Chat with Email History via Telegram, Mistral & Pgvector RAG

This n8n workflow transforms your email inbox into a conversational AI assistant. Users send natural language queries through Telegram, like 'What hotel did I stay in last summer?' or 'What were my semester grades?'. It leverages Retrieval-Augmented Generation (RAG) with vector similarity searches on embeddings stored in Pgvector, combined with structured SQL queries for precise results.

Built for privacy and cost-efficiency, it runs 100% locally using open-source tools: Ollama for LLM (Mistral compatible), nomic-embed-text for embeddings, and Postgres with Pgvector extension as the vector database. No cloud dependencies mean full data control and no API costs. It integrates semantic (fuzzy, context-aware) and structured (exact filters) searches, outperforming basic keyword tools.

Setup is straightforward: Use companion workflows to ingest Gmail emails into Pgvector embeddings and enable SQL composition. Customize Telegram credentials, AI models, and database connections, then activate. Ideal for personal knowledge retrieval, travel recaps, expense tracking, or professional correspondence analysis.

Benefits include time savings on manual email hunts, advanced AI without enterprise setups, and scalability for large inboxes. Perfect for individuals or small teams seeking a bleeding-edge, customizable email RAG system via n8n's no-code power.