Build a DAX Code Generator

Author(s): Barrett Studdard

Originally published on Towards AI.

Utilize LLM AI to Create A Power BI Coding Assistant

I’ve been wanting to play around with recent technologies and approaches used in building LLM-powered applications — one interesting idea that came to mind was an assistant to generate DAX. Now, this could be done via ChatGPT currently (simply ask it to generate sum of sales for example), but for real assistance the model needs context awareness of how a data model is setup. Here is a sneak peek at the final result and then some details on how I built it.

With the problem/idea, the overall workflow that would be implemented:

A user could upload a Power BI model.The application parses the model and extracts relevant components (tables, columns, relationships) into plain text.Additional DAX awareness is added such as formatting suggestions, general rules, etc.All this is wrapped up and passed to the LLM along with the user’s question.An AI judge is used to score the confidence in the response.A testing approach to see how well the LLM performs.

In addition to the above, the goal was to build a custom and extensible app using FastAPI and Vue — common backend and frontend technologies that could be built upon in many different ways.

The first piece to build was a method to stream out… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI