xinPlugin_Chroma_fastMCP

xinPlugin_Chroma_fastMCP

Enables agents to retrieve and cite text from MinIO-hosted knowledge bases by indexing PDF/txt/md documents into Chroma and exposing search, ingestion, and source listing as MCP tools.

Category
Visit Server

README

xinPlugin_Chroma_fastMCP

MinIO 知识库的向量检索层：把上传到 MinIO 的文档（PDF/txt/md）抽取文本 → 分块 → 向量入库（Chroma），并通过 FastMCP 把「检索」暴露成 MCP 工具，供 DSH（DeepSeek Harness）的 dsh-mcp-client 连接、让 agent 在问答时直接检索原文。

组成

文件 作用
chroma_store.py Chroma 持久化 + 分块（带页码/行号）+ 混合检索（语义 + 字符二元组 BM25，RRF 融合）
ingest.py CLI 入库：python ingest.py <文件> [source名]，输出 JSON 摘要（供 MinIO 插件联动调用）
server.py FastMCP stdio 服务，暴露 search / ingest_file / list_sources
requirements.txt chromadb / fastmcp / pypdf

安装

# 一键安装 + 启动（依赖 + 校验 + 后台拉起 HTTP 服务，默认 127.0.0.1:8000）
powershell -ExecutionPolicy Bypass -File .\install.ps1
# 参数：-Port 8000  -BindHost 127.0.0.1  -NoLaunch(只安装)  -SkipInstall(跳过安装直接启动)

或手动：

pip install -r requirements.txt
# 首次检索会下载默认 embedding（all-MiniLM-L6-v2，约 80MB，缓存在 ~/.cache/chroma）

使用

# 入库
python ingest.py "广州十五五规划.pdf" "广州十五五规划.pdf"

# 检索（或经 MCP 工具 search）
python -c "from chroma_store import search; import json; print(json.dumps(search('广州 人工智能+ 大模型 算力 数据要素', 6), ensure_ascii=False))"

MCP 工具

  • search(query, top_k=6)：语义+关键词混合检索，返回原文片段及出处（文件 + 页码 + 行号）。
  • ingest_file(path, source_name)：本地文件入库。
  • list_sources()：已入库来源清单。

DSH 端以 stdio 连接 server.py（dsh-mcp-client），工具名形如 mcp__chroma__search。

检索原理

  • 分块：按页提取文本，过滤页眉/页脚/页码噪声，每 6 行一块（重叠 1 行），元数据记录 source/page/line_start/line_end。
  • 混合检索：Chroma 语义向量（余弦） + 字符二元组 BM25 稀疏检索，RRF 融合——中文语义 embedding 偏弱时，BM25 兜住「大模型/算力/数据要素」等精确关键词，保证出处定位稳定。
  • 入库即失效稀疏缓存，重复入库覆盖更新。

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured