FastMCP Learning Path Vector Search Server

This project provides a Model Context Protocol (MCP) server that enables semantic search over a set of learning resources using FAISS and SentenceTransformers embeddings. It is compatible with Amazon Q and exposes a tool for finding relevant learning resources by query.

Features

• learning_path_search: Finds learning resources relevant to a text query using semantic similarity.
• Fast, local search using FAISS and all-MiniLM-L6-v2 embeddings.
• Deduplicates results by URL.
• Returns titles, snippets, and URLs for matched resources.

Requirements

• Python 3.10+
• Amazon Q CLI
• fastmcp Python package
• faiss-cpu Python package
• sentence-transformers Python package
• pyyaml and numpy Python packages

Installation

1. Clone this repository:

   git clone https://github.com/JoeStech/fastmcp-learning-path-search.git
   cd fastmcp-learning-path-search
   

2. Create and activate a virtual environment:

   python3 -m venv venv
   source venv/bin/activate
   

3. Install the required packages:

   conda is required here for faiss to install properly, which is why there's a two-step package install.

   pip install fastmcp sentence-transformers

   and then run

   conda install --file requirements.txt

4. Configure Amazon Q to use this MCP server by adding the following to ~/.aws/amazonq/mcp.json:

   {
       "mcpServers": {
           "learning_path_server": {
           "command": "/path/to/your/venv/bin/python3",
           "args": ["/path/to/your/server.py"],
           "env": {},
           "timeout": 60000
           }
       }
   }
   

Usage

1. Start Amazon Q CLI:

   q chat
   

2. View available tools by typing:

   /tools
   

This will display all available tools, including your custom MCP tools.

3. Use the tool to get back the best learning path based on your query:

   ask the learning_path_search tool to search for "are there GitHub hosted Arm runners for windows?" and please provide the best reference URL that it returns