [plugins]Add plugins crate with format adapters, install, and chat integration
• [gateway]Add Images page for managing sandbox container images
• [telegram]Add /model, /sandbox commands and improve /context display
• [gateway]Add setup code at startup and configurable config/data dirs
• [gateway]Clean auth state separation, gon pattern, and identity improvements
• [env]Add write-only environment variables with sandbox injection
• [security]Wrap secrets in secrecy::Secret<String> to prevent leaks
• [hooks]Add hook dispatch system with native and shell handlers
• [hooks]Add hook discovery, eligibility, metadata, CLI commands, and bundled hooks
• [memory]Wire memory system into gateway with tools, compaction, and session hooks
• [memory]Add embedding cache, local GGUF, fallback chain, batch API, file watcher, and pre-compaction flush
• [tools]Add web_search and web_fetch agent tools
• [sandbox]Add tracing to Apple Container and exec tool lifecycle
• [ui]Show container name in /context sandbox section
• [web]Forward client Accept-Language to web_fetch and web_search
• [sandbox]Auto-provision curl, python3, nodejs, npm in containers
• [sandbox]Expand default packages inspired by GitHub runner images
• [sandbox]Apple Container pre-built images, CLI commands, default config
• [sandbox]Expand packages from GitHub runner images, use Secret for web search keys
• [hooks]Wire all 15 hook events and add examples
• [chat]Add per-session run serialization to prevent history corruption
• [chat]Add configurable agent-level timeout enforcement
• [agents]Retry agent loop after compaction on context window overflow
• [tools]Add SpawnAgentTool for sub-agent / nested agent support
• [chat]Add message queue modes for concurrent send handling
• [agents]Sanitize tool results before appending to LLM message history
• [agents]Execute tool calls concurrently with join_all
• [mcp]Add MCP client support with discovery UI
• [gateway]Add nav sidebar count badges and MCP UI improvements
• [mcp]MCP context in chat, duplicate name handling, and misc improvements
• [mcp]Add McpTransport and McpClientTrait trait abstractions
• [mcp]Wire MCP tool bridges into agent ToolRegistry
• [mcp]SSE transport, health polling, auto-restart, and edit config
• [gateway]Add tailscale serve/funnel management, UI consistency overhaul, and HTTP/2 support
• [memory]Log status with DB size after initial sync
• [tailscale]Add Start Tailscale button when daemon is not running
• [channels]Assign default model to new telegram sessions
• [agents]Add model failover with per-provider circuit breakers
• [gateway]Add report an issue link to nav sidebar
• [config]Support MOLTIS_* env var overrides for all config fields
• [cron]Add heartbeat feature with persistent run history
• [cli]Add cargo-binstall support for binary installation
• Add Homebrew tap and auto-update workflow
• Generate random port on first run and make gateway the default command
• [ci]Add multi-arch Docker build workflow
• [agents]Enable streaming responses with tool support
• [metrics]Add Prometheus metrics with feature-gated support
• [metrics]Expand metrics to all crates with tracing feature
• [metrics]Add provider alias support for metrics differentiation
• [metrics]Add SQLite persistence and per-provider charts
• [db]Add sqlx migrations with per-crate ownership
• [local-llm]Add local LLM provider with GGUF/MLX backend selection
• [local-llm]Add MLX models and filter by backend
• [local-llm]Add HuggingFace search and custom model support
• [agents]Add unified local-llm provider with pluggable backends
• [ui]Add auto-search with debounce for HuggingFace search
• [ui]Show chat-only mode notice when selecting model without tools
• [ui]Show all configured models in providers page
• [providers]Add per-model removal and disable support
• [local-llm]Add tracing and metrics to GGUF provider
• [cli]Add database management commands (db reset, clear, migrate)
• [telegram]Auto-disable channel when another bot instance is running
• Add install script and update URLs to moltis.org
• Add Pi-inspired features — skill state, self-extension, branching, hot-reload
• [auth]Add scope support to API keys
• Typed ToolSource, per-session MCP toggle, debug panel convergence, docs & changelog
• [gateway]Add Fork button to chat header
• [gateway]Show warning banner when running on non-main git branch
• [gateway]Add mobile PWA support with push notifications (#40)
• [tls]Default HTTP redirect port to gateway port + 1 (#49)
• [memory]Add QMD backend support, citations, session export, and LLM reranking (#27)
• [deploy]Add --no-tls flag and one-click cloud deploy configs
• [gateway]Display process and system memory usage in header
• [ui]Red favicon and branch-prefixed title for non-main branches
• [ui]Add security warnings to MCP and Plugins pages
• [ui]Add click-to-view-source detail panel to enabled plugins table
• [deploy]Add MOLTIS_DEPLOY_PLATFORM env var to hide local-only providers on cloud
• [ui]Add allowlist field to onboarding channel step
• [ui]Prefill agent name and emoji in onboarding identity step
• [hooks]Add hooks web UI page
• [hooks]Seed example hook on first run
• [hooks]Add Preview/Source tabs for HOOK.md content
• [hooks]Click-to-copy hook source path
• [hooks]Show built-in hooks in the web UI
• [hooks]Wire up built-in hooks (boot-md, command-logger, session-memory)
• [channels]Replace allowlist textarea with tag-style input
• [workspace]Move persona and startup context to markdown files
• [browser]Add full browser automation support via CDP
• [browser]Enable browser tool by default
• [settings]Add Load Template button and docs link
• [browser]Display screenshot thumbnails in chat UI
• Add restart API and improve browser tool integration
• [browser]Add fullscreen lightbox for screenshot thumbnails
• [local-llm]Auto-detect backend based on model type
• [local-llm]Support Homebrew-installed mlx-lm
• [local-llm]Download and cache MLX models locally
• [local-llm]Notify user when downloading missing models
• [ui]Show download progress in chat when model is missing
• [browser]Add sandboxed browser support using Docker containers
• [browser]Add sandbox mode visibility in logs and response
• [browser]Pre-pull container image at startup with UI feedback
• [browser]Include page content in snapshot response
• [cli]Add browser command for managing browser configuration
• [browser]Add detailed logging for browser execution mode
• [ui]Show browser action and execution mode in chat
• [browser]Add screenshot download functionality
• [browser]Increase default viewport to 1920x1080
• [browser]Increase viewport to 2560x1440 with 2x Retina scaling
• [telegram]Add screenshot support for browser tool
• [browser,telegram]Improve screenshot handling and display
• [browser]Memory-based pool limits instead of fixed count
• [telegram]Send tool execution status during chat
• [logs]Show crate/module target in log output
• [ci]Add local validation gate with CI fallback
• [security]Add emergency disable for third-party skills
• [security]Surface commit provenance in skills UI
• [skills]Add websocket install progress and loading states
• [skills]Use websocket install status with honest UI messaging
• [scripts]Support local-only validation without a PR
• [workspace]Align context files with OpenClaw semantics
• [heartbeat]Skip LLM turns when HEARTBEAT.md is empty
• [heartbeat]Surface empty-file skip status in UI
• [gateway]Surface GitHub release update banner
• [voice]Add voice crate with TTS and STT providers
• [gateway]Integrate voice services with TTS and STT support
• [voice]Add voice UI with feature flag support
• [voice]Add multiple STT providers
• [voice]Add voice provider management UI with auto-detection
• [voice]Add ElevenLabs Scribe STT provider
• [voice]Integrate ElevenLabs Scribe STT and upgrade to v2
• [voice]Improve STT/TTS testing UX and fix ElevenLabs API
• [voice]Send voice messages directly to chat
• [telegram]Transcribe voice messages before dispatching to chat
• [telegram]Add info logging for unhandled media types
• [telegram]Add image support for multimodal LLM messages
• [channels]Improve voice handling and typed STT config
• [voice]Add typed provider metadata and voice preference flows
• [chat]Prefer same reply medium with text fallback
• [voice]Add provider list allowlists and narrow template defaults
• [chat]Add runtime host+sandbox prompt context
• [chat]Add Prompt button to inspect full system prompt
• [sandbox]Add tmux to default packages
• [tools]Add process tool and tmux skill for interactive terminal sessions
• Ship provider onboarding and model discovery improvements
• [ui]Redesign Voice onboarding step with settings-like experience
• [ui]Personalize TTS test phrases with user and bot names
• [gateway]Voice pending UI, TTS phrase generation, and empty message filtering
• [gateway]Persist TTS audio, per-session media, and silent replies
• [gateway]Add Context button showing full LLM messages array
• [gateway]Add queued message UI with cancel support
