Back to Gap Map Product / UX / Orchestration & agents

RAGFlow

open source / Overall score: 3.7

RAGFlow is a RAG engine built around deep document understanding, with template-based chunking and layout-aware parsing of complex PDFs, tables, figures and scans, producing answers with visible citations. Unlike a general orchestration framework it is a batteries-included RAG server deployed with Docker Compose, and it includes agent workflows and MCP support.

A managed cloud is sold beside the published engine, and there is no first-party package beyond the SDK client.

Openness

5 high confidence

5.0

license: Apache-2.0(OSI)
source: public
self-host: primary(Docker)
commercial: cloud/enterprise(separate)
core features withheld: no

Apache-2.0 self-hostable engine; cloud is the commercial tier.

https://github.com/infiniflow/ragflow recorded 2026-08-13
Apache-2.0 license, full engine source published for Docker Compose self-hosting, 87.9k stars and 10.3k forks; managed cloud sold beside the published engine
https://raw.githubusercontent.com/infiniflow/ragflow/main/LICENSE recorded 2026-09-29
The LICENSE file is the unmodified Apache License text, with no additional restriction or commercial-use clause: "Apache License" and "Version 2.0, January 2004".
https://raw.githubusercontent.com/infiniflow/ragflow/main/README.md recorded 2026-09-29
The README documents running the full service from a public source checkout: "Launch Service from Source for Development" and "Clone the source code and install Python dependencies".
https://raw.githubusercontent.com/infiniflow/ragflow/main/README.md recorded 2026-09-29
The README badge names the license, "License-Apache--2.0", and the hosted service is offered separately: "Try our cloud service at". Searched the body for paid, pricing, subscription, license key, enterprise edition and ee/: none; its edition table lists Docker image variants only.

Adoption

3 medium confidence

3.0

RAGFlow ships as a Docker Compose stack, so Docker Hub image pulls are the main distribution signal, and the ragflow-sdk package on PyPI is only an API client rather than the server. The pull count is a cumulative total since the image was published, so it understates the current pace of use.

https://hub.docker.com/v2/repositories/infiniflow/ragflow/ recorded 2026-09-29
Docker Hub lists a cumulative pull count for the image, "pull_count" ":3843434," with the description "Try our cloud service at" and "Retrieval-Augmented Generation".
https://pypistats.org/api/packages/ragflow-sdk/recent recorded 2026-09-29
{"data":{"last_day":750,"last_month":19603,"last_week":3553},"package":"ragflow-sdk","type":"recent_downloads"}

Capability

4 medium confidence

4.0

Opinionated, document-centric RAG application; deeper parsing than general frameworks, narrower scope.

https://github.com/infiniflow/ragflow recorded 2026-08-13
README key features - deep document understanding based knowledge extraction from unstructured data with complicated formats, template chunking, grounded citations, hybrid retrieval; no benchmark figure
https://raw.githubusercontent.com/infiniflow/ragflow/main/README.md recorded 2026-09-29
The README key features list document understanding, template chunking, grounded citations, fused re-ranking recall and agent capabilities, with no benchmark figure: "Template-based chunking" and "Grounded citations with reduced hallucinations".

Verified 2026-09-29