context-engine

Cross-repo code search locates function implementations across multiple repositories by running hybrid semantic and lexical queries with neural reranking. It exposes interfaces like repo_search to handle cross-repo queries and trace usage across projects accurately.

Semantic code search captures underlying intent rather than exact string matches, differing from lexical search by using neural reranking. Combining both techniques in a hybrid approach yields faster, more accurate results for locating relevant implementations and workflows.

Yes, you can use a symbol graph to discover code relationships and trace callers in a large codebase. The feature provides symbol graph lookups to map relationships, helping developers understand workflows and find relevant usage patterns.

Yes, the code analysis tool supports storing insights for future reuse through memory-backed knowledge retrieval. It includes specific memory tools designed for knowledge capture, allowing developers to save and retrieve context during history exploration.

The best way to understand workflows and locate implementations in a single repo is using hybrid semantic and lexical queries with neural reranking. Single-repo queries accurately trace callers and map symbol relationships within the codebase.