• [gateway]Add Copy button to full context panel
• [gateway]Add client-side message sequence number for ordering diagnostics
• [gateway]Add run_id and seq to persisted messages for parent/child linking
• [channels]Add reply threading support for Telegram messages
• [gateway]Auto-detect browser timezone via Intl.DateTimeFormat
• [metrics]Populate cache token counters from provider responses
• [tools]Add get_user_location tool with browser geolocation
• [sandbox]Add image, audio, media, and data processing packages
• [sandbox]Add document, office, and search packages
• [sandbox]Add GIS and OpenStreetMap packages
• [tools]Add sandbox_packages tool for on-demand package discovery
• [sandbox]Add communication packages, hybrid sandbox query, and mise
• [telegram]Add location sharing with live location tracking
• [onboarding]Redesign provider step as multi-provider list
• [gateway]Show no-providers card when no LLM models available
• [gateway]Validate-first provider setup in onboarding and settings
• [oauth]Import auto-detected tokens into central store at startup
• [providers]Add model selection for auto-detected providers
• [onboarding]Add summary step and improve channel UX
• [gateway]Add Telegram-style waveform audio player
• [gateway]Add generic session upload endpoint replacing base64-over-WS
• [gateway]Add drag-and-drop image upload to chat and auth tests
• [voice]Add TTS text sanitization and voice-friendly LLM prompting
• [tools]Reverse geocode user location to human-readable place names
• [agents]Retry once on transient LLM provider errors
• [tools]Add show_map tool and location precision modes
• [telegram]Consolidate bot responses into single message with logbook
• [auth]Passkey UX improvements and mDNS origin support
• [benchmarks]Add boot-path performance benchmarks for CodSpeed
• [cli]Add moltis memory subcommand with search and status
• [telegram]Send native location pin for show_map tool
• [telegram]Send voice replies with text transcript
• [gateway]Add search to project filter and fix button alignment
• [gateway]Attach project to new sessions from filtered view
• [sessions]Add server-side unread tracking via last_seen_message_count
• [gateway]Auto-detect models after saving provider API key (#83)
• [sessions]Add entity versioning to prevent stale updates
• [gateway]Improve logs filter UX and branch favicon contrast
• [gateway]Add logout button to header bar (#86)
• [gateway]Add comprehensive e2e test suite for web UI (#87)
• [gateway]Add auth-aware endpoint throttling and login retry UX
• [gateway]Auto-detect WebAuthn RP ID from PaaS environment variables
• [cli]Add --version flag
• [gateway]Centralize SPA routes and preserve config TOML comments
• [models]Tighten allowlist matching and support probe output (#91)
• [config]Add moonshot to default offered providers
• [onboarding]Preselect passkey auth method when available
• [config]Auto-create SOUL.md with default content on first run
• [gateway]Disconnect all WS clients on credential changes
• [providers]Prefer configured models and merge live model discovery
• [providers]Multi-select preferred models, keyOptional, createdAt sorting
• [providers]Filter non-chat models and fix per-model tool support
• [providers]Show model dates, probe on select, spacebar pause
• [onboarding]Multi-select model picker with probe badges
• [memory]Inject MEMORY.md into system prompt and fix file watcher
• Auto-select and install browser tool backends (#130)
• [agents]Strip <think> tags from OpenAI-compatible providers and add MiniMax
• [chat]Preserve reasoning text in tool cards across page reloads
• [agents]Add Z.AI (Zhipu) as OpenAI-compatible provider
• Env injection, sandbox recovery, UI fixes, provider improvements (#108)
• [gateway]Show server start time at bottom of onboarding
• [browser]Auto-inject low-memory Chromium flags on constrained systems
• [gateway]Add generic OpenAI-compatible provider support
• [tools]SSRF allowlist for Docker inter-container networking (#146)
• [config]Enable openrouter in default offered providers
• [mcp]Add OAuth 2.1 support for remote MCP servers (#148)
• [telegram]Add channel streaming with stream_mode fallback (#165)
• [tools]Add calc tool for safe arithmetic evaluation
• [cron]Add per-job model and execution target controls (#170)
• [gateway]Cache session histories and show switch loader
• [map]Add provider-aware show_map links
• [gateway]Render markdown and ansi tables in chat
• [browser]Persist Chrome profile across sessions (#162)
• [memory]Reduce baseline memory footprint for lightweight devices
• [telegram]Save voice audio to session media for web UI playback
• [cron]Add event-driven heartbeat wake with system events queue
• [telegram]Render markdown tables as formatted pre blocks
• [gateway]Static file caching for public share pages
• [cron]Deliver agent turn output to Telegram channels
• [gateway]Seed dcg-guard hook and polish onboarding badges
• [graphql]Add GraphQL API exposing all RPC methods (#200)
• [tools]Add send_image tool for channel image delivery (#224)
• [gateway]Expand identity emoji picker options (#206)
• [web]Browsable skills list in repo cards
• [providers]Add configurable OpenAI websocket stream transport (#227)
• [sandbox]Install latest gogcli in default sandbox images (#232)
• [agents]Add multi-agent personas with CRUD UI (#97)
• [caldav]Add CalDAV integration for calendar CRUD (#84)
• [channels]Add Microsoft Teams channel integration (#231)
• [import]Add OpenClaw import crate, CLI, gateway RPC, and UI (#217)
• [web]Add vault algorithm description to encryption settings page
• [web]Internationalization (i18n) with English and French locales (#237)
• [whatsapp]Add WhatsApp channel support (#73)
• [web]Add sandbox shared-home settings UI and config
• [protocol]Upgrade WebSocket protocol from v3 to v4 (#247)
• [agents,providers,chat]Universal tool support for all models
• [openclaw-import]Import workspace files and add safety messaging
• [providers]Promote Gemini to first-class OpenAI-compatible provider
• [i18n]Add full zh-CN (Simplified Chinese) localization (#260)
• Add channel-aware heartbeat delivery and send_message agent tool (#270)
• [memory]Add tree-sitter code splitter and RRF search merge
• [web]Add Shiki syntax highlighting to code blocks
• [sandbox]Add GitHub runner parity packages and enable corepack (#284)
• [providers]Add first-class LM Studio provider (#286)
• [agents]Enrich spawn_agent presets with identity, policies, memory (#271)
• [web]Show running version at bottom of identity settings
• [channels]Channel architecture phase 5, contract suites, and observability baseline (#289)
• [ci]Add release dry-run mode
• [browser]Add container_host for Docker-in-Docker connectivity (#300)
• [ios]Auto-discover server identity and show emojis (#297)
• [website]Migrate cloudflare website into monorepo (#302)
• [local-llm]Allow arbitrary HuggingFace model IDs for MLX models
• [web,tools]AOT WASM pre-compilation and Shiki CDN loading
• [cli]Remove wasm from default features to reduce memory
• [gateway]Make provider discovery startup non-blocking
• [monitoring]Track memory history and improve local-llm memory reporting (#325)
• [ios]Add local llama cpp memory field to GraphQL schema
• [providers]Include reasoning fields for kimi models (#323)
• [chat]Tabs to filter chats between sessions and cron (#338)
• [oauth]Support pasted callback URL fallback (#365)
• [providers]Add reasoning effort support for models with extended thinking (#363)
• [providers]Add Responses API support to GitHub Copilot provider (#393)
• [release]Migrate to date-based versioning (YYYYMMDD.NN) (#394)
• Support secret remote MCP URLs and headers (#416)
• [docker]Support generic provider env bootstrap (#401)
• [skills]Support safe agent-written sidecar files (#413)
• [mcp]Add custom display names for MCP servers
• [mcp]Add displayName to iOS GraphQL schema
• [skills]Gate installer behind install feature
• [gateway]Embedded web chat UI at root endpoint
• [gateway]Add services, pairing, expanded methods and auth
• [agents]Add LLM chat with streaming, multi-provider support and feature flags
• [gateway]Add Tailwind-based chat UI with dark/light theme
• [config]Add multi-format config file with provider enable/disable support
• [gateway]Add model selector and WebSocket auto-reconnect
• [oauth]Add OpenAI Codex OAuth provider and reusable OAuth infrastructure
• [tools]Add LLM code execution with agent loop, tool calling, and security layers
• [agents]Add debug logging and end-to-end exec tool test
• [agents]Text-based tool calling fallback for non-native providers
• [gateway]Log user message on chat.send
• [memory]Implement memory management system with hybrid search
• [tools]Wire approval gating into exec tool with UI
• [brew]Add Homebrew formula for tap-based installation
• [website]Add static site and roadmap for moltis features
• [website]Rewrite with Tailwind CSS, Inter/JetBrains fonts, and polish
• [packaging]Add Debian package builds for amd64 and arm64
• [packaging]Add Arch Linux package builds for x86_64 and aarch64
