Features
- Thinking mode support: Parse
<think>/<thinking>/<thought>tags from LLM output and display reasoning in WeCom's collapsiblethinking_contentfield - Passive reply thinking UI: First sync response shows thinking mode UI immediately via
thinking_contentfield - Markdown fallback: response_url fallback now sends
msgtype: "markdown"instead oftextfor richer formatting
Fixes
- Stream routing: Fix concurrent message race by checking
ctxStream.finishedbefore reusing async-context stream; fall back to latest recoverable stream - Agent inbound whitelist: Filter non-message event types (subscribe/unsubscribe etc.) to prevent them from triggering LLM replies
- MEDIA regex: Match only line-start
^MEDIA:directives to align with upstream OpenClaw behavior - Grace timer: Reduce post-dispatch grace timer from 3000ms to 200ms for faster stream finalization
- Remove auto-detect /workspace/ paths: Remove overly aggressive workspace path auto-detection in outbound delivery; rely on upstream MEDIA directives and payload.mediaUrls instead
Docs
- Add table of contents navigation to README
- Add streaming capabilities documentation (Markdown, thinking mode, images)
- Add CHANGELOG.md
Install
npm install @sunnoy/wecom@1.7.0