Kotaemon
title: kotaemon type: tool created: 2026-05-26 last_updated: 2026-05-26 related: ["radar/platforms/R2R", "radar/tools/Ollama", "radar/techniques/RagChunkingStrategies", "radar/techniques/SearchBasedRAG"] sources: ["https://github.com/Cinnamon/kotaemon"] radar_quadrant: Tools radar_ring: Assess radar_position: inner
Open-source RAG chat-with-documents web application with polished citation display, multi-modal support, hybrid retrieval, and multi-user access control. Self-hostable via Docker Compose.
kotaemon handles the full document-to-answer pipeline behind a usable web UI. Users upload PDFs, Word documents, or web URLs; the system ingests, indexes, and enables conversational Q&A with source citations highlighted inline in a document viewer.
Core capabilities:
- Multi-model backend — OpenAI, Claude, Gemini, Azure OpenAI, and local models via Ollama; model selection per conversation
- Hybrid retrieval — combines dense vector search and BM25 full-text search; addresses the named-entity retrieval gap of pure vector approaches
- Multi-modal — extracts and reasons over images, tables, and charts in PDFs, not just text
- Citation display — every answer cites source passages with jump-to-highlight in the original document viewer
- Multi-user — user accounts, document sharing, collection access control for team use
- Self-hostable — Docker Compose deployment; no mandatory cloud dependency; local-only operation possible with Ollama
kotaemon and R2R occupy different layers of the RAG stack:
| kotaemon | R2R | |
|---|---|---|
| Target user | End users: ask questions about documents | Developers: build RAG applications |
| Interface | Web chat UI with document viewer | REST API |
| Customisation | Configuration, not code | Full SDK extensibility |
| Use case | Team knowledge base, document Q&A | Production RAG backend |
kotaemon is the ready-made UI layer; R2R is the programmable platform to build custom UI on top of. For teams who want to query documents without building, kotaemon is the faster path.
kotaemon sits in the Assess ring of the Tools quadrant, at inner position. First studied via GitHub (2024-08-31). 20,000+ stars; actively maintained by Cinnamon. The citation display with inline document highlighting is the most differentiating UX feature in the chat-over-docs space — it makes the RAG pipeline trustworthy for non-technical users. Hybrid retrieval (dense + BM25) and multi-modal support align with best practices already on the radar. Docker Compose deployment path is straightforward. Remaining gate before Trial is a running instance serving a real document corpus to at least one non-developer user with citation quality verified against known answers.
