Skip to content

SQL Tools

A suite of composable tools for text‑to‑SQL workflows and database insights. Use them end‑to‑end or pick individual blocks that fit your agent.

  • Relevant Table Schema Finder (LLM)

    Selects the most relevant tables based on summaries and returns a schema context to guide query generation.

    Documentation

  • SQL Query Generator

    Produces a single, syntactically valid T‑SQL query from a refined question and provided schema context.

    Documentation

  • SQL Executor

    Executes the generated SQL with SQLAlchemy and returns rows as dictionaries for downstream formatting.

    Documentation

  • Final Answer Formatter

    Converts raw rows into a clear, user‑friendly narrative and/or markdown table suitable for UI display.

    Documentation

  • DB Insights Tool

    Answers conceptual questions about schema design and relationships using table metadata (no query execution).

    Documentation

Flow (typical)

  1. Relevant Table Schema Finder (LLM) → build focused schema context
  2. SQL Query Generator → produce a valid query
  3. SQL Executor → run the query and collect rows
  4. Final Answer Formatter → present the result to the user
    (Use DB Insights Tool anytime you need architectural/relationship explanations.)

All SQL tools are standard BaseTool subclasses. Register them in a Toolkit, validate inputs with ToolSpec.check_inputs(), and chain state via ToolResponse.extras (e.g., schema_context, db_results).