AI SQL Agent - Ask your database anything with natural language

Architected a production-grade Natural Language to SQL agent that allows non-technical stakeholders to retrieve data from PostgreSQL/SQLite databases using plain English. Unlike standard chatbots, this system treats SQL generation as a compilation target, enforcing strict schema validation using Pydantic to ensure that the AI generates executable, syntactically correct code every time.

The agent implements a 'Self-Correction' execution loop (Reflection). If a generated query fails or returns an empty set, the error is captured and fed back into the context window, allowing the agent to replan and fix its own syntax automatically. This robust error handling mimics a human developer's workflow, ensuring high reliability for complex joins and aggregations without requiring user intervention.

Security and performance are handled through scoped read-only permissions and dynamic context injection. The system retrieves only relevant table schemas based on the user's intent to minimize token usage and latency. It outputs strictly structured JSON responses containing both the raw SQL and the interpreted data, making it easy to integrate into frontend dashboards or reporting tools.