Project RAG

What it does\nConnects AI assistants to local codebases, allowing them to perform deep semantic search and architectural analysis using Retrieval-Augmented Generation (RAG). It indexes files locally using FastEmbed and LanceDB, enabling high-performance code understanding without sending data to external APIs.\n\n## Tools\n- index_codebase: Smartly indexes a directory with support for .gitignore and incremental updates.\n- query_codebase: Hybrid semantic and keyword search across the indexed code.\n- search_git_history: Performs semantic search over commit messages, diffs, and authors.\n- find_definition: Locates the definition of a symbol using a hybrid approach (Stack-graphs or RepoMap).\n- find_references: Identifies all usages of a symbol across the codebase.\n- get_call_graph: Analyzes function call relationships for impact analysis.\n- get_statistics: Provides counts of files, chunks, and language breakdowns.\n- clear_index: Wipes the vector database for a fresh start.\n- search_by_filters: Advanced search filtered by extension, language, or path.\n\n## Installation\nAdd to claude_desktop_config.json:\njson\n{\n \"mcpServers\": {\n \"project-rag\": {\n \"command\": \"/absolute/path/to/project-rag/target/release/project-rag\",\n \"env\": { \"RUST_LOG\": \"info\" }\n }\n }\n}\n\n\n## Supported hosts\n- claude