vector-index-mcp

# MCP Server: Software Project Indexing & Semantic Search This project provides a Model Context Protocol (MCP) server designed to index software project files and offer semantic search capabilities over the indexed content. It monitors project directories for changes and maintains an up-to-date index. For detailed design and architectural decisions, please refer to [ARCHITECTURE.md](ARCHITECTURE.md). ## Usage (End Users) This server is designed to be run easily by pointing it to the software project you want to index. ### 1. Run the Server Navigate to a convenient location in your terminal. The server is started using the `python -m` command, specifying the module and the path to the project you wish to index. ```bash python -m vector_index_mcp.main_mcp <path_to_your_project> ``` For example, to index a project located at `~/dev/my_cool_project`, you would run: ```bash python -m vector_index_mcp.main_mcp ~/dev/my_cool_project ``` Or, to index the current directory: ```bash python -m vector_index_mcp.main_mcp . ``` The server will start, begin watching the specified project path for changes, and create/use a LanceDB database within that project's directory (by default). ### 2. Configuration * **`PROJECT_PATH` (Required, via Command Line):** * This is now passed as a direct command-line argument to the server, as shown above. It specifies the root directory of the software project you want to index. * **Environment Variables (Optional, via `.env` file):** * Other settings can be configured using an `.env` file placed in the directory **from where you run the `python -m ...` command**. The server uses `pydantic-settings` and will automatically load this `.env` file. * The following variables are supported: * `LANCEDB_URI`: Path where the LanceDB vector database will be stored. * Default: `./.lancedb` (relative to the indexed project's path, meaning it's stored within the project itself). * `EMBEDDING_…