semantic-search

Semantic search uses vector embeddings to find code by meaning rather than text matching. It analyzes Python functions, classes, and code elements through AST extraction, generates embeddings via OpenAI, and retrieves matches by semantic similarity in pgvector-backed PostgreSQL, enabling discovery of functionally related code you'd miss with keyword search.

Yes. Beyond code search, semantic search applies to production messages and facts. Index your data into PostgreSQL with pgvector embeddings, then query semantically to retrieve records by meaning—useful for log analysis, event correlation, and fact discovery without exact text matching.

Install Docker and Docker Compose, configure your OPENAI_API_KEY in superpowers/.env, then run docker-compose up -d to start PostgreSQL with pgvector and the indexing pipeline. The Dockerized environment handles embeddings generation and vector storage so you can immediately index and search code or data.

Yes. Semantic search is built for Python code search via the code-search CLI. It extracts code elements using AST analysis, generates embeddings, and enables you to find functions, classes, and related code blocks by semantic meaning within Python projects.

Keyword search matches exact text or simple patterns; semantic search understands code intent and meaning. By converting code to vector embeddings, semantic search finds functionally similar code even when names and syntax differ, making it stronger for discovery and refactoring tasks.

Yes. The Docker-based pipeline supports full codebase indexing. Run the indexing process to extract all Python code elements via AST, generate embeddings for each, and store them in pgvector. Once indexed, use code-search find to query the entire codebase semantically.