Ragbits is an open-source Python framework from deepsense.ai that provides building blocks for rapidly developing generative AI applications. Instead of another monolithic RAG wrapper, it offers modular, typed components — document ingestion, vector-store access, prompt management, guardrails, evaluation and agent primitives — that snap together into production LLM systems, with a matching UI for chatting with and testing what you built. The project has around 1,700 stars and is actively maintained, with code last updated in 2026.
Ragbits grew out of deepsense.ai’s client work building RAG systems, and it shows: the abstractions cover the unglamorous parts teams repeatedly rebuild, like source ingestion pipelines, prompt versioning, response evaluation and optimisation loops. Components are designed to be swapped — change vector store or model provider without rewriting the application — and everything is typed and testable rather than notebook glue.
Free. Ragbits is open source under the MIT license; you self-host it and pay only for any model APIs you plug in.
Python developers building RAG and GenAI applications who want reusable, production-minded components instead of starting from a blank LangChain script.
The framework’s components are designed to compose: a document-search application, for example, can reuse the same ingestion and retrieval blocks as an agent workflow, which keeps prototypes and production systems on one codebase.