Advanced Data Analysis Feature using OpenAI API

API
1

I have been using the beta feature Advanced Data Analysis through ChatGPT 4, by uploading a set of csv files and asking questions related to the
data stored in these CSV files.

I am working on implementing a chatbot using the OpenAI API. The API to upload a file seems to only accept JSON files for fine tuning purpose. Is there actually a way to upload a CSV file and then reference this CSV file using the ChatCompletion API?

2

Fine-tuning is only meant to instruct how the model should answer.

If you want to do some QA over CSV fikes you could use embeddings to get the important part and then pass the result to the model so it considers it when creating the response.

3

Thanks for the prompt reply, but not sure to fully understand. I am a newbie when it comes to OpenAI API and not familiar with embeddings. I have the chatbot in place with a simple front end, but I am not finding a way to upload a set of csv files using OpenAI API and then reference these files through the ChatCompletion API.

4

The “advanced data analysis” aka “code interpreter” is a python function-call framework provided to the AI, where the AI is informed that python function output will be run in a jupyter sandbox and a function value output returned to it.

Implementation would require setting up a python environment that can be called by the function output to run the code without the AI running rampant.