vezlo/assistant-server

Implements chunk-based semantic RAG with adjacent retrieval using pgvector and OpenAI embeddings, WebSocket support for real-time multi-agent conversations, and integrates with Slack and external Supabase databases for natural language queries. Features LLM-as-Judge validation, conversation persistence with human handoff workflows, and Knex.js migrations for schema management, all containerized for production deployment.

Available on npm.