• [packaging]Add RPM, Flatpak, Snap, AppImage, Nix, and Homebrew packaging
• [agents]Register all Codex models in provider registry
• [gateway]Structured error handling, exec cards, and approval UI
• [gateway]Add provider management UI with multi-model support
• [gateway]Persist API keys and add session management
• [gateway]Add session sidebar UI and fix session management
• [gateway]Route chat events by session key, add unread dots and thinking restore
• [agents]Add missing LLM providers and GitHub Copilot OAuth
• [gateway]Add session search with autocomplete, scroll-to and highlight
• [gateway]Include model and provider name in chat final events
• [claude]Save plans and sessions to prompts/ via hooks
• [gateway]Multi-page SPA with nav panel, crons page and methods page
• [cron]Wire cron callbacks and register CronTool for LLM use
• [cron]Implement production-grade cron scheduling system
• [projects]Add project management with context loading and session binding
• [gateway]Searchable model selector, per-session model, chat history
• [gateway]Persist model/provider in chat history and style model footer
• [gateway]Add token usage display per-message and per-session
• [gateway]Move model selector to chat page, add providers to nav panel
• [projects,sessions]Migrate project and session metadata storage to SQLite
• [gateway]Reorganize navigation, move providers to dedicated page
• [sandbox]Per-session sandbox toggle with sandbox-on-by-default
• [gateway]Show LLM thinking text, persist token usage, fix chat scroll
• [gateway]Add slash commands (/clear, /compact, /context) with autocomplete
• [context]Display sandbox details in /context command
• [providers]Add Kimi Code OAuth (device flow) provider
• [gateway]Add Channels navigation page
• [gateway]Move enabled toggle to last column in cron job table
• [telegram]Add username allowlist matching and message log store
• [compact]Auto-compact on context limit, use session provider, show summary card
• [worktree]Implement workspace worktree lifecycle features
• [ui]Project selector combo in chat header with session filtering
• [telegram]Per-channel sessions, slash commands, and default model config
• [gateway]Add logs/forensic page with real-time streaming and persistence
• [telegram]Command autocomplete, /new session, /sessions with inline keyboard
• [gateway]Live session updates, channel icons, and active session indicator
• [gateway]Add amber ping dot indicator for active sessions
• [gateway]UI improvements and provider/context enhancements
• [skills]Add agent skills system crate with discovery, registry, and CLI
• [gateway]Add Skills navigation page to web UI
• [skills]Repository-based skills with per-skill enable/disable
• [skills]Accept GitHub URLs in skill source input
• [gateway]Add native HTTPS/TLS support behind tls cargo feature
• [gateway]Hybrid asset serving with filesystem dev and embedded release
• [sandbox]Configurable images, on-demand caching, Apple Container auto-detection
• [onboarding]Replace wizard with inline identity editing in Settings
• [plugins]Add plugins crate with format adapters, install, and chat integration
• [gateway]Add Images page for managing sandbox container images
• [telegram]Add /model, /sandbox commands and improve /context display
• [gateway]Add setup code at startup and configurable config/data dirs
• [gateway]Clean auth state separation, gon pattern, and identity improvements
• [env]Add write-only environment variables with sandbox injection
• [security]Wrap secrets in secrecy::Secret<String> to prevent leaks
• [hooks]Add hook dispatch system with native and shell handlers
• [hooks]Add hook discovery, eligibility, metadata, CLI commands, and bundled hooks
• [memory]Wire memory system into gateway with tools, compaction, and session hooks
• [memory]Add embedding cache, local GGUF, fallback chain, batch API, file watcher, and pre-compaction flush
• [tools]Add web_search and web_fetch agent tools
• [sandbox]Add tracing to Apple Container and exec tool lifecycle
• [ui]Show container name in /context sandbox section
• [web]Forward client Accept-Language to web_fetch and web_search
• [sandbox]Auto-provision curl, python3, nodejs, npm in containers
• [sandbox]Expand default packages inspired by GitHub runner images
• [sandbox]Apple Container pre-built images, CLI commands, default config
• [sandbox]Expand packages from GitHub runner images, use Secret for web search keys
• [hooks]Wire all 15 hook events and add examples
• [chat]Add per-session run serialization to prevent history corruption
• [chat]Add configurable agent-level timeout enforcement
• [agents]Retry agent loop after compaction on context window overflow
• [tools]Add SpawnAgentTool for sub-agent / nested agent support
• [chat]Add message queue modes for concurrent send handling
• [agents]Sanitize tool results before appending to LLM message history
• [agents]Execute tool calls concurrently with join_all
• [mcp]Add MCP client support with discovery UI
• [gateway]Add nav sidebar count badges and MCP UI improvements
• [mcp]MCP context in chat, duplicate name handling, and misc improvements
• [mcp]Add McpTransport and McpClientTrait trait abstractions
• [mcp]Wire MCP tool bridges into agent ToolRegistry
• [mcp]SSE transport, health polling, auto-restart, and edit config
• [gateway]Add tailscale serve/funnel management, UI consistency overhaul, and HTTP/2 support
• [memory]Log status with DB size after initial sync
• [tailscale]Add Start Tailscale button when daemon is not running
• [channels]Assign default model to new telegram sessions
• [agents]Add model failover with per-provider circuit breakers
• [gateway]Add report an issue link to nav sidebar
• [config]Support MOLTIS_* env var overrides for all config fields
• [cron]Add heartbeat feature with persistent run history
• [cli]Add cargo-binstall support for binary installation
• Add Homebrew tap and auto-update workflow
• Generate random port on first run and make gateway the default command
• [ci]Add multi-arch Docker build workflow
• [agents]Enable streaming responses with tool support
• [metrics]Add Prometheus metrics with feature-gated support
• [metrics]Expand metrics to all crates with tracing feature
• [metrics]Add provider alias support for metrics differentiation
• [metrics]Add SQLite persistence and per-provider charts
• [db]Add sqlx migrations with per-crate ownership
• [local-llm]Add local LLM provider with GGUF/MLX backend selection
• [local-llm]Add MLX models and filter by backend
• [local-llm]Add HuggingFace search and custom model support
• [agents]Add unified local-llm provider with pluggable backends
• [ui]Add auto-search with debounce for HuggingFace search
• [ui]Show chat-only mode notice when selecting model without tools
• [ui]Show all configured models in providers page
• [providers]Add per-model removal and disable support
• [local-llm]Add tracing and metrics to GGUF provider
• [cli]Add database management commands (db reset, clear, migrate)
• [telegram]Auto-disable channel when another bot instance is running
• Add install script and update URLs to moltis.org
• Add Pi-inspired features — skill state, self-extension, branching, hot-reload
• [auth]Add scope support to API keys
• Typed ToolSource, per-session MCP toggle, debug panel convergence, docs & changelog
• [gateway]Add Fork button to chat header
• [gateway]Show warning banner when running on non-main git branch
• [gateway]Add mobile PWA support with push notifications (#40)
• [tls]Default HTTP redirect port to gateway port + 1 (#49)
• [memory]Add QMD backend support, citations, session export, and LLM reranking (#27)
• [skills]Show confirmation hint after skill creation/update
• [skills]Add skill editing and forking from the web UI
• [skills]Show confirmation hint after skill creation/update
• [skills]Add skill editing and forking from the web UI
• [mcp]Make request timeout configurable
• [agents]Lazy tool registry with tool_search meta-tool
• [local-llm]Add opt-in vulkan gguf support
• [providers]Add MiniMax M2.7 and missing M2.1 highspeed models
• [prompt]Stabilize system prompt for local LLM KV cache
• [scripts]Skip local-validate.sh when commit already passed
• [docker]Add Node.js/npm to Docker image for MCP servers
• [ci]Add pre_release option to release workflow

• [website]Extract inline CSS to separate styles.css file
• [gateway]Split monolithic app.js into 24 ES modules
• [gateway]Extract inline JS styles to CSS and add message dedup
• [gateway]Preact skills page, REST APIs, biome linting, perf fixes
• [gateway]Reduce cognitive complexity in JS modules and add biome CI
• [gateway]Faster log loading with memory-first reads and batch rendering
• Simplify auth, onboarding, and settings code
• Simplify hooks and memory code
• [env]Wrap env var values in secrecy::Secret
• Remove all unsafe code and add workspace-wide deny(unsafe_code)
• Keep API keys wrapped in Secret<String> through provider construction
• [agents]Simplify tool result sanitization
