kotaemon

kotaemon is an open source document chat app for people who want to ask questions across their files and inspect the evidence behind an answer. Developers can also use it to build custom document question-answering applications.

• Answers across documents: Ask questions spanning multiple files, including material in figures and tables; search combines full-text and vector retrieval with re-ranking.
• Inspectable citations: Check highlighted passages and relevance scores in the browser's PDF viewer, with warnings when retrieved material has low relevance.
• Shared collections: Multi-user login, private and public file collections, and chat sharing support collaborative use.
• Complex questions: Question decomposition and ReAct or ReWOO agents support questions that require several reasoning steps.
• Custom pipelines: Developers can extend the Gradio interface and replace indexing or reasoning pipelines, including using GraphRAG indexing.

The Apache 2.0 app runs on Windows, macOS and Linux through a browser interface, with Docker supported. Local LLM backends include Ollama and llama-cpp-python; Docling and PaddleOCR provide local document parsing. This combination supports private processing on your hardware.

Optional cloud models include OpenAI, Azure OpenAI, Cohere and Groq. Azure Document Intelligence and Adobe PDF Extract provide cloud parsing; those choices send the corresponding processing to external services. A Hugging Face Space offers a separate online deployment.