swapnanil/vectr

What it does

Vectr is a semantic search and persistent working memory system designed for AI code editors and development environments. It enables developers to quickly find relevant code sections using natural language queries through hybrid dense vector search combined with BM25 keyword matching. The tool maintains an AST-aware symbol graph for precise code location and tracing, and provides typed notes functionality that survives context window limitations and session restarts.

How to set up

Installation is straightforward via pip: pip install vectr. The tool requires no API keys or external service configuration, operating entirely locally on your machine. Simply install the package and configure it within your MCP-compatible AI editor or development environment.

Key features

• Hybrid dense vector + BM25 semantic search over codebases
• AST-aware code chunking for accurate context understanding
• Symbol graph for code navigation and dependency tracing
• Typed notes with persistence across context compaction and restarts
• Local embeddings with zero configuration
• No API keys or external dependencies required
• Cross-platform support (Windows, macOS, Linux)