• [mcp]Remove dead code and deduplicate config parsing
• [memory]Skip redundant work on sync restart
• [gateway]Use inline script for identity instead of server-side HTML replace
• [api]Split /api/skills and /api/plugins into separate endpoints
• [ui]Use CSS classes for tailscale status bar, single line layout
• [ui]Rename ts-status-bar to generic info-bar classes and fix image page font size
• [tailscale]Remove Start Tailscale button and up endpoint
• [ui]Move session rename/delete to chat header and clean up
• Remove unsafe set_var in detect.rs tests, resolve merge conflicts
• [gateway]Reorder nav items alphabetically with Chat first
• [ci]Extract signing logic into composite action
• [ui]Move install hint markup to HTML template
• [ui]Move model notice card HTML to template
• Rename crate moltis-cli to moltis
• [deps]Centralize all dependency versions in workspace root
• [ci]Run lightweight lint jobs on GitHub-hosted runners
• [ci]Restore Docker jobs to release workflow, remove docker.yml
• [plugins]Replace hand-rolled date arithmetic with time crate
• [gateway]Use time::Duration::days in tls expiry check
• Use moltis_config::data_dir() for all path resolution
• [browser]Add browser detection and simplify session handling
• [settings]Rename Tools to Configuration and edit full config
• [browser]Use typed structs for OpenAI tool schemas
• [local-llm]Add modular ResponseParser trait for output parsing
• [browser]Use unified sandbox infrastructure for browser containers
• [gateway]Add defense-in-depth auth checks and DRY server.rs
• [browser]Send execution mode from server in tool_call_start
• [channels]Use ChannelType enum instead of string matching
• [browser]Sandbox mode follows session, fix tall screenshot lightbox
• [ci]Move local status check logic into script
• [ci]Avoid blocking lint and test on local zizmor
• [security]Use gitoxide metadata and improve trust UX
• [scripts]Rename local-validate-pr.sh to local-validate.sh
• [chat]Simplify runtime host+sandbox prompt context
• [sandbox]Remove redundant mkdir exec from ensure_ready
• Simplify branch review fixes across UI, gateway, and config
• [gateway]Use typed structs for chat broadcast payloads
• [ui]Rename /images and /api/images routes to /sandboxes
• [gateway]Consolidate GatewayState per-field RwLocks into single RwLock<GatewayInner>
• [mcp]Consolidate McpManager per-field RwLocks into single RwLock<McpManagerInner>
• [gateway]Replace project select with custom combo dropdown
• [gateway]Unify auth into single check_auth() gate
• [browser]Make stale cleanup path expression-based
• [ci]Move E2E job into CI workflow
• [build]Enable thin LTO and binary stripping in release profile
• [gateway]Extract shared voice, identity, and channel utils
• [tools]Resolve browser sandbox mode from SandboxRouter directly
• [gateway]Reorder onboarding screens
• [session]Typed params for patch and voice_generate (#131)
• [gateway]Remove standalone /crons route, use /settings/crons
• [agents]Simplify prompt builder and runtime context
• [prompt]Compact prompt sections and add server runtime time
• [web]Extract web UI into dedicated moltis-web crate
• [identity]Consolidate creature+vibe into single theme field
• [gateway]Extract moltis-service-traits crate (Phase 0)
• [gateway]Extract moltis-tls and moltis-tailscale crates (Phase 1)
• [gateway]Extract moltis-auth crate (Phase 2)
• [gateway]Extract moltis-provider-setup crate (Phase 3)
• [gateway]Extract moltis-chat crate (Phase 4)
• [web]Move share_render.rs from gateway to moltis-web (Phase 1c)
• [providers]Extract provider implementations into new crate
• [errors]Move crates to typed thiserror enums (#226)
• [tools]Replace anyhow bridge with crate::Result for internal APIs (#257)
• [ffi]Tighten unsafe_code allowances
• [channels]Registry-driven dispatch for cheap new channels (#277)
• [gateway]Fetch updates from releases manifest instead of GitHub API
• [web]Move settings nav icons from JS to CSS
• Externalize web/wasm assets and reduce memory footprint (#321)
• [web]Move chat history hydration to paged HTTP
• [web]Paginate sessions and auto-load older history
• [tools]Split sandbox.rs into sandbox/ module directory
• [website]Extract inline CSS to separate styles.css file
• [gateway]Split monolithic app.js into 24 ES modules
• [gateway]Extract inline JS styles to CSS and add message dedup
• [gateway]Preact skills page, REST APIs, biome linting, perf fixes
• [gateway]Reduce cognitive complexity in JS modules and add biome CI
• [gateway]Faster log loading with memory-first reads and batch rendering
• Simplify auth, onboarding, and settings code
• Simplify hooks and memory code
• [env]Wrap env var values in secrecy::Secret
• Remove all unsafe code and add workspace-wide deny(unsafe_code)
• Keep API keys wrapped in Secret<String> through provider construction
• [agents]Simplify tool result sanitization
• [mcp]Remove dead code and deduplicate config parsing
• [memory]Skip redundant work on sync restart
• [gateway]Use inline script for identity instead of server-side HTML replace
• [api]Split /api/skills and /api/plugins into separate endpoints
• [ui]Use CSS classes for tailscale status bar, single line layout
• [ui]Rename ts-status-bar to generic info-bar classes and fix image page font size
• [tailscale]Remove Start Tailscale button and up endpoint
• [ui]Move session rename/delete to chat header and clean up
• Remove unsafe set_var in detect.rs tests, resolve merge conflicts
• [gateway]Reorder nav items alphabetically with Chat first
• [ci]Extract signing logic into composite action
• [ui]Move install hint markup to HTML template
• [ui]Move model notice card HTML to template
• Rename crate moltis-cli to moltis
• [media]Replace manual MIME lookup with mime_guess crate
• [auth]Extract GatewayState::is_secure() to centralise cookie Secure logic
• [browser]Use match instead of nested if/else for dir creation
• [httpd]Extract moltis-httpd crate as HTTP transport facade
• [ci]Restore release.yml clippy runner to self-hosted
• [tools]Unify rescue helper, add tracing, exclude action-level keys
• [docker]Split runtime installs into separate layers
• [dev]Deduplicate release-preflight by delegating to lint

• Merge branch 'main' into claude/remove-unsafe-code-ehZlQ
• [ui]Remove decimal digits from memory display
• [ci]Add latest Docker tag on tag pushes, remove unused branch tag
• [ci]Remove template-expanded gate debug output
• [gateway]Use danger button for session delete
• [gateway]Size-match delete button with fork and add danger style
• [gateway]Match delete button size to fork button
• [browser]Remove linux-only unused mut in stale cleanup
• [codspeed]Remove noisy n=1 session_store_list case
• Remove cargo-binstall install path
• [browser]Remove needless return in match arm
• [gateway]Remove unnecessary path qualification flagged by clippy
• [cron]Remove oneOf from tool schema for OpenAI responses
• [readme]Drop logo icon to save vertical space
• [readme]Restore 64px favicon, remove link from title
• Remove trailing whitespace in validate.rs
• [cli]Remove binstall metadata and fix feature indentation
• [sandbox]Remove redundant error conversions
• [web]Remove nested onboarding scroll and restore settings nav icons
• [web]Declutter chat controls and fix dropdown positioning
• Merge branch 'main' into claude/remove-unsafe-code-ehZlQ
• [local-llm]Remove unused gguf runtime helper

• [oauth]Use correct OpenAI client_id, add config layer and tests
• [agents]Prefer tool-capable provider when tools are registered
• [agents]Register builtin providers before genai to enable tool calling
• [gateway]Auto-focus chat input on page load
• [sessions]Add missing sandbox_enabled column migration, stop swallowing errors
• [gateway]Move sandbox button next to model select, fix default state
• Resolve all clippy warnings (collapsible_if, map_flatten, or_default, unused imports)
• [gateway]Open nav panel by default
• [gateway]Add session icons and replace spinner in-place
• [gateway]Add missing renderCompactCard export to page-chat.js
• [gateway]Fix routing, styling, and layout issues in JS modules
• [telegram]Remove channel prefix from user messages and log LLM responses
• [ci]Fix biome version and worktree test default branch
• [ci]Restore biome version pin and fix cargo fmt
• Update CLAUDE.md pre-commit checks and fix clippy warnings
• [ci]Update setup-biome to v2.7.0 and fix sort_by_key clippy lint
• [tests]Configure git user in worktree test setup for CI
• Replace sort_by with sort_by_key in project store
• [assets]Fix biome formatting and empty catch block in sandbox.js
• [ui]Match settings nav item height and style to main navigation
• [security]Wrap API keys in Secret<String> for memory subsystem
• Use RecommendedCache for cross-platform watcher, sort_by_key for clippy
• [hooks]Ignore broken pipe on stdin write when child doesn't read it
• [ui]Show URL and query in web tool call cards
• [sandbox]Restart stopped Apple containers in ensure_ready
• [sandbox]Pass sleep infinity to Apple Container run
