The Project
Chat RAG is a Retrieval-Augmented Generation platform that transforms static documents into interactive knowledge bases. Users upload PDFs or text files, and the system creates vector embeddings that enable AI-powered conversations with accurate source references.
Key Features
- Document upload & processing — PDFs and text files are chunked, embedded, and indexed automatically
- Semantic search via pgvector — finds relevant document passages using vector similarity
- Source citations — every AI response includes references to the specific document sections used
- Multi-user chat — each user has their own document library and conversation history
- Real-time streaming — AI responses stream token-by-token via WebSockets
- Role-based access — superadmin manages all documents, users manage their own
Technical Highlights
Rails 8.1 with PostgreSQL + pgvector extension for vector storage and similarity search. RubyLLM for LLM orchestration (OpenAI, OpenRouter). tiktoken_ruby for accurate token counting. Background document processing via Solid Queue. Real-time streaming via Turbo Streams.