- Move LLMHelper and related logic from app.helper.llm to app.agent.llm.helper
- Update all imports to reference new LLMHelper location
- Introduce app/agent/llm/__init__.py for internal LLM adapter exports
- Add llm.py API router with endpoints for model listing, provider auth, and test calls
- Remove legacy LLM endpoints from system.py
- Update requirements for langchain-anthropic and anthropic
- Refactor test_llm_helper_testcall.py for async LLMHelper usage and new import paths
- Implement batch AI re-organize endpoint for transfer history with progress tracking
- Add batch_manual_transfer_redo system task template and prompt generation
- Refactor agent_manager to support generic background prompt execution
- Add AIRecommendChain for search result recommendation using agent background prompt
- Update search endpoints to use new AIRecommendChain and remove legacy code
- Enhance test cases for batch manual transfer redo
- Minor code cleanup and style fixes
- Add SYSTEM_INTERNAL_USER_ID constant and helpers to app.utils.identity
- Ensure internal user ID is normalized to None before dispatching notifications, preventing misrouting to external channels
- Refactor MessageChain to use normalization for all message dispatch methods
- Add tests for internal user ID normalization and notification dispatch behavior
- Integrate voice message handling: detect and extract audio references from Telegram and WeChat messages, route to agent with voice reply preference.
- Add voice provider abstraction and OpenAI-based TTS/STT implementation.
- Implement agent tool `send_voice_message` for generating and sending voice replies, with fallback to text if voice is unavailable.
- Extend agent prompt and context to support voice reply instructions.
- Update notification and message schemas to support audio fields.
- Add Telegram and WeChat voice sending logic, including audio file conversion and temporary media upload for WeChat.
- Add tests for voice helper and agent voice routing.
Add /stop_agent command that cancels the currently running agent reasoning
task without clearing the session or memory. Unlike /clear_session which
destroys the entire session, this allows users to stop a long-running or
stuck agent process and continue the conversation afterward.
