lex-rag

RAG (Retrieval-Augmented Generation) system over Brazilian federal legislation.

Built with Laravel 11 + PostgreSQL/pgvector + OpenAI.

Status: 🚧 Active development. Currently in prototype phase.

Architecture

1. Ingestion — PDF/text → chunks → embeddings → pgvector
2. Retrieval — query → embedding → similarity search
3. Generation — context + question → LLM → answer with sources

Tech stack

• Laravel 11 (PHP 8.3)
• PostgreSQL + pgvector
• OpenAI API (text-embedding-3-small + gpt-4o-mini)
• Docker

Status

• RAG prototype working in Python notebook
• Laravel API structure
• Ingestion pipeline
• Query endpoint
• OpenAPI documentation

License

MIT