Langchain-Chatchat

Use it when

• You must offer chat Q and A over internal documents and can configure the entire pipeline to use local inference frameworks, local models, and local stores so document-derived content stays on your infrastructure
• You need a ready-made knowledge base app with a web UI, file upload, and hybrid BM25 plus vector retrieval for Chinese-language documents instead of wiring LangChain, a vector store, and a chat frontend together yourself

What it solves

Not the fit when

• Managed multi-tenant SaaS deployments
• Model fine-tuning or training workflows
• English-first teams wanting a minimal library rather than a full app
• Hosted-provider configurations; local-only inference must be selected when document-derived content cannot leave the infrastructure
• no-cross-session-recall
• pdf-table-garbled
• embedding-recall-poor

Install

pip install langchain-chatchat -U (or docker pull chatimage/chatchat:0.3.1.3-93e2c87-20240829); optionally pip install "langchain-chatchat[xinference]" -U

Invoke

chatchat init to generate yaml config, edit model_settings.yaml to point at a running inference framework (Xinference, Ollama, LocalAI, FastChat, One API), chatchat kb -r to build the knowledge base, then chatchat start -a and use the Streamlit WebUI or FastAPI endpoints

Alternatives

No reviewed alternatives recorded yet.