• [sandbox]Promote exec routing and container logs to info level
• [sandbox]Handle Apple Container inspect returning empty for nonexistent containers
• [sandbox]Default working_dir to "/" when running inside container
• [security]Reject cross-origin WebSocket upgrades (CSWSH protection)
• [sandbox]Inject env vars in Apple Container backend
• [exec]Redact env var values from command output
• [exec]Redact base64 and hex encoded env var values from output
• [config]Use correct defaults for ToolsConfig when not deserialized
• [gateway]Nav badge counts and cron page instant load
• [memory]Write memory files to data dir instead of cwd
• [gateway]Eliminate identity flash on page load
• [ui]Use consistent button class for skills Remove button
• [ui]Fix skills Disable button style and add error feedback
• [skills]Route disable to correct RPC for plugin-sourced skills
• [ui]Use inline style for tailscale status green dot
• [ui]Use status-dot connected class for tailscale green dot
• [ui]Replace Tailwind arbitrary classes with inline styles in tailscale templates
• [ui]Make tailscale status bar fit-content width
• [ui]Make info-bar full width to match warning/error divs
• [ui]Constrain info-bar to max-width 600px matching alert divs
• [ui]Add alert-info-text to shared alert base styles
• [ui]Add btn-row and btn-row-mt CSS classes for button spacing
• [ui]Space cancel/create buttons apart and normalize height
• [ui]Improve tailscale loading message to set expectations
• [tailscale]Open Tailscale.app on macOS instead of running tailscale up
• [ui]Preserve funnel security warning when rebuilding tailscale DOM
• [ui]Use alert-warning-text style for funnel auth warning
• [ui]Replace auth text with green button linking to security settings
• [ui]Move auth button below the funnel security warning panel
• [ui]Remove @layer components wrapper to fix nav specificity
• [ui]Update session list after /clear and select next on delete
• [ui]Format environment variable timestamps with luxon
• [gateway]Stop echoing web UI messages to Telegram channel
• Collapse nested if statements in config loader
• [ci]Fix package build failures
• [ci]Correct cargo-deb assets order
• [local-llm]Detect available package managers for MLX install
• [local-llm]Detect mlx-lm installed via brew
• [ui]Show install commands on separate lines
• [local-llm]Rename provider from local-gguf to local-llm
• [local-llm]Fix HuggingFace API response parsing
• [ui]Show searching state on HuggingFace search button
• [ui]Close modal and prevent multiple clicks on HF model selection
• [chat]Remove per-message tools warning broadcast
• [ui]Remove scroll from local model list
• [ci]Install cmake for llama.cpp build and fix Biome errors
• [ci]Fix local-llm build in CI
• [migrations]Use set_ignore_missing to allow multi-crate migrations
• [cli]Run gateway migrations from db migrate command
• [telegram]Don't delete channel from database on conflict
• [gateway]Load images nav count async to avoid blocking page serve
• [gateway]Replace branch icon, fix fork UX issues
• Streaming tool calls, skill directory, skills page UX
• [skills]Handle disable for personal/project skills
• [skills]Use app modal for delete confirmation instead of system confirm()
• [skills]Danger button on delete modal, fix disable routing for SkillsService
• [ui]Use decimal units (MB, GB) for memory display
• [ui]Strengthen skills security warning
• [ui]Prevent branch banner error from blanking the page
• [ci]Deploy docs to gh-pages branch instead of using deploy-pages action
• [ci]Use GitHub-hosted runners for PRs to protect self-hosted runners
• [ci]Add persist-credentials: false to docs workflow checkout
• [ci]Use macOS runners for apple-darwin builds
• [ci]Use project-local cargo config for cross-compilation
• [ci]Fix all build failures in release workflow
• [ci]Migrate release builds to GitHub-hosted runners
• [ci]Build Docker images natively per arch instead of QEMU emulation
• [ci]Only build Docker images on release tags, not main pushes
• [deploy]Persist config directory on cloud providers
• [ci]Use ubuntu-24.04-arm for Linux ARM jobs
• [ci]Correct cargo-sbom CycloneDX format flag
• [ci]Derive release package versions from git tags
• [ui]Compact memory info display in header bar
• [docker]Add runtime libgomp and image smoke test
• [gateway]Route setup through onboarding and stabilize hosted WS auth
• [ui]Disable sandbox controls when no backend is available
• [ci]Remove gh dependency from homebrew workflow
• [ci]Run homebrew update after release assets
• [ci]Avoid invalid secrets context in release workflow
• [hooks]Show GitHub source link for built-in hooks instead of editor
• [hooks]Use checked_div for avg latency calculation
• [gateway]Keep telegram typing active until reply send
• [sandbox]Skip prebuild when disabled and require daemon access
• [gateway]Use instance callback URL for OAuth flows
• [ci]Apply nightly rustfmt and clarify formatting checklist
• [ci]Add CUDA static libs and linker paths
• [ci]Map Debian CUDA libs into CUDA root for static link
• [browser]Register BrowserTool in tool registry
• [agents]Correctly pass tools through ProviderChain and detect tool result failures
• [browser]Include install instructions when browser launch fails
• [browser]Check macOS app bundles before PATH for browser detection
• [browser]Improve tool description with explicit examples
• [agents]Add strict mode to OpenAI-compatible tool schemas
• [browser]Default to navigate action when only url is provided
• [chat]Preserve message order when tool calls interleave with text
• [chat]Hide schema validation error cards from UI
• [chat]Show validation errors as muted informational cards
• [browser]Recursively add additionalProperties to nested schemas
• [browser]Ensure all properties in required for OpenAI strict mode
• [browser]Use shared to_openai_tools in openai_codex provider
• [browser]Add to_responses_api_tools for Codex API format
• [browser]Handle objects without properties in strict mode
• [browser]Use data URI format for screenshots
• [browser]Handle both raw base64 and data URI in screenshot display
• [local-llm]Reject MLX models in GGUF loader with helpful error
• [local-llm]Route MLX models to MLX backend via local_llm provider
• [local-llm]Allow HuggingFace repo IDs for MLX models
• [local-llm]Parse mlx_lm CLI output correctly
• [gateway]Use LocalLlmProvider for UI-added models
• [local-llm]Check legacy registry for MLX models
• [local-llm]Use local cache for MLX models instead of HF repo path
• [local-llm]Multiple fixes for model download and registration
• [browser]Improve container detection and settings error handling
• [lint]Resolve biome errors for CI
• [browser]Update deprecated rand API calls
• [browser]Improve tool description to trigger on 'browse' keyword
• [browser]Don't write screenshot files to disk
• [sandbox]Resolve race condition where sandbox shows disabled at startup
• [gateway]Add missing config and metrics routes to push-notifications build
• [ui]Remove duplicate browser info from result area
• [browser]Retry on dead WebSocket connections
• [browser]Validate URLs to reject LLM garbage
• [browser]Improve screenshot handling and suppress chromiumoxide logs
• [browser]Ensure viewport is applied to pages and add debug logging
• [browser]Wait for Chrome to be ready before connecting
• [browser]Auto-track session_id to prevent pool exhaustion
• [browser]Use navigation timeout for sandboxed browser connections
• [telegram]Re-send typing indicator after tool status message
• [ci]Pin sccache action to commit SHA
• [ci]Publish local validation statuses to PR head repo
• [ci]Fallback to gh auth token for local status publishing
• [ci]Clean stale llama cmake dirs in local validator
• [ci]Use non-CUDA local validation defaults on macOS
• [ci]Wait for local statuses before failing PR checks
• [ci]Correct parallel local check PID handling
• [ci]Stabilize local validator output and show run URL
• [security]Harden skill execution and web asset safety
• [security]Require trust before enabling installed skills
• [security]Harden tarball extraction and pin install provenance
• [security]Require re-trust when skill source drifts
• [security]Block suspicious dependency install commands
• [security]Recover orphaned installs and protect seed skills
• [skills]Seed template and quiet invalid skill warnings
• [memory]Align indexing scope with openclaw defaults
• [memory]Coalesce stale-index cleanup logs
• [validation]Require clean tree for local PR status
• [agents]Replace serde_json::Value with typed ChatMessage in LlmProvider
• [scripts]Add lockfile check and enforce clean worktree in all modes
• [clippy]Address nested-if and duplicate branch warnings
• [ui]Remove redundant heartbeat token-saver panel
• [chat]Keep skill management tools available at runtime
• [gateway]Support configured GitHub repo for update checks
• [update]Disable checks when repo URL is unset
• [ci]Pin workflow actions to commit SHAs
• [ci]Expose CUDA compat libs for test runtime
