GitHub user matebenyovszky added a comment to the discussion: Vambery AI Agent — AI-powered SQL assistant extension for SQL Lab (public beta)
Thank you for the feedback and for your questions! Your tool is great! I wrote my first text2SQL tool last spring which supported multiple dialects, and last autumn it was able to create complex query plans, evaluate and update plan step by step and create a final report based on that. The main context and sample queries were fed into the context, and the main logic and steps were predefined (e.g. I gave samples to the AI how to create a plan etc.). In this extension I wanted to test the built-on "thinking" or reasoning approach of modern "tool using" LLM-s like GPT5, and to be honest it works really well! So far it solved all questions, however I still plan to enrich context with addition descriptions from the scheme or even add skills (so the AI would know how to use properly complex databases). Regarding to the context window question it works pretty smart, first it checks tables overall, and try to find tables that could be relevant to answer the prompt, than check table columns etc. Here is an example: <img width="1962" height="1543" alt="image" src="https://github.com/user-attachments/assets/959fe1f9-1cfd-4e42-8ac5-a56295528cb5"; /> I am really thankful and open to your suggestions! 1) Do you know an good validation libraries worth adding? 2) Will double check, if database backend type is known by the AI 3) See sample above + now I have an initial charting tool also beyond auto-execution GitHub link: https://github.com/apache/superset/discussions/38356#discussioncomment-15987003 ---- This is an automatically sent email for [email protected]. To unsubscribe, please send an email to: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
