fix/windows gateway encoding v2

[XiaoXiao0221]

• test(tools): add unit tests for tool_backend_helpers module
• test(tools): add unit tests for budget_config module
• docs: fix ASCII diagram width mismatch in architecture.md
• fix: enable Matrix Reactions in platform comparison table
• fix: add gpt-5.4 and gpt-5.4-mini to openai-codex curated model list (fix: add gpt-5.4 to openai-codex model picker #7670)
• fix: honor session-scoped gateway model overrides
• feat: background process monitoring — watch_patterns for real-time output alerts
• fix(qwen): correct context lengths for qwen3-coder models and send max_tokens to portal
• fix(streaming): adaptive backoff + cursor strip to prevent message truncation (fix(streaming): adaptive backoff + cursor strip to prevent message truncation #7683)
• fix: unify openai-codex model list — derive from codex_models.py (fix: add gpt-5.3-codex-spark to openai-codex curated model list #7844)
• fix(matrix): pass required args to MemoryCryptoStore for mautrix ≥0.21 (fix(matrix): pass required args to MemoryCryptoStore for mautrix ≥0.21 #7848)
• fix: resolve three high-impact community bugs (Bot connects and syncs old messages, but silently ignores all new messages (no log output) #5819, [Bug] Error code 200340 when approving dangerous commands via Feishu/Lark gateway #6893, [Bug]: Copilot GPT-5.4 intermittently routes to OpenRouter /chat/completions and fails with unsupported_api_for_model #3388) (fix: resolve three high-impact community bugs (#5819, #6893, #3388) #7881)
• fix(vision): auto-resize oversized images, increase default timeout, fix vision capability detection
• docs: add warning about summary model context length requirement (docs: warn that compaction summary model must match main model context length #7879)
• feat(xiaomi): add Xiaomi MiMo as first-class provider
• docs: add Xiaomi MiMo to all provider docs + fix MiMo-V2-Flash ctx len
• refactor(auxiliary): config.yaml takes priority over env vars for aux task settings (refactor(auxiliary): config.yaml takes priority over env vars #7889)
• fix(migration): update OpenClaw migration for schema drift
• feat(migration): preview-then-confirm UX + docs updates
• fix(vision): preserve aspect ratio during auto-resize
• fix(weixin): keep multi-line messages in single bubble by default (fix(weixin): keep multi-line messages in single bubble by default #7903)
• feat: warn at session start when compression model context is too small (feat: warn at session start when compression model context is too small #7894)
• fix(gateway): propagate user identity through process watcher pipeline
• fix(api-server): keep chat-completions SSE alive
• fix: detect truncated tool_calls when finish_reason is not length
• fix(agent): prevent false thinking-exhaustion for non-reasoning models
• fix(gateway): break stuck session resume loops on restart ([Gateway] Stuck session resumes on restart — creates unrecoverable loop #7536)
• fix: update thinking-exhaustion test for think-tag gating
• fix(gateway/weixin): ensure atomic persistence for critical session state
• fix(wecom): handle appmsg attachments (PDF/Word/Excel) from WeCom AI Bot
• refactor: extract shared helpers to deduplicate repeated code patterns (refactor: extract shared helpers to deduplicate repeated code patterns #7917)
• fix(weixin): add per-chunk retry with backoff for text delivery
• fix: reap orphaned browser sessions on startup (fix: reap orphaned browser sessions on startup #7931)
• fix: scope tool interrupt signal per-thread to prevent cross-session leaks (fix: scope tool interrupt signal per-thread to prevent cross-session leaks #7930)
• fix(whatsapp): pin Baileys to fix/abprops-abt-fetch for bad-request fix
• add basic twilio signature checking and tests
• handle port variants in Twilio signatures
• change Twilio signature verification from opt-in to opt-out
• update docks with changes made
• remove unused import and fix misleading log
• add new test covering edge case where both insecure_no_sig and _webhook_url are set
• fix: mock aiohttp server in startup guard tests to avoid port binding
• fix(custom-providers): propagate model field from config to runtime so API receives the correct model name
• fix: propagate model through credential pool path + add tests
• fix(tools): prevent command argument injection and path traversal in checkpoint manager
• fix(tools): neutralize shell injection in _write_to_sandbox via path quoting (fix(tools): neutralize shell injection in _write_to_sandbox via path quoting #7940)
• fix(gateway): use source.thread_id instead of undefined event in queued response
• fix: deduplicate reasoning items in Responses API input
• test: add dedup coverage for reasoning item ID deduplication
• fix(gateway): preserve queued voice events for STT
• fix: scope gateway status to the active profile
• fix: add all_profiles param + narrow exception handling
• feat(gateway): add all missing platforms to interactive setup wizard (feat(gateway): add all missing platforms to interactive setup wizard #7949)
• fix: is_local_endpoint misses Docker/Podman DNS names (fix: is_local_endpoint misses Docker/Podman DNS names #7950)
• fix(cron): pass skip_context_files=True to AIAgent in run_job (fix(cron): pass skip_context_files=True to AIAgent in run_job #7958)
• fix(claw): warn if gateway is running before migrating bot tokens
• fix: normalize checkpoint manager home-relative paths
• feat(gateway): add WeCom callback-mode adapter for self-built apps
• fix: reset compression_attempts and primary_recovery_attempted on fallback activation
• fix: remove dead hasattr checks for retry counters initialized in reset block
• feat: add --env and --preset support to hermes mcp add
• fix(gateway): add HERMES_SESSION_KEY to session_context contextvars
• fix(cli): add ChatConsole.status for /skills search
• docs: add platform adapter developer guide + WeCom Callback docs (docs: add platform adapter developer guide + WeCom Callback docs #7969)
• fix: always log outer loop exception traceback at DEBUG level
• fix: prevent agent from stopping mid-task — compression floor, budget overhaul, activity tracking
• docs: expand tool-use enforcement documentation (docs: expand tool-use enforcement documentation #7984)
• feat(gateway): surface natural mid-turn assistant messages in chat platforms
• fix(tools): enforce ID uniqueness in TODO store during replace operations
• fix: use ceiling division for token estimation, deduplicate inline formula
• feat(nix): container-aware CLI — auto-route into managed container (feat(nix): container-aware CLI — auto-route into managed container #7543)
• feat: standardize message whitespace and JSON formatting
• refactor: remove budget warning injection system (dead code)
• feat(cli): add native /model picker modal for provider → model selection
• refactor(terminal): remove check_interval parameter (refactor(terminal): remove check_interval parameter #8001)
• feat: per-platform display verbosity configuration (feat: per-platform display verbosity configuration #8006)
• feat: component-separated logging with session context and filtering (feat: component-separated logging with session context and filtering #7991)
• perf(ssh,modal): bulk file sync via tar pipe and tar/base64 archive (perf(ssh,modal): bulk file sync via tar pipe and tar/base64 archive #8014)
• fix(matrix): replace pickle crypto store with SQLite, fix E2EE decryption (fix(matrix): replace pickle crypto store with SQLite, fix E2EE decryption #7981)
• feat: add hermes backup and hermes import commands (feat: add hermes backup and hermes import commands #7997)
• fix(discord): decouple readiness from slash sync
• feat: /compress — guided compression with focus topic (feat: /compress <focus> — guided compression with focus topic #8017)
• fix(gateway): add missing RedactingFormatter import
• fix: improve context compaction to prevent model answering stale questions (fix: improve context compaction to prevent model answering stale questions #8107)
• fix(gateway): Windows compatibility fixes

[XiaoXiao0221]

Superseded by #8180 and #8563.