• [concurrency]Remove Mutex<()> sentinels and harden exec working dir
• [startup]Ensure data dir exists and isolate exec tests
• [startup]Fail fast when workspace dirs cannot be created
• [exec]Satisfy clippy lints in sandbox detection tests
• Add voice schema to build_schema_map for validation
• [voice]Auto-enable provider and share ElevenLabs key between TTS/STT
• [voice]Align transcribing indicator right and fix recording delay
• [voice-ui]Align mic state timing and preserve settings scroll
• [chat-ui]Show reply medium badge in assistant footer
• [sandbox]Create /home/sandbox directory in generated images
• [sandbox]Create /home/sandbox at container startup via exec
• [sandbox]Default working directory to /home/sandbox
• [sandbox]Keep apt index in pre-built images
• [gateway]Remove no-op or_else closure flagged by clippy
• [ui]Bust Safari cache in dev mode and fix detected providers border
• [tools]Handle empty session_name in process tool start action
• [gateway]Validate config before restart to prevent crash loops
• [gateway]Use YAML list for allowed-tools in seeded tmux skill
• [gateway]Auto-enable ElevenLabs counterpart in onboarding voice test
• [gateway]Persist empty assistant messages for LLM history coherence
• [gateway]Remove separate voice-generating indicator during TTS
• [gateway]Enable TTS counterpart when saving ElevenLabs/Google STT key
• [gateway]Reconstruct tool messages in full context view
• [gateway]Fix queued message detection and ordering
• [gateway]Move queued messages to dedicated bottom tray
• [gateway]Clear queued tray on session switch and add debug logs
• [gateway]Never attach model footer or timestamp to user messages
• [gateway]Prevent multi-move race in queued message tray
• [gateway]Defer user message persist until after semaphore
• [gateway]Move queued messages only after response is rendered
• [tools]Fall back to DuckDuckGo when web_search API key is missing
• [onboarding]Skip finish screen and redirect to chat directly
• [ui]Rename Images nav item to Sandboxes
• [ui]Revert API routes to /api/images, remove "live" status text
• [cron]Skip heartbeat when no prompt configured, fix duplicate runs
• [tools]Collapse nested if in location tool (clippy)
• [gateway]Collapse nested if in ws timezone handling (clippy)
• [agents]Make openai-codex token refresh async to prevent runtime panic
• [gateway]Ensure session row exists before setting channel binding
• Sync Cargo.lock with workspace version 0.3.1
• [gateway]Use is_none_or instead of map_or for clippy
• [sessions]Use blocking file lock to prevent concurrent write failures
• Auto-sync Cargo.lock in local-validate.sh
• [docker]Resolve deployment errors on DigitalOcean Docker containers
• [install]Add missing -1 revision to .deb filename
• [gateway]Show real IP in banner when bound to 0.0.0.0
• [oauth]Downgrade missing token file log from warn to debug
• [oauth]Downgrade routine token store logs to debug
• [providers]Show model selector after OAuth in onboarding
• [onboarding]Improve scroll padding in channel and summary steps
• [gateway]Use route-specific body limit for upload endpoint
• [gateway]Reduce global request body limit to 2 MB
• [agents]Pass multimodal content to LLM providers instead of stripping images
• [gateway]Unlock audio autoplay via AudioContext on user gesture
• [tools]Instruct LLM to use lat/lon for searches, not place names
• [tools]Use place names in map links and skip Brave when unconfigured
• [tools]Block DuckDuckGo fallback for 1h after CAPTCHA
• [voice]Add autoplay debug logging, stop audio on mic, Esc cancels
• [ci]Move coverage job into CI workflow
• [ci]Slim down coverage job to avoid disk space exhaustion
• [ci]Exclude moltis-tools from coverage and drop sccache
• Collapse nested if blocks per clippy
• Collapse nested if blocks in provider_setup per clippy
• [tools]Use if-let instead of is_some/unwrap in map test
• Deny unwrap/expect in production code via clippy lints
• [gateway]Mark boot_time test as ignored for CI coverage
• [gateway]Fix session project filter and add clear all sessions
• [gateway]Prevent layout shift from active session border
• [gateway]Collapse nested if per clippy collapsible_if lint
• [ci]Consolidate release uploads into single job
• [gateway]Reactive session badge with optimistic updates
• [gateway]Bump session badge for channel messages
• [gateway]Stop auto-switching session on channel messages
• [gateway]Broadcast session_cleared so /clear from channels syncs web UI
• [voice]Add media-src to CSP for TTS playback and improve voice diagnostics
• [auth]Clear setup_complete when last passkey is removed
• [gateway]Bump session badge on every WS event
• [gateway]Shrink passkey action buttons and hide danger zone when no auth configured
• [ci]Align release clippy with nightly flags and fix test lints
• [browser]Add lifecycle management for browser containers (#88)
• [auth]Break onboarding redirect loop for non-local connections
• [settings]Disable browser password autofill for env fields
• [gateway]Reconnect WebSocket after remote onboarding auth
• [docker]Bind 0.0.0.0 and expose CA cert port for TLS setup
• [ci]Gate release builds on E2E tests and fix onboarding-auth flake
• [ci]Gate release builds on E2E tests and fix onboarding-auth regex
• [docker]Add chromium and demote codex token warning to debug
• [scripts]Skip --all-features for clippy on macOS without CUDA
• [scripts]Allow dirty working tree in local-only validation
• [exec]Use host data dir when no container runtime is available
• [onboarding]Update low memory warning message
• [ui]Rename "provider" to "LLM" in navigation and headings
• [gateway]Preserve chat stream event order
• [auth]Enforce auth after passkey/password setup (#93)
• [e2e]Make NoopChatService::clear() succeed and handle no-provider errors in tests
• [config]Resolve lock reentrancy deadlock in update_config
• [onboarding]Defer WS connection until auth completes
• [config]Prevent SOUL.md re-seeding after explicit clear
• [onboarding]Update test for save_soul empty-file behavior
• [ci]Restore node cache hardening and gate on zizmor
• [onboarding]Keep voice step visible in auth-gated flow
• [gateway]Stabilize anthropic onboarding model selection
• Pin deploy templates to explicit versions and auto-update on release (#98)
• [ci]Sync workspace version in Docker build
• [gateway]Update session preview immediately after clear
• [onboarding]Persist selected llm models on continue
• [agents]Cap tool-call IDs to OpenAI 40-char limit
• [config]Isolate stale-key test from env overrides
• [gateway]Include saveProviderKey extraction in provider-validation
• [e2e]Handle pre-configured providers in onboarding tests
• [gateway,tools]Fix sessions.patch camelCase, streaming artifacts, and host working dir
• [tools]Gate macOS-only sandbox helpers with cfg to fix dead-code errors on Linux CI
• [tools]Browser tool falls back to host when no container runtime
• [agents]Add memory_save and memory-anchor hints to system prompt
• [agents]Guide memory_save toward topic files to keep context lean
• [gateway]Fix disabled button rendering and add validation progress hints
• [gateway]Restore voicePending state on session reload
• [tools]Is_sandboxed() returns false when no container runtime exists
• [gateway]Constrain sandbox warnings to max-w-form width
• [memory]Normalize embeddings base_url and add disable_rag mode (#147)
• [gateway]Hide crons/heartbeat submenu when embedded in settings
• [cron]Use data_dir for job storage instead of hardcoded ~/.clawdbot/cron
• [plugins]Set working_dir on shell hooks so relative paths resolve correctly
• [gateway]Improve onboarding and provider validation flows
• [gateway]Allow openrouter model discovery without model id
• [chat]Retry provider 429s across web and channel flows (#149)
• [gateway]Queue channel replies per message with followup default (#150)
• [onboarding]Restore saved model selection and stabilize Anthropic e2e
• [tools]Harden apple container sandbox workdir bootstrap
• [cron]Normalize shorthand cron tool inputs
• [e2e]Avoid response listener race in sandboxes test
• [chat]Compact based on next-request context pressure (#166)
• [agents]Require explicit /sh for forced exec fallback (#161)
• [map]Support show_map multipoints and grouped map links (#168)
• [config]Make agent loop limit configurable and sync docs
• [skills]Support marketplace skills[] repositories
• [onboarding]Persist browser timezone and expose prompt today
• [web-search]Make DuckDuckGo fallback opt-in
• [gateway]Align show_map listing ratings to the right
• [prompt]Clarify sandbox vs data_dir path semantics
• [agents]Append runtime datetime at prompt tail
• [web-search]Load runtime env keys and robustify brave parsing
• [prompt]Preserve datetime tail ordering and add profile plan
• [mcp]Strip internal metadata from MCP tool call arguments
• [crons]Fix modal default validation and form reset on schedule change (#181)
• [gateway]Deduplicate voice replies on Telegram channels (#173)
• [browser]Centralize stale CDP connection detection (#172)
