Agentic AI · Gen AI

Agentic Multi-Dialect SQL Assistant

A natural-language-to-SQL agent that reasons over your schema and generates correct, executable queries across multiple SQL dialects.

Live Demo All Projects

Technical Information

Project Type
NLP to SQLAgentic AssistantMulti-Dialect
Tech Stack
PythonFlaskMistral AIPhiDataPrompt EngineeringRDBMSSQLiteMySQLPostgreSQLOracleSQL ServerAmazon Aurora
Topics / Modules
Schema ReasoningQuery GenerationOrder of ExecutionAgentic Workflow
Links
Live Demo

Project Summary

An agentic assistant that converts plain-English questions into accurate SQL across SQLite, MySQL, PostgreSQL, Oracle, SQL Server and Amazon Aurora. Upload a database or schema, ask a question, and the agent reasons about tables and relationships before generating and executing the query.

Skills Demonstrated

NL-to-SQLAgentic AIPrompt EngineeringSchema UnderstandingQuery ValidationMulti-Dialect SQL

Solution

A Mistral-powered agent built on PhiData inspects the uploaded schema and sample data, picks the right dialect, then drafts, validates and runs the SQL – returning both the result set and the query's order of execution for learning.

Approach

  1. Ingest: Accept an uploaded database/schema and surface tables, columns and sample rows.
  2. Understand: Let the agent reason about relationships and the target SQL dialect.
  3. Generate: Draft dialect-correct SQL from the natural-language question.
  4. Execute: Run the query safely and return results with the order of execution.
  5. Explain: Show the generated SQL and step-by-step execution for transparency.
  6. Deploy: Ship as a Flask app on DigitalOcean.