GitHubGitHub RepoNeeds Review

infiniflow/ragflow

RAGFlow packages document ingestion, retrieval, citations, agent workflows, memory and MCP into one self-hostable context layer. It is useful when an agent needs traceable answers across mixed private data.

GitHub

Open Source Open Markdown Open JSON

Source Summary

RAGFlow combines document parsing, chunk inspection, fused retrieval, reranking and traceable citations with agent templates, memory, MCP and code execution. The documented image is **v0.26.4**, with **Docker 24+** and **Compose 2.26.1+** required.

Practical Implication

Consider it when coding agents need one context service across files, structured sources and synced services such as Confluence, S3, Notion or Google Drive. Its visible chunking and references provide useful inspection points when retrieval quality matters.

Agent-Ready Context

RAGFlow combines document parsing, chunk inspection, fused retrieval, reranking and traceable citations with agent templates, memory, MCP and code execution. The documented image is **v0.26.4**, with **Docker 24+** and **Compose 2.26.1+** required.

Consider it when coding agents need one context service across files, structured sources and synced services such as Confluence, S3, Notion or Google Drive. Its visible chunking and references provide useful inspection points when retrieval quality matters.

Prebuilt images target **x86 only**; ARM64 users must build their own. Code execution requires gVisor, switching document engines can clear stored volumes, and the repository’s quality claims are not supported here by comparative evaluation.

Connected Context · Feed7 Judgment

This packages several context-engineering concerns into one inspectable service: ingestion, chunking, retrieval, reranking, citations, memory, and agent access. It broadens the prior graph- and code-specific approaches to heterogeneous enterprise sources, while its operational constraints and lack of comparative evaluation leave retrieval quality, architecture choice, and deployment cost to local testing.

AI on Your Lakehouse: Context Comes in Shapes, Not Queries — Zach Blumenfeld, Neo4jRAGFlow provides fused retrieval and visible chunks, while the candidate argues that graph relationships can expose structure those retrieval surfaces may miss; together they imply benchmarking retrieval shapes on the actual corpus.tirth8205/code-review-graphcode-review-graph is a narrow dependency-aware MCP surface for review; RAGFlow is a broader document and service context layer, so the former could remain preferable when caller, dependent, and test relationships are the needed evidence.Citation Needed: Provenance for LLM-Built Knowledge Graphs — Daniel Chalef, Zep AIRAGFlow’s traceable citations reinforce the need for provenance, but the candidate identifies harder cases—merged, changing, or deleted facts—that simple source references may not resolve.RLM: Recursive Language Models for Large Codebases - Shashi, Superagentic AIRLMs retrieve bounded evidence through agent-written inspection code; RAGFlow offers a prebuilt ingestion and retrieval service, presenting different implementation paths for avoiding whole-corpus context loading.

Context Map

contextcodingdata#retrieval#mcp #context-engineering

Uncertainty

Prebuilt images target **x86 only**; ARM64 users must build their own. Code execution requires gVisor, switching document engines can clear stored volumes, and the repository’s quality claims are not supported here by comparative evaluation.