GH Repository · superlinear-ai
raglite
🥤 RAGLite is a Python toolkit for Retrieval-Augmented Generation (RAG) with DuckDB or PostgreSQL
- stars
- 1,200
- 30-day movement
- starts with the next reading
- Related entries
- 60
- Connections
- 3
dockerPythonduckdbchainlitquery-adaptercolbertpythonllmpdflate-chunkingpgvectorragevalsretrieval-augmented-generationsqlitepostgresvector-searchpostgresqlrerankingmarkdownlate-interaction
RAGLite is a Python toolkit for building Retrieval-Augmented Generation (RAG) pipelines backed by DuckDB or PostgreSQL. Its topics indicate support for PDF and Markdown ingestion, vector search with pgvector, reranking, late chunking and late interaction (ColBERT), query adaptation, evals, and a Chainlit interface.
It gives you a self-contained Python RAG stack that runs on lightweight DuckDB or an existing PostgreSQL database instead of a dedicated vector store.
Use it to
- Build a RAG pipeline over PDF and Markdown documents
- Run vector search on DuckDB or PostgreSQL with pgvector
- Apply reranking and ColBERT-style late interaction retrieval
- Evaluate retrieval quality with built-in evals
- Prototype a chat interface with Chainlit
For Python developers building RAG applications
- Role
- rag
- Language
- Python
- Licence
- MPL-2.0
- Forks
- 110
- Open issues
- 12
- Last push
- 2026-08-17
- Latest release
- v0.1.0 · 2024-10-07
topicsragpythonvector-searchpostgresduckdbreranking