• [browser]Enable profile persistence on Apple Container
• [terminal]Force tmux window resize on client viewport change
• [telegram]Default DM policy to allowlist for new channels
• Update expired Discord invite link
• [telegram]Dispatch voice messages with empty transcription to chat
• [gateway]Broadcast voice audio filename over WebSocket for real-time playback
• [voice]Improve voice reply prompt and audio player placement
• [voice]Deliver TTS audio even when text was streamed
• [tools]Don't register web_search tool when API key is missing
• [agents]Don't emit regular text as ThinkingText when alongside tool calls
• [gateway,telegram,voice]Correct web UI stuck on thinking dots, table rendering, and TTS
• [telegram]Use vertical card layout for wide tables, deliver logbook for streamed voice
• [telegram]Send activity logbook as raw HTML, not markdown
• [e2e]Update share page nonce assertion after script externalization
• [cli]Skip jemalloc on Windows
• [gateway]Redirect plain HTTP to HTTPS on TLS port and optimize metrics SQLite
• [agents]Register DeepSeek via OpenAI-compatible provider for tool support
• [gateway]Sync cron modal signals when editing a job
• [providers]Prefer subscription OAuth and expose Codex onboarding
• [gateway]Surface insufficient_quota without retries
• [voice]Reuse OpenAI LLM API key for TTS and STT providers (#198)
• [ci]Only tag Docker image as latest for highest semver release (#211)
• [e2e]Fix provider selector and onboarding voice step tests
• [e2e]Use exact text match for onboarding provider row selection
• [graphql]Fix session id type, missing ok fields, chat routing, and add session.active query (#218)
• [cron]Document delivery fields in tool description and schema (#213)
• Pass speaking_rate, pitch and speed from config to voice providers (#212)
• [cli]Forward feature flags to moltis-web optional dependency
• [gateway]Harden reverse-proxy tls and websocket handling (#230)
• [cli]Skip jemalloc on linux/aarch64 (#229)
• [gateway,web]Normalize service error mapping and nav expectations
• [e2e]Rebuild gateway when startup binary is stale
• [e2e]Harden session creation wait and anthro cargo env
• [e2e]Handle no-provider chat state in agent specs
• [e2e]Skip openclaw step in onboarding openai flow
• [import]Resolve OpenClaw workspace from config and cross-machine paths
• [openclaw-import]Collapse nested if chains for clippy
• [docker]Use modern Docker CLI in published image (#238)
• [web]Update qmd package references to @tobilu/qmd
• [ci]Exclude swift-bridge from coverage and increase E2E poll timeout
• [e2e]Match error code instead of localized message in full context tests
• [gateway]Remove duplicate /certs/ca.pem route in start_gateway
• [ci]Unblock clippy and macOS bridge wasm build
• [oauth]Make OpenAI Codex callback port 1455 work in Docker (#258)
• [tests]Accept restricted-host auto sandbox backend
• [ios]Add missing .none case in NWTXTRecord switch
• [web]Read DOM value in identity blur handlers to prevent stale closure skip
• [ios]Add cancel button to connection banner for unreachable servers
• [agents]Resolve clippy collapsible-if and expect-used lints
• [ci]Add WASM component build step to all release jobs
• [ci]Parallelize macOS app with release package builds
• [release]Ship v0.10.2 packaging fixes
• [sandbox]Make apple container keepalive portable (#269)
• [local-llm]Combine compile-time and runtime Metal detection
• [auth]Auto-detect new WebAuthn hosts and prompt passkey refresh (#268)
• [web]Replace rg with grep in changelog guard and deduplicate passkey status refresh
• [web]Lazy-load Shiki to prevent blocking page mount
• [web]Fix Shiki highlighter init failures in E2E tests
• [web]Make thinking stop button smaller with left spacing
• [chat]Surface error when LLM returns empty response with zero tokens
• [providers]Emit StreamEvent::Error on non-success finish_reason
• [e2e]Make sandboxes container tests deterministic
• [e2e]Replace remaining racy waitForResponse with route interceptors
• [mcp]Make optional MCP tool params nullable to prevent empty string errors (#283)
• [provider-setup]Reorder validation probes to prefer fast models (#280)
• [sandbox]Resolve host gateway IP for Podman < 5.0 (#287)
• [e2e]Fix flaky "deleting unmodified fork" test
• [ci]Stale lockfile, missing Tailwind in macOS job, OAuth e2e setup
• [ci]Use standalone Tailwind binary for macOS app job
• [e2e]Fix OAuth token-exchange failure test and add error-context capture
• [web]Auto-install node_modules in Tailwind build script
• [web]Retry openclaw onboarding scan until ws is ready
• [ci]Add Tailwind CSS build step to release workflow, Dockerfile, and snapcraft
• [e2e]Wait for session history render before DOM injection in chat-abort
• [ci]Harden tailwindcss cli downloads
• [swift-bridge]Stabilize gateway migration and stream tests
• [config]Support provider url alias for remote Ollama config (#299)
• [ci]Make release dry-run job conditions valid
• [providers]Use Ollama capabilities for tool support detection (#301)
• [scripts]Roll back heavy local validation parallelism
• [web]Skip npm install when TAILWINDCSS binary is provided
• [ci]Update website/releases.json on release
• [web]Add missing i18n button key for preferredModels
• [local-llm]Use sampler API for mlx-lm >= 0.20
• [gateway]Break redirect loop when onboarded but auth not configured (#310)
• [gateway]Reduce idle CPU from metrics loop and log broadcast feedback
• [gateway]Speed up startup by deferring tailscale and log scan
• [gateway]Improve browser warmup integration
• [scripts]Run local nextest with ci timeout profile
• [ci]Build macOS app arm64 in fast path
• [web]Move session history off websocket and cap payload size
• [web]Use combo select for session header selectors
• [web]Externalize SVG icons and restore empty-chat centering
• [web]Align e2e with controls modal and daily model refresh
• [ci]Stage wasm assets for cargo-deb packaging
• [packaging]Use cli-relative web assets in cargo-deb
• Install rustls CryptoProvider before channel startup (#336)
• [ci,tools]Unblock dependabot and support wasmtime 36
• [auth]Honor forwarded host for proxy session cookies
• [config]Include tmux in default sandbox packages
• [mdns]Use stable host label to avoid mDNSResponder conflict and double-.local suffix (#349)
• [web]Prevent Enter key from triggering actions during IME composition (#341)
• [biome]Update schema to 2.4.6 and move noUnusedExpressions to suspicious
• [ci]Update biome version to 2.4.6 in CI workflows
• [macos]Extract makeTextView to fix function body length lint violation
• [providers]Report compatible client_version for Codex model discovery (#359)
• [prompt]Omit sandbox/node info from runtime prompt when disabled (#362)
• [web]Allow deleting cron sessions from chat sidebar (#357)
• [chat]Skip duplicate text fallback when TTS disabled and voice streamed (#373)
• [web]Break redirect loop when accessing via Tailscale Serve (#356)
• Node WebSocket connection and UI connection string (#382)
• [config]Write IDENTITY.md and SOUL.md to agents/main/ instead of root (#384)
• [auth]Bypass auth for local API requests during onboarding (#386)
• [whatsapp]Sled persistence, graceful shutdown, and review fixes (#387)
• [cron]Add delay_ms to avoid LLM computing absolute timestamps (#377)
• [gateway]Retain proxy shutdown sender to prevent immediate proxy exit (#368)
• [agents]Include tool_result messages in LLM conversation history (#389)
• [sandbox]Auto-detect host data dir for docker-in-docker (#396)
• [chat]Compact the active channel session (#399)
• [providers]Strip stop tokens from MLX streaming output (#397)
• [config]Support legacy memory embedding keys (#400)
• [web]Address installation feedback from user testing (#398)
• [tools]Harden apple container bootstrap execs (#405)
• [telegram]Strip HTML tags from plain fallback (#404)
• [web]Clarify cron setup modal copy (#409)
• [web]Keep onboarding accessible after auth reset (#415)
• [web]Improve onboarding password autofill hints (#406)
• [chat]Persist aborted partial history (#418)
• [agents]Retry empty structured tool names (#410)
• [tools]Make sandbox cfg gates consistent for cross-platform CI
• [local-llm]Restore custom GGUF setup without restart (#417)
• [browser]Scope cached browser sessions per chat (#412)
• [browser]Align sandbox browserless timeout with pool lifecycle (#403)
• [providers]Keep minimax system messages in history
• [tools]Only expose exec node parameter when nodes are connected
• [gateway]Address greptile review feedback
• [gateway]Create heartbeat cron job on update when missing
• [gateway]Address review feedback on heartbeat cron job creation
• [agents]Sanitize model-mangled tool names from parallel calls
• [agents]Pass sanitized tool name to hook dispatch
• [agents]Add suffix-stripping invariant comment and functions_ edge case test
• [mcp]Fix JS bugs in display name edit flow
• [mcp]Add missing display_name field to doctor test structs
• [mcp]Destructure props in renderServerName and send null for blank display_name
