can we change the AI model?

[feiyihx]

Open AI just release ChatGPT API using "gpt-3.5-turbo" model. Not only because that is the actual ChatGPT, the popular one, but also it is much cheaper than "text-davinci-003".

[jimzer]

Hey @feiyihx, thanks for proposing new features!

Right now, the app uses gpt-3.5-turbo. You can see it here: https://github.com/bitswired/ai-chat-app/blob/c99ea2c9bbe54a725c5c24958ba0233aca3b6bdc/src/server/api/routers/chats.ts#L73

Would you be interested in using other models, like text-davinvi-003?

[feiyihx]

@jimzer Hi thank you so much for the reply. Yes, I would like to use another model for myself.

Because I am in China now and my father would like to try ChatGPT, which requires an proxy server that could not be accessed from my apartment, I just deployed an docker on my NAS for him.

But I myself would like to try some other models. And your app is the most convenience one to use, especially when I am not at home I can just access the docker through web by my cell without the whole deployment and configuration.

Now that you've point out the file for me, I think I can just modify it.

Thanks again for your help and provide this great app.

[jimzer]

No problem :)

I will deploy today a version that handles multiple users. You can create an account for your father, and you will have separated data (chats, templates, ...).

Soon I will add support for changing the model and its parameters.

[turnercore]

We need to be able to update the models to the chat-gpt models, I believe right now there is just gpt-3.5-turbo, gpt-3.5-turbo with a specific version, and a couple of gpt-4 models. But as some users have beta access to the gpt-4 models already it's probably important to include them (and will be more so moving forward).

BTW love the app so far, can't wait to contribute to it when i get some time!