When Your SME Ghosts You: Using GenAI to Build a Database Documentation Agent
summary
Builds a working GenAI-powered Database Documentation Agent that auto-summarizes schemas, answers schema Q&A, generates SQL from business questions, and draws ER diagrams using Python, SQLite, and Gemini 2.5 flashlight. The prototype (called SME Ghost Protocol) parses table metadata into JSON, produces human‑readable table summaries, returns runnable SQL with plain‑English explanations, and visualizes primary/foreign key links — shipped as a Kaggle notebook and demoed live.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:46Undocumented databases problem
- 3:37Automating docs with GenAI
- 5:02Technical stack overview
- 6:48Prompt engineering & MVP
- 7:33SME Ghost Protocol
- 9:18Lessons learned
- 10:42Future roadmap
- 12:07Live demo
- 15:37Setting up the database and schema extraction
- 17:29Table summaries with AI
- 18:52Natural language Q&A
- 20:20Schema exploration demo
- 22:08SQL for business questions
- 23:57ER diagram visualization
- 25:02Q&A
related talks
19mHow to use generative AI to deliver self-service data products
50mLeveraging AI as a PostgreSQL DBA
15mGraphRAG-Powered AI Agents: Transforming Corporate Metadata into Intelligent Knowledge
45mForget SQL! Create your own AI Data Analyst
34mBuilding Multi-Agent Generative-AI Applications with AutoGen
22mLearning to Apply Generative AI at Enterprise Level
10mImprove Developer Productivity by Leveraging GenAI
29mAI-Driven Code Generation and Testing: Accelerating Development with GenAI