TeaRAGs

Provides semantic code search across large codebases using vector embeddings and AST-aware chunking with git metadata integration for intelligent code discovery and analysis.

A high-performance semantic code search server built on Qdrant vector database with support for multiple embedding providers including Ollama, OpenAI, Cohere, and Voyage AI. Features AST-aware chunking using Tree-sitter for precise semantic boundaries, hybrid search combining dense vectors with BM25, and git metadata integration for filtering by authorship, code age, and churn rate. Designed for enterprise-scale codebases with incremental indexing that only reprocesses changed files.