Introduction
bigRAG is an open-source, self-hostable RAG platform for document ingestion, vector search, and retrieval-augmented generation.
bigRAG is an open-source RAG (Retrieval-Augmented Generation) platform. It provides a complete pipeline for document ingestion, chunking, embedding, and vector search — all behind a simple REST API.
Key Features
- End-to-end RAG pipeline — upload documents, auto-chunk, embed, and search in one platform
- Any document format — PDF, DOCX, PPTX, HTML, Markdown, images (with OCR), and more via Docling
- Any embedding model — OpenAI and Cohere providers with per-collection configuration
- Milvus vector database — production-grade vector search with hybrid capabilities
- Self-hostable — Docker Compose, no external dependencies
- MIT licensed — run it anywhere, forever free
Architecture
| Component | Purpose | Default Address |
|---|---|---|
| bigRAG API | REST API server (FastAPI) | http://localhost:6100 |
| PostgreSQL | Metadata storage | localhost:5433 |
| Milvus | Vector storage and search | localhost:19530 |
| Redis | Ingestion job queue | localhost:6380 |