flexible-graphrag

概述

Flexible GraphRAG 是一个全面的 AI 上下文平台，与 MCP 集成，提供从 13 个数据源（9 个支持自动同步）摄取文档、执行跨向量、图和全文数据库的混合搜索，以及进行 AI 查询的工具。MCP 服务器将系统的核心功能暴露给 Claude Desktop 和其他 MCP 客户端，实现与 AI 工作流程的无缝集成。它通过使用 LlamaIndex 和 LangChain 的双框架架构支持多种图数据库、向量存储和 LLM 提供商。

试试问 AI

装完之后，这里有 5 个你可以让 AI 做的事：

• ingest_documents

  Ingest documents from various data sources into the knowledge graph

  试试:Ingest documents from the Alfresco repository试试:Upload documents from Amazon S3 bucket试试:Process documents from local filesystem directory
• search_documents

  Perform hybrid search across vector, full-text, and graph databases

  试试:Search for documents related to machine learning research试试:Find documents containing information about customer contracts试试:Search using semantic similarity and keyword matching
• query_documents

  Query documents using natural language or SPARQL for RDF stores

  试试:What are the latest financial reports?试试:Find information about product development timeline试试:Query the knowledge graph for relationships between entities
• ingest_text

  Directly ingest text content into the knowledge graph

  试试:Ingest this text: 'Company policy requires all employees to complete annual training'试试:Add this document content to the knowledge base试试:Process and index the following text snippet
• system_diagnostics

  Run system diagnostics and health checks

  试试:Check system health and performance试试:Run diagnostic tests on the knowledge graph试试:Verify all database connections are working
• health_checks

  Perform health checks on system components

  试试:Check if the vector database is responsive试试:Verify the graph database connection status试试:Confirm all services are running normally
• get_processing_status

  Get the status of ongoing document processing tasks

  试试:Check status of the document ingestion job试试:Show progress of knowledge graph extraction试试:List all active processing tasks
• cancel_processing_task

  Cancel an ongoing document processing task

  试试:Cancel the document ingestion job试试:Stop the knowledge graph extraction process试试:Abort the text processing task
• list_data_sources

  List available data sources for document ingestion

  试试:Show all available data sources试试:List configured repositories and storage systems试试:What data sources can I use for ingestion?

可对比工具

安装

{
  "mcpServers": {
    "flexible-graphrag": {
      "command": "python",
      "args": ["-m", "flexible_graphrag_mcp"]
    }
  }
}

FAQ

MCP 服务器支持哪些数据源？

MCP 服务器支持 13 种数据源，包括文件系统、云存储（S3、Azure）、企业知识库（Alfresco、SharePoint、Box、CMIS）和 Web 源。其中 9 种支持自动增量更新。

我可以同时将 LlamaIndex 和 LangChain 框架与 MCP 服务器一起使用吗？

是的，MCP 服务器同时支持 LlamaIndex（默认）和 LangChain 框架。每个处理阶段都可以通过环境变量独立配置使用任一框架。

flexible-graphrag 对比