Codebase files embedding finetuning support

[mlaugharn]

It would be very useful if this plugin allowed finetuning a chatgpt model on document embeddings of the files in the current project/codebase. One tool that does this already is qqbot - https://marketplace.visualstudio.com/items?itemName=qqbot.qqbot

if this already exists, my apologies!

[carlrobertoh]

Hi @mlaugharn

Thank you for the feature request. Unfortunately, at the moment the plugin doesn't support model fine-tuning, however this is indeed a really good idea and can be useful in many different cases.

I will try to plan and implement the first phase in the next upcoming days.

Best regards

[versecafe]

Instead of fine tuning to massively improve the use of the context window, making vector embeddings from documentation and using vector similarity search to find relevant documentation to pass into the context window could be much more effective and far far cheaper then retraining, using openAI ada-002 for it cost less then a cent to generate embeddings for my companies internal documentation breaking down MDX files.