> IMO, you shouldn't use the assistant API unless your use-case fits it well. I believe it's good if you don't want to bother hosting an agent or implementing RAG yourself. So, it's great for PoCs and MVPs. I wouldn't use it for local agents. I'd only use it as a temporary solution in order to get product out the door quickly. Then when I have more time, I would switch to the chat API + RAG in order to get more control. I'd write an API wrapper class over the assistant API to make this move easier. However, if your devs aren't skilled enough to do that or you just can't find the time, sticking with the assistant api might be best. ## All https://allgpts.co/ ## Parameters - name - instructions - tools - model ## Benefits - multiple custom instructions going at once, which was a big pain with the previous iteration of ChatGPT - It also saves uploaded documents via [[GPT Retrieval]]. So instead of uploading the same documents over and over I can create a baseline for each new chat I have with it.