# RAGFlow

> RAGFlow is an open-source Retrieval-Augmented Generation (RAG) engine from InfiniFlow Inc. It pairs
> deep document understanding (DeepDoc parsing of PDFs, images, tables and scans) with hybrid retrieval
> — dense vector, BM25 full-text and tensor/multi-vector re-ranking — and an integrated visual agent
> platform. It runs self-hosted under Apache-2.0 (Docker) or as RAGFlow Cloud at cloud.ragflow.io.

Generated by API Evangelist on 2026-08-27 from the provider's own public documentation and repositories.
RAGFlow does not publish an llms.txt of its own (https://ragflow.io/llms.txt returns 404), and it
publishes no OpenAPI, AsyncAPI or GraphQL SDL. The HTTP API below is documented as a human reference
only; every path here is copied from that reference, not inferred.

## API basics

- Base URL (RAGFlow Cloud): https://cloud.ragflow.io/api/v1
- Base URL (self-hosted): http://{address}/api/v1  — {address} is your own deployment, default port 9380
- Authentication: `Authorization: Bearer <YOUR_API_KEY>` on every request. Keys are issued from the
  RAGFlow UI (avatar -> API). On RAGFlow Cloud, API keys are NOT available on the Free plan.
- Content type: application/json (multipart/form-data for document upload)
- Response envelope: `{"code": <int>, "data": <object|null>, "message": <string>}` — `code` 0 means
  success; HTTP 401 returns `{"code":401,"data":null,"message":"<Unauthorized '401: Unauthorized'>"}`
- Pagination: `page` (default 1) and `page_size` on every list endpoint, with `orderby` and `desc`
- No `Idempotency-Key` header, no rate-limit response headers, and no documented request-rate limits.
  Usage is metered as monthly credits per RAGFlow Cloud plan.

## Documentation

- Docs home: https://ragflow.io/docs
- HTTP API reference: https://ragflow.io/docs/http_api_reference
- Python API reference: https://ragflow.io/docs/python_api_reference
- Acquire an API key: https://ragflow.io/docs/acquire_ragflow_api_key
- FAQ: https://ragflow.io/docs/faq
- Release notes: https://ragflow.io/docs/release_notes
- What is RAG: https://ragflow.io/basics/what-is-rag

## API surface (95 documented operations)

- OpenAI-compatible: POST /api/v1/openai/{chat_id}/chat/completions,
  POST /api/v1/agents_openai/{agent_id}/chat/completions
- Datasets: POST|DELETE /api/v1/datasets, PUT /api/v1/datasets/{dataset_id},
  GET /api/v1/datasets, plus knowledge-graph and RAPTOR construction/status endpoints
- Documents: POST|DELETE /api/v1/datasets/{dataset_id}/documents,
  PATCH|GET /api/v1/datasets/{dataset_id}/documents/{document_id},
  POST|DELETE /api/v1/datasets/{dataset_id}/chunks (start/stop parsing),
  POST /api/v1/documents/ingest
- Chunks: POST|GET|DELETE /api/v1/datasets/{dataset_id}/documents/{document_id}/chunks,
  GET|PATCH .../chunks/{chunk_id}, GET /api/v1/datasets/{dataset_id}/metadata/summary
- Retrieval: POST /api/v1/retrieval
- Chat assistants: POST|DELETE /api/v1/chats, GET|PUT|PATCH|DELETE /api/v1/chats/{chat_id}
- Sessions: POST|GET|DELETE /api/v1/chats/{chat_id}/sessions, POST /api/v1/chat/completions,
  POST /api/v1/agents/chat/completions, POST /api/v1/chat/audio/speech,
  POST /api/v1/chat/audio/transcription, POST /api/v1/chat/mindmap,
  POST /api/v1/chat/recommendation
- Agents: GET|POST /api/v1/agents, PUT|DELETE /api/v1/agents/{agent_id}
- Memory: POST|GET /api/v1/memories, PUT|DELETE /api/v1/memories/{memory_id},
  POST|GET /api/v1/messages, GET /api/v1/messages/search
- Files and workspaces: POST|GET|DELETE /api/v1/files, POST /api/v1/files/move,
  POST /api/v1/files/link-to-datasets, GET|POST /api/v1/workspaces/{workspace_id}/commits,
  GET /api/v1/workspaces/{workspace_id}/commits/diff
- Search apps: POST|GET /api/v1/searches, GET|PUT|DELETE /api/v1/searches/{search_id},
  POST /api/v1/searches/{search_id}/completions
- System: GET /api/v1/system/healthz (unauthenticated; returns db / doc_engine / redis / storage status)

## Agent surfaces

- MCP server: first-party, shipped in the repo at mcp/server/server.py, DISABLED BY DEFAULT and
  self-hosted. Transports: streamable-HTTP at /mcp and legacy SSE at /sse. Tools: ragflow_retrieval,
  ragflow_list_datasets, ragflow_list_chats. No InfiniFlow-hosted MCP endpoint exists.
  https://ragflow.io/docs/launch_mcp_server
- Agent Skill: https://github.com/infiniflow/ragflow-skills — `ragflow-dataset-ingest`
- A2A Agent Card: none published on any RAGFlow host.

## SDKs and distribution

- Python SDK: https://pypi.org/project/ragflow-sdk/ (0.27.0, 2026-08-19, Apache-2.0)
- Server image: https://hub.docker.com/r/infiniflow/ragflow
- No JavaScript/TypeScript, Java, Go, Ruby or Rust client is published.

## Source, releases and security

- Source: https://github.com/infiniflow/ragflow (Apache-2.0)
- Releases: https://github.com/infiniflow/ragflow/releases — latest v0.27.0, 2026-08-19
- Changelog: https://ragflow.io/changelog
- Security advisories + private vulnerability reporting:
  https://github.com/infiniflow/ragflow/security/advisories

## Commercial

- Pricing (RAGFlow Cloud): https://ragflow.io/#pricing-plan — Free $0 (no API key),
  Starter $29/month, Pro $129/month, Enterprise contact sales
- Contact: https://ragflow.io/contact-us
- Terms of Use: https://ragflow.io/policies/terms
- Privacy Policy: https://ragflow.io/policies/privacy
- Data Processing Agreement: https://ragflow.io/policies/dpa