QueryGPT with LangGraphjs : Build a Natural Language to SQL Agent Workflow #langgraph #llm #sql #ai

In this video, I walk through how I built a QueryGPT-style system using LangGraphJS and GPT-4o, where users can ask natural language questions and get back live SQL results — not just queries.

We break it down step by step:
- Intent detection
- Table & column selection
- SQL generation
- Query execution with SQLite

I use a lightweight, real-world schema (Northwind) and share the open-source repo so you can run it yourself. Perfect for anyone exploring LLM agents, LangGraph, or building internal analytics copilots.

00:00 - Intro: What is QueryGPT?
00:21 - Natural Language to SQL Flow
00:50 - Intent Agent: Detecting Business Domain
01:20 - Tables Agent: Determining Relevant Tables
01:46 - Columns Agent: Optimizing Query Columns
02:23 - SQL Agent: Generating SQL from Context
02:50 - Execution Agent: Running the Final Query
03:21 - Full Circle: Returning Query Results
03:44 - LangGraphJS Architecture Overview
04:18 - SQLite + Northwind Dataset for Demo
04:58 - Tooling Breakdown (Workspaces, Tables, Query)
06:15 - Agents and Tools in LangGraphJS
07:24 - ReAct Agent Design with Tool Access
08:22 - SQL Agent and Final Execution
09:14 - Wrapping Up + GitHub Link

Author: Rabea Abdelwahab

[felixbel]

Hi, I want to ask you how can I integrate this with ollama or Deepsek, and to do it with the IMDB data base. Thanks!