1c8d45c31b6fcd0da04942b1c480090e73741c43
- Memory service: asyncpg pool auto-reconnect on connection loss, IVFFlat lists 10→100 - Bot: parallel RAG/memory/chunk queries (asyncio.gather), parallel tool execution - Bot: skip memory extraction for trivial messages (<20 chars, no personal facts) - Bot: persist voice call transcripts as searchable conversation chunks - RAG: JSON parse safety in AI metadata, embedding_status tracking, fetch timeouts - Drive sync: token refresh mutex to prevent race conditions, fetch timeouts Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Description
Matrix AI voice agent - LiveKit + LiteLLM
Languages
Python
99.1%
Dockerfile
0.8%
Shell
0.1%