• [tools]Replace hand-rolled html_to_text with html2text crate
• [tools]Address PR review for html_to_text
• [tools]Address second round of PR review
• [tools]Collapse consecutive blank lines, fix doc comment
• [sessions]Use write(true)+seek instead of append(true) for fd_lock on Windows
• [release]Update conditions for jobs to handle dry-run scenarios correctly
• [oauth]Use correct OpenAI client_id, add config layer and tests
• [agents]Prefer tool-capable provider when tools are registered
• [agents]Register builtin providers before genai to enable tool calling
• [gateway]Auto-focus chat input on page load
• [sessions]Add missing sandbox_enabled column migration, stop swallowing errors
• [gateway]Move sandbox button next to model select, fix default state
• Resolve all clippy warnings (collapsible_if, map_flatten, or_default, unused imports)
• [gateway]Open nav panel by default
• [gateway]Add session icons and replace spinner in-place
• [gateway]Add missing renderCompactCard export to page-chat.js
• [gateway]Fix routing, styling, and layout issues in JS modules
• [telegram]Remove channel prefix from user messages and log LLM responses
• [ci]Fix biome version and worktree test default branch
• [ci]Restore biome version pin and fix cargo fmt
• Update CLAUDE.md pre-commit checks and fix clippy warnings
• [ci]Update setup-biome to v2.7.0 and fix sort_by_key clippy lint
• [tests]Configure git user in worktree test setup for CI
• Replace sort_by with sort_by_key in project store
• [assets]Fix biome formatting and empty catch block in sandbox.js
• [ui]Match settings nav item height and style to main navigation
• [security]Wrap API keys in Secret<String> for memory subsystem
• Use RecommendedCache for cross-platform watcher, sort_by_key for clippy
• [hooks]Ignore broken pipe on stdin write when child doesn't read it
• [ui]Show URL and query in web tool call cards
• [sandbox]Restart stopped Apple containers in ensure_ready
• [sandbox]Pass sleep infinity to Apple Container run
• [sandbox]Promote exec routing and container logs to info level
• [sandbox]Handle Apple Container inspect returning empty for nonexistent containers
• [sandbox]Default working_dir to "/" when running inside container
• [security]Reject cross-origin WebSocket upgrades (CSWSH protection)
• [sandbox]Inject env vars in Apple Container backend
• [exec]Redact env var values from command output
• [exec]Redact base64 and hex encoded env var values from output
• [config]Use correct defaults for ToolsConfig when not deserialized
• [gateway]Nav badge counts and cron page instant load
• [memory]Write memory files to data dir instead of cwd
• [gateway]Eliminate identity flash on page load
• [ui]Use consistent button class for skills Remove button
• [ui]Fix skills Disable button style and add error feedback
• [skills]Route disable to correct RPC for plugin-sourced skills
• [ui]Use inline style for tailscale status green dot
• [ui]Use status-dot connected class for tailscale green dot
• [ui]Replace Tailwind arbitrary classes with inline styles in tailscale templates
• [ui]Make tailscale status bar fit-content width
• [ui]Make info-bar full width to match warning/error divs
• [ui]Constrain info-bar to max-width 600px matching alert divs
• [ui]Add alert-info-text to shared alert base styles
• [ui]Add btn-row and btn-row-mt CSS classes for button spacing
• [ui]Space cancel/create buttons apart and normalize height
• [ui]Improve tailscale loading message to set expectations
• [tailscale]Open Tailscale.app on macOS instead of running tailscale up
• [ui]Preserve funnel security warning when rebuilding tailscale DOM
• [ui]Use alert-warning-text style for funnel auth warning
• [ui]Replace auth text with green button linking to security settings
• [ui]Move auth button below the funnel security warning panel
• [ui]Remove @layer components wrapper to fix nav specificity
• [ui]Update session list after /clear and select next on delete
• [ui]Format environment variable timestamps with luxon
• [gateway]Stop echoing web UI messages to Telegram channel
• Collapse nested if statements in config loader
• [ci]Fix package build failures
• [ci]Correct cargo-deb assets order
• [local-llm]Detect available package managers for MLX install
• [local-llm]Detect mlx-lm installed via brew
• [ui]Show install commands on separate lines
• [local-llm]Rename provider from local-gguf to local-llm
• [local-llm]Fix HuggingFace API response parsing
• [ui]Show searching state on HuggingFace search button
• [ui]Close modal and prevent multiple clicks on HF model selection
• [chat]Remove per-message tools warning broadcast
• [ui]Remove scroll from local model list
• [ci]Install cmake for llama.cpp build and fix Biome errors
• [ci]Fix local-llm build in CI
• [migrations]Use set_ignore_missing to allow multi-crate migrations
• [cli]Run gateway migrations from db migrate command
• [telegram]Don't delete channel from database on conflict
• [gateway]Load images nav count async to avoid blocking page serve
• [gateway]Replace branch icon, fix fork UX issues
• Streaming tool calls, skill directory, skills page UX
• [skills]Handle disable for personal/project skills
• [skills]Use app modal for delete confirmation instead of system confirm()
• [skills]Danger button on delete modal, fix disable routing for SkillsService
• [graphql]Add skill_save to MockSkills trait impl
• [gateway]Warm container_cli OnceLock at startup via spawn_blocking
• [security]Prevent stored XSS via HTML inline and block script extensions
• [e2e]Use correct event bus dispatch for send_document tests
• [e2e]Use system-event RPC to dispatch tool events in send_document tests
• [sandbox]Fix clippy CI failures on Linux
• [sandbox]Gate apple-only tests with cfg(target_os = "macos")
• [e2e]Mock slow sandbox APIs in container tests
• [e2e]Make send_document icon test more robust on CI
• [mcp]Address review feedback on is_alive and timeout field naming
• [mcp]Reorder config update to write memory before disk
• Add missing request_timeout_secs field and allow expect in test
• [agents]Address PR review feedback for lazy tool registry
• [e2e]Mock sandbox backend in container e2e tests
• [e2e]Add afterEach unrouteAll to sandbox tests
• [auth]Add Secure attribute to session cookie when TLS is active
• [tools]Make sandbox off test explicit
• [ci]Replace glslc with glslang-tools for Ubuntu 22.04
• [providers]Address PR #408 review feedback
• [ci]Install glslc from LunarG Vulkan SDK on Ubuntu 22.04
• [whatsapp]Improve discoverability and debug logging (#460)
• [whatsapp]Address PR review feedback
• [channels]Redact secrets in channel config API responses
• [channels]Address PR review feedback
• [channels]Compute serialize_struct field counts dynamically
• [providers]Update stale MiniMax context-window comment to include M2.7
• [tools]Ignore exec node param when no nodes are connected (#427)
• [tools]Error on configured default_node when disconnected, not silent fallthrough
• [anthropic]Document system message merging for alternating-role constraint
• [chat]Re-inject datetime in context-overflow retry path
• [browser]Set world-writable permissions on container profile directory
• [browser]Replace expect() with ? in test to satisfy clippy
• [server]Support IPv6 bind addresses (#447)
• [server]Address PR review feedback on IPv6 bind fix
• [server]Include bind address in oauth error, assert address family in test
• [import]Preserve config template comments during OpenClaw import
• [import]Assert on import report in comment-preservation test
• [httpd]Address greptile review findings on PR #465
• Address second round of greptile review on PR #465
• [node-host]Update systemd unit test for quoted ExecStart args
• [web]Allow delete button on cron sessions
• Address greptile review round 3 on PR #465
• [httpd]Align metrics history window and cleanup interval with gateway
• [providers]Address PR review — double-Done, missing ProviderRaw, URL normalization
• [tools]Rescue stringified JSON and flat params in cron tool
• [tools]Exclude "patch" and "job" from flat-param rescue keys
• [ci]Replace non-existent glslc package with glslang-tools
• [docker]Use Node 22 LTS via NodeSource, persist npm cache
• [docker]Avoid silent curl failure, cache npx in container example
• [docker]Use GPG key + apt source instead of NodeSource setup script
• [docker]Ensure /etc/apt/keyrings exists before NodeSource GPG import
• [web]Await event subscriptions before accepting broadcasts
• [ci]Install vulkan-sdk from LunarG instead of Ubuntu's old libvulkan-dev
• [dev]Make local checks OS-aware for LLM backends
• [dev]Merge main and update OS-aware local-LLM checks
• [dev]Address PR review feedback
• [skills]Use slug as fallback when skill name fails validation
• [skills]Address PR review feedback on slug fallback
• [skills]Address second round of PR review feedback
• [skills]Use tempdir in slug fallback error tests for isolation
• [skills]Improve slug error messages and test isolation
• [gateway]Suppress update banner for dev builds