FAQ

What is Crader?

Crader indexes Git repositories into a code property graph stored in PostgreSQL. It parses code into chunks using Tree-sitter, stores structural relationships, and supports keyword and vector search.

Does Crader do incremental indexing?

Yes. Crader supports file-incremental indexing. When re-indexing a repository, only changed files are processed. Embeddings are deduplicated across snapshots by vector_hash.

Which languages are supported?

Indexing scans files by extension:

• .py
• .js, .jsx
• .ts, .tsx
• .java
• .go
• .rs
• .c, .cpp
• .php
• .html, .css

Semantic tagging via Tree-sitter queries is currently provided for Python, JavaScript, and TypeScript.

Why PostgreSQL instead of a separate vector database?

Crader stores the graph, full-text index, and vectors in one system. PostgreSQL provides transactional consistency and flexible joins between nodes, edges, and embeddings.

index() returns "queued". What does it mean?

The repository is already locked by another indexing run (a snapshot is still marked indexing). Wait for it to finish or mark the stale snapshot as failed or completed in the database before retrying.

Search returns empty or low-quality results

• Confirm that a snapshot is active for the repository.
• Run the embedding pipeline if you are using strategy="vector" or strategy="hybrid".
• Remove restrictive filters like path_prefix or language.
• Use strategy="keyword" to validate FTS results independently of embeddings.

Can I use local embeddings?

Yes. Use FastEmbedProvider for local embeddings or implement your own EmbeddingProvider.