Modular RAG MCP Server
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Modular RAG MCP Serversearch my ingested docs for how hybrid retrieval and RRF fusion work"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Modular RAG MCP Server
这是一个模块化 RAG 服务,通过 MCP(Model Context Protocol)向 AI 工具暴露知识检索能力。项目把摄取、检索、响应组装、Trace 和评估拆成独立模块,便于替换组件、定位问题和单独测试。
核心能力
PDF / Markdown 文档加载与语义分块
LLM 辅助 Chunk 精炼和元数据增强
文档图片 Vision Caption
Dense、Sparse 和 Hybrid Retrieval
RRF 融合与可配置 Reranker
MCP 知识检索工具
摄取与查询链路 Trace
Ragas + 自定义评估和固定回归集
Streamlit Dashboard,用于查看文档、Trace 和评估结果
Related MCP server: Modular RAG MCP Server
架构
MCP Client / Streamlit
│
▼
Query Engine
├── Query Rewrite
├── Dense Retrieval
├── Sparse Retrieval
├── RRF Fusion
├── Reranker
└── Response Assembly
Ingestion Pipeline
├── Document Loader
├── Semantic Splitter
├── Chunk Refinement
├── Metadata Enrichment
├── Image Captioning
└── Vector / Keyword Index Writers
Storage
├── Chroma Vector Store
├── Sparse Index
├── SQLite Ingestion History
└── Trace Store仓库结构
路径 | 职责 |
| 配置、检索、融合、重排、响应和 Trace |
| 加载、分块、增强、多模态处理和索引写入 |
| MCP Server、Tools、Resources 和 Prompts |
| Ragas 和自定义评估 |
| Streamlit 调试界面 |
| 配置、Prompt 和凭据模板 |
| ingest、query、evaluate、dashboard 入口 |
| Unit、Integration 和 E2E 测试 |
| 功能规格和验收标准 |
| 架构、AI 工程和测试说明 |
环境准备
python -m venv .venv
# Windows: .venv\Scripts\activate
pip install -e ".[dev]"
cp .env.example .env在 config/settings.yaml 中选择 Provider。真实 API Key 只放在环境变量或本地凭据文件中,不要提交到仓库。
运行
# 摄取文档
python scripts/ingest.py path/to/document.pdf
# 执行查询
python scripts/query.py "Hybrid Retrieval 是怎么工作的?"
# 运行评估
python scripts/evaluate.py
# 启动 Dashboard
python scripts/start_dashboard.py
# 启动 MCP Server
python main.py测试
pytest -m unit
pytest -m "not llm"
ruff check .标记为 llm 的测试需要真实模型服务,默认不参与本地快速验证。
评估
评估覆盖检索质量、答案忠实度、上下文精度、延迟和 Trace 完整性。Golden Case 和阈值作为版本化回归输入维护,不使用一次性 Notebook 结果代替工程评估。
当前边界
Provider 的可用性和限流会影响评估结果。
检索和重排质量依赖配置的 Embedding 与 Reranker 模型。
Dashboard 是调试工具,不是生产级可观测平台。
本地 Chroma 和 SQLite 适合开发及作品集规模语料;更大规模需要独立存储、并发控制和运维体系。
Mypy 已配置,但旧代码仍存在类型错误,尚未设为 CI 阻断项。
This server cannot be deployed
Maintenance
Related MCP Connectors
Cloud or self-hosted knowledge for AI agents: hybrid search, reranking, GraphRAG, scoped MCP tools.
Ingest, manage, and retrieve documents for RAG-powered AI applications
Make your knowledge agent-ready. One MCP endpoint, 5 connectors, 3 search modes.
The CustomGPT.ai MCP server is a fully managed, RAG-powered endpoint that connects large language models with private knowledge bases and external data sources. It provides tools for retrieval-augmented generation queries (send_message), data ingestion (upload_file), and source listing, enabling AI agents to query private documents like PDFs with high accuracy and real-time citations.
Related MCP Servers
- FlicenseNot gradedqualityCmaintenanceProvides RAG-based knowledge retrieval and document management as MCP tools, supporting hybrid search, reranking, and retrieval process visualization.-
- AlicenseNot gradedqualityCmaintenanceA modular RAG framework exposing knowledge retrieval tools via MCP, enabling AI assistants to perform hybrid search, reranking, and multimodal document queries with full observability and evaluation.MIT
- AlicenseNot gradedqualityCmaintenanceA modular RAG framework exposing MCP tools for hybrid document search, enabling AI assistants to query private knowledge bases with multimodal support and full observability.MIT
- AlicenseNot gradedqualityCmaintenanceEnables AI assistants to perform retrieval-augmented generation with hybrid search, reranking, multi-modal image processing, and RAG evaluation through standardized MCP tools.MIT