Facilitates natural language search over relational database schemas to quickly locate tables and their relationships.
Schema Search is a powerful tool designed to simplify navigating complex relational database schemas. Instead of manually sifting through hundreds of tables or relying on LLMs that struggle with large contexts, it allows users to query database metadata using natural language. It builds schematic embeddings of tables, stores them in-memory without needing a separate vector database, and provides millisecond query latency. This tool integrates seamlessly as an MCP Server or can be used via its Python API, offering various search strategies (semantic, BM25, fuzzy, hybrid) and optional reranking for enhanced accuracy, making it efficient for developers and data professionals alike.