AI导哈

RAG知识检索平台

Humbowo是一个生产级的知识检索与RAG平台，提供多租户知识库、语义搜索、混合检索和MCP集成，适用于学术、商业和运营等文档密集型场景。

  • by Kanyuchi
  • Python
  • 本地部署
  • 1 星标
  • 更新于
在 GitHub 查看 Kanyuchi/literature-rag-mcp

使用说明

Humbowo

Repo note: the repository name (retrievo) is legacy; the product is Humbowo at humbowo.com (by TheNerdsInt).

Humbowo is a production‑grade retrieval and knowledge platform for document‑heavy workflows — academic, business, and operational. It pairs a scalable RAG backend with a modern web app, supports multi‑tenant knowledge bases, and is designed to grow from thousands to millions of documents without changing the core architecture.

Want to deploy this for your own domain? See the Replication Guide for step‑by‑step instructions.

Features

  • Multi‑tenant knowledge bases with job‑scoped search, chat, and indexing
  • Semantic search with OpenAI embeddings (configurable model)
  • Hybrid retrieval (BM25 + dense) and optional reranking
  • ChromaDB vector storage for fast, scalable retrieval
  • FastAPI REST API (port 8001) with auth + admin‑friendly endpoints
  • Web app for upload, search, chat, and knowledge base management
  • MCP integration available for Claude/agent tooling (optional)

MCP Server Setup (Optional)

Claude Desktop

Config: ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "literature-rag": {
      "command": "/Users/fadzie/Desktop/lit_rag/literature_review_rag_api/venv/bin/python",
      "args": ["-m", "literature_rag.mcp_server"],
      "cwd": "/Users/fadzie/Desktop/lit_rag/literature_review_rag_api",
      "env": {
        "PYTHONPATH": "/Users/fadzie/Desktop/lit_rag/literature_review_rag_api",
        "INDICES_PATH": "/Users/fadzie/Desktop/lit_rag/literature_review_rag_api/indices",
        "CONFIG_PATH": "/Users/fadzie/Desktop/lit_rag/literature_review_rag_api/config/literature_config.yaml"
      }
    }
  }
}

Claude Code

Config: ~/.claude.json (add to mcpServers object)

"literature-rag": {
  "type": "stdio",
  "command": "/Users/fadzie/Desktop/lit_rag/literature_review_rag_api/venv/bin/python",
  "args": ["-m", "literature_rag.mcp_server"],
  "env": {
    "PYTHONPATH": "/Users/fadzie/Desktop/lit_rag/literature_review_rag_api",
    "INDICES_PATH": "/Users/fadzie/Desktop/lit_rag/literature_review_rag_api/indices",
    "CONFIG_PATH": "/Users/fadzie/Desktop/lit_rag/literature_review_rag_api/config/literature_config.yaml"
  }
}

Key Configuration Notes

  1. Absolute paths required - Relative paths fail in sandboxed environments
  2. PYTHONPATH - Required for module discovery
  3. INDICES_PATH & CONFIG_PATH - Override defaults to ensure correct file access
  4. Restart required - After config changes, restart Claude Desktop/Code

Available MCP Tools

Tool Description
semantic_search Search papers with phase/topic/year filters
get_context_for_llm Pre-formatted context with citations
list_papers Discover available papers in collection
find_related_papers Find similar papers via embeddings
get_collection_stats Collection statistics and metadata
answer_with_citations Structured sources with bibliography
synthesis_query Multi-topic comparative analysis

Quick Start

# Install dependencies
cd literature_review_rag_api
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
pip install fastmcp

# Run MCP server
python -m literature_rag.mcp_server

# Or run FastAPI server
python -m literature_rag.api

Project Structure

retrievo/
├── literature_review_rag_api/
│   ├── literature_rag/
│   │   ├── mcp_server.py      # MCP server (FastMCP)
│   │   ├── api.py             # FastAPI server
│   │   ├── literature_rag.py  # Core RAG engine
│   │   └── config.py          # Configuration
│   ├── config/
│   │   └── literature_config.yaml
│   ├── indices/               # ChromaDB storage
│   └── venv/
├── webapp/                     # Humbowo web UI (search/chat/upload)
└── README.md

相关 MCP