Fine-Tuning X document storage

Hello guys when I want to create a assistant that doesn’t exactly make searchs on documents or talk with data base is just a assistant with some fixed documents as base on how some thing work and the user needs some help.

It’s better to make something like that with Fine-Tuning, api document storage or assistants vector storage?

And if none of these are the way embeddings are rly the only way?

Thanks in advance.

You cannot produce any model fine-tuning that emulates the internal tools within Assistants.

The best you can do is instruct the AI, and understand the existing tool internal instruction language (not published but you can have it repeated back) that may work against your use-case so you can countermand that.

The provided file search uses embeddings. It is just general purpose, and anyone developing a specialized application would write something better to ground AI with knowledge.

2 Likes