Skip to content

Crader

Crader turns Git repositories into a code property graph stored in PostgreSQL. It parses code into semantic chunks using Tree-sitter, stores structural relations, and supports keyword and vector search with snapshot isolation.

What you can do

  • Index a repository into immutable snapshots
  • Generate embeddings with a staged, deduplicated pipeline
  • Run hybrid search (vector plus keyword) with Reciprocal Rank Fusion
  • Expand results with parent context and outgoing definitions
  • Read files and navigate code blocks from the graph

Key notes

  • Semantic tagging via Tree-sitter queries is currently provided for Python, JavaScript, and TypeScript.
  • The indexer always uses PostgreSQL; SQLite is intended for local experiments and tests. You can implement additional storage providers by extending the GraphStorage interface.

Documentation