Database Toolset for Pydantic AI

Empower AI agents with SQL database capabilities

SQL database toolset that lets AI agents query, analyze, and interact with databases. Supports schema introspection, safe query execution, and result formatting.

Installation

pip install database-pydantic-ai

Five tools give agents full database interaction: list_tables, get_schema, describe_table, explain_query, and query. Supports both SQLite (via aiosqlite) and PostgreSQL (via asyncpg) with a unified interface. Security-first: read-only mode blocks 15+ dangerous SQL keywords including DROP, DELETE, and ALTER. Resource control with configurable query timeouts (30s default) and row limits (100 default). Comment-aware parsing detects dangerous keywords hidden behind -- and /* */ comments.

Features

Quick Start

from pydantic_ai import Agent
from database_pydantic_ai import (
    SQLiteDatabase,
    SQLDatabaseDeps,
    SQLITE_SYSTEM_PROMPT,
    create_database_toolset,
)

async with SQLiteDatabase("data.db") as db:
    deps = SQLDatabaseDeps(database=db, read_only=True)
    agent = Agent(
        "openai:gpt-4o",
        deps_type=SQLDatabaseDeps,
        toolsets=[create_database_toolset()],
        system_prompt=SQLITE_SYSTEM_PROMPT,
    )
    result = await agent.run("What are the top 5 most expensive products?", deps=deps)

Use Cases