I want to build an AI based report generation model in which the input to the LLM model will be a Json file with score’s of users result on answering a survey (our product). This score will be given in json format. Using this scores I have to generate bar, pie and line chart (as of now) all the charts should be interactive.
Where do I start from ?
what is the rationale for using AI in this scenario?
why not use a normal deterministic approach using a tried and tested library like Chartjs?
The reason behind AI is that
I need to add text content that relates to the survey
Including the charts
For example : There is this psychometric test called “RIASEC” here after the taking the test the user will have score for each tag score for R , I , A, S, E,C.
Using these score we can provide career path suggestions. This suggestion and visualization of scores using AI should be done(Charts).
One other reason is that:
Developers need not write code for these templates because we have many survey and psychometric tests.
I found this article with Bing: Generate Charts using OpenAI Code Interpreter — The API Way | by Sumit Sahoo | Medium
Might be of use?
(no affiliation)
I checked it out
It wont work for my use case
Check this website:
Piktochart
They generate reports
It’s also worth noting this on:
https://help.openai.com/en/articles/8437071-data-analysis-with-chatgpt
Q: How does ChatGPT analyze and visualize data with charts?
A: ChatGPT uses pandas to analyze your data and Matplotlib to create both static and interactive charts with your data.
You could probably reverse engineer that fairly easily
Thank you,
The problem posed here is that how can pass the chart to the front end.
For users
Can i send a react component code through api call in the backend.
All the contents that is generated using this LLM will be sent in JSON format.
This is what i think
I do this for our ‘weekly report’ - I give it a json with relevant data and let a write a weekly update with text and charts. You prompt the whole thing. And make sure that code completion is on. I shared some of that here earlier : How do you use the Assistants API? - #8 by jlvanhulst