semantic-search

Semantic code search lets you query a codebase by describing concepts and behavior. It uses an indexed embedding store to match descriptions like "retry logic with exponential backoff" to relevant code, going beyond simple text matching.

Conceptual codebase exploration helps you understand unfamiliar architecture by querying code meaning. You can ask questions like "how does the shard allocation system work" to discover relevant modules and trace data flow without reading every file.

Description-based code search finds implementations by behavior rather than function name. You provide a behavioral description, and the indexed embedding store matches it to the code responsible for that functionality, simplifying architecture discovery.

Yes, conceptual code search requires an absolute path to the codebase for indexing and searching. This path allows the embedding store to process and map the code structure so you can perform description-based queries.

Semantic code search supports tracing data flow by finding conceptually related code across modules. By querying the indexed embedding store for specific data behaviors, you can discover how data moves through the architecture.

Semantic search relies on an indexed embedding store, meaning it requires an initial indexing phase with an absolute codebase path. It focuses on conceptual and architectural discovery rather than exact text or regex matching.