Overview: RAGFlow is a retrieval-augmented generation engine that combines document ingestion, parsing, retrieval, context management, and agent workflows.
What beginners can do: Users can create datasets, select chunking and embedding strategies, connect models, test retrieval, build chat assistants, and extend them with agent workflows.
Recommended start: Read the current prerequisites before cloning. The repository currently calls for Docker 24+, Docker Compose 2.26.1+, and Python 3.13 for development. Start the stable Docker image, wait for all services to become healthy, then import one small document set.
Requirements and cost: The default stack includes several infrastructure services and x86 prebuilt images; ARM64 users need a separate build path. Storage, memory, model services, and embedding costs must be planned.
Who it suits: Useful for technical teams that have outgrown a desktop document-chat tool and need deeper control over parsing and retrieval.
Limits and safety: This is not the simplest first RAG project. Changing the document engine or running destructive Docker volume commands can clear data, so upgrades and migrations need tested backups. Change default credentials, protect model and database secrets, restrict code-executor features, use gVisor when required, isolate outbound access, and never expose the default stack directly to the internet.
Project and license: Repository content is provided under Apache-2.0; check the current LICENSE before reuse. Repository: https://github.com/infiniflow/ragflow