When mention_only=true is set, the bot should not respond to non-text
messages (photos, documents, videos, stickers, voice) in group chats
unless the caption contains a bot mention.
Changes:
- Add mention_only check in try_parse_attachment_message() for group messages
- Check if caption contains bot mention before processing
- Skip attachment if no caption or no mention
- Add mention_only check in try_parse_voice_message() for group messages
- Voice messages cannot contain mentions, so always skip in groups
- Add unit tests for the new behavior
Fixes#1662
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add background sub-agent orchestration tools that extend the existing
delegate tool with async execution, session tracking, and lifecycle
management.
New tools:
- subagent_spawn: Spawn delegate agents in background via tokio::spawn,
returns session_id immediately. Respects security policy, depth limits,
rate limits, and configurable concurrent session cap.
- subagent_list: List running/completed/failed/killed sessions with
status filtering. Read-only, allowed in all autonomy modes.
- subagent_manage: Kill running sessions via CancellationToken or
query status with partial output. Enforces Act policy for kill.
Shared state:
- SubAgentRegistry: Thread-safe session store using
Arc<parking_lot::RwLock<HashMap>> with lazy cleanup of sessions
older than 1 hour. Tracks session metadata, status, timing, and
results.
Test coverage: 56 tests across all 4 modules covering happy paths,
error handling, security enforcement, concurrency, parameter
validation, and edge cases.
No new dependencies added. No existing tests broken.
- lark: convert send to rich-text post format with markdown parsing
- lark: add draft edit throttling and shell polling guidance
- lark: auto-detect receive_id_type from recipient prefix
- lark: deliver heartbeat as interactive card
- lark: use valid Feishu API emoji_type keys for ack reactions
- lark: handle flat post format from WS and add diagnostic logging
- lark: replace unsupported code_inline tag and strip leaked tool blocks
- lark: gate LarkChannel behind channel-lark feature flag
- whatsapp: render WhatsApp Web pairing QR in terminal
- channels: update_draft returns Option<String> for new draft IDs
- config: add draft_update_interval_ms and max_draft_edits to Lark/FeishuConfig
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add query classification support to channel message processing (Telegram,
Discord, Slack, etc.). When query_classification is enabled with model_routes,
each incoming message is now classified and routed to the appropriate model
with an INFO-level log line.
Changes:
- Add query_classification and model_routes fields to ChannelRuntimeContext
- Add classify_message_route function that logs classification decisions
- Update process_channel_message to try classification before default routing
- Initialize new fields in channel runtime context
- Update all test contexts with new fields
The logging matches the existing agent.rs implementation:
- target: "query_classification"
- fields: hint, model, rule_priority, message_length
- level: INFO
Closes#1367
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
`supports_vision` is currently hardcoded per-provider. The same Ollama instance can run `llava` (vision) or
`codellama` (no vision), but the code fixes vision support at the provider level with no user override.
This adds a top-level `model_support_vision: Option<bool>` config key — tri-state:
- **Unset (default):** provider's built-in value, zero behavior change
- **`true`:** force vision on (e.g. Ollama + llava)
- **`false`:** force vision off
Follows the exact same pattern as `reasoning_enabled`. Override is applied at the wrapper layer (`ReliableProvider` /
`RouterProvider`) — no concrete provider code is touched.
## Changes
**Config surface:**
- Top-level `model_support_vision` field in `Config` struct with `#[serde(default)]`
- Env override: `ZEROCLAW_MODEL_SUPPORT_VISION` / `MODEL_SUPPORT_VISION`
**Provider wrappers (core logic):**
- `ReliableProvider`: `vision_override` field + `with_vision_override()` builder + `supports_vision()` override
- `RouterProvider`: same pattern
**Wiring (1-line each):**
- `ProviderRuntimeOptions` struct + factory functions
- 5 construction sites: `loop_.rs`, `channels/mod.rs`, `gateway/mod.rs`, `tools/mod.rs`, `onboard/wizard.rs`
**Docs (i18n parity):**
- `config-reference.md` — Core Keys table
- `providers-reference.md` — new "Ollama Vision Override" section
- Vietnamese sync: `docs/i18n/vi/` + `docs/vi/` (4 files)
## Non-goals
- Does not change any concrete provider implementation
- Does not auto-detect model vision capability
## Test plan
- [x] `cargo fmt --all -- --check`
- [x] `cargo clippy --all-targets -- -D warnings` (no new errors)
- [x] 5 new tests passing:
- `model_support_vision_deserializes` — TOML parse + default None
- `env_override_model_support_vision` — env var override + invalid value ignored
- `vision_override_forces_true` — ReliableProvider override
- `vision_override_forces_false` — ReliableProvider override
- `vision_override_none_defers_to_provider` — passthrough behavior
## Risk and Rollback
- **Risk:** Low. `None` default = zero behavior change for existing users.
- **Rollback:** Revert commit. Field is `#[serde(default)]` so old configs without it will deserialize fine.
(cherry picked from commit a1b8dee785)
The Relation enum in the Matrix SDK is marked as non-exhaustive,
causing a compilation error when building with the channel-matrix feature.
Add a wildcard pattern to handle any future relation types.
Fixes#1693
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Voice messages were being silently ignored when transcription was disabled
or user was unauthorized, making it difficult to diagnose configuration
issues. This change adds:
- Debug log when voice/audio message received but transcription disabled
- Debug log when voice message skipped due to unauthorized user
- Info log on successful voice transcription
Closes#1469
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
parse_attachment_markers used .find(']') which returns the first ']', so
filenames containing brackets (e.g. yt-dlp output 'Video [G4PvTrTp7Tc].mp4')
were truncated at the inner bracket, producing a wrong path and a send failure.
Replace the naive search with find_matching_close, a depth-tracking scanner
that correctly skips nested '[...]' pairs and returns the index of the
outermost closing bracket.
Adds regression tests for the bracket-in-filename case and for the
unclosed-bracket fallback (no match → message passed through unchanged).
format_attachment_content was matching only Photo for [IMAGE:] routing.
Documents with image extensions (jpg, png, gif, webp, bmp) were formatted as
[Document: name] /path, bypassing the multimodal pipeline entirely.
Extend the match arm to cover Document when is_image_extension returns true,
so both Photos and image Documents produce [IMAGE:/path] and reach the provider
as proper vision input blocks.
Adds regression tests covering Document+image extension → [IMAGE:] and
Document+non-image extension → [Document:] paths.
Chummy
argenis de la rosa
dave
Allen Huang
FlashFamily
reidliu41
guitaripod