v0.0.5-beta.1 Release Note

✨ Added

• Channel Message Filtering: Added per-channel configuration to hide tool execution steps and show only final results, improving readability in chat channels like DingTalk and Feishu (#449)
• DingTalk Rich Text Images: Added support for images in rich text messages (#471)
• DingTalk Message Deduplication: Implemented message deduplication using message ID tracking and content hashing to prevent duplicate processing when DingTalk resends messages (#489)
• MCP Client Auto-Recovery: Added automatic reconnect/rebuild for closed MCP sessions with graceful fallback; prevents session state overwrite on failures (#431)
• Roadmap Documentation: Added project roadmap and community contribution guidelines to website and README with bilingual support (#428, #439)

🔄 Changed

• Memory System: Upgraded to reme-ai 0.3.0.1 with vLLM embedding support for better memory compaction and token management (#383)
• OpenAI Model IDs: Corrected model list by replacing invalid gpt-5-chat with gpt-5 and adding gpt-5.2 (#498)
• Browser Tool: Added system browser detection with automatic fallback to WebKit on macOS; improved installation workflow with better error messages (#451)
• Provider Status Logic: Simplified provider status with unified available/unavailable status, removing redundant has_api_key field (#442)
• Skill Hub: Enhanced error handling for GitHub API with better rate limit hints and smarter default branch detection (#506)

🐛 Fixed

• Configuration Persistence (Critical for Docker users):
  • Fixed configuration loss on container restart by moving providers.json and envs.json to persistent SECRET_DIR (~/.copaw.secret) with automatic migration (#346, #345)
  • Applied security hardening with restrictive permissions (dir: 0700, file: 0600)
• DingTalk Channel: Fixed webhook handling to prevent stale session reuse (#446)
• Feishu Channel: Fixed cron task message delivery by correctly attaching channel metadata and receiver IDs (#464)
• OpenAI-Compatible Providers:
  • Fixed multi-turn chat failures by removing unsupported top-level name fields from messages (#452)
  • Fixed crashes from malformed streaming tool-call chunks by adding stream sanitizer (#488)
• Tool System: Fixed "already registered in toolkit" errors with automatic duplicate handling (#438)
• CLI Commands: Removed redundant provider validation in config key command (#457)

👥 Contributors

Special thanks to all contributors in this release:

@jinliyl, @gongpx20069, @zhijianma, @xieyxclack, @rayrayraykk, @cuiyuebing, @lalaliat, @fancyboi999

Full Changelog: v0.0.4...v0.0.5-beta.1