mcp-server

# mcp-server (Knowledge Base + Components) This repository provides a set of Python components (including knowledge base data and vector database) for **Knowledge Base Retrieval / Matching / Reasoning**, for upstream projects to call through the **MCP protocol**. > This repository itself is not responsible for starting the complete MCP Server (the entry point is in another upstream project). > You can think of it as: the "capability implementation layer / tool collection" of the upstream MCP Server. --- ## What’s inside - Case/knowledge item matching (e.g. BM25/similarity matching, etc.) - Text normalization, disease/entity matching - Embedding/semantic retrieval (based on `sentence-transformers`) - PubMed search tool (if used) - `knowledge_base/`: JSON documents + vector database (small size, released with the repository) --- ## Requirements - Python `>= 3.12` (see `pyproject.toml`) - It is recommended to use `uv` to manage dependencies and environment (fast installation, reproducible) --- ## Install ### Option A: uv (Recommended) ```bash uv sync ``` ### Option B: pip (General) python -m venv .venv # Windows .\.venv\Scripts\activate # macOS/Linux source .venv/bin/activate pip install -e .