Support additional models for ChatGPTBrowser

[adamrb]

Added models listed by the API for Plugins, Code Interpreter, & Browser: https://pastebin.com/raw/gB67XfWf

[danny-avila]

I'm curious what your responses look like with those added models. I can see the plugins model being useful as it has expanded context, but you wouldn't be able to utilize tools (without some changes to the request that I don't know how to handle).

Also curious because browsing and code interpreter also have special embeddings.

can you test this out and see what kind of responses are generated, in particular with browsing and code interpreter? I don't have access to those.

[adamrb]

I don't have access to browsing or code interpreter yet either. My thinking was that they could be interesting for the eventual langchain integration. I'm curious if the code interpreter or browsing plugins could be better at generating code, debugging, or understanding web pages. in addition to perhaps also having longer token lengths.

I'm definitely seeing value from the additional context length of running the plugin model even without any plugins enabled.

[danny-avila]

I’m going to test the browsing soon as I just realized I got access.

I don’t think I could leverage it with use for langchain though, not in a plug-and play sort of way. I could test it out though

[danny-avila]

@adamrb so I'm actually going to close this PR because it's not necessary, but it was helpful that you brought it up!

No modification to the code is necessary, you simply need to add the models to this env var in your api/.env file

CHATGPT_MODELS=text-davinci-002-render-sha,text-davinci-002-render-paid,gpt-4,gpt-4-browsing,gpt-4-plugins (i only have access to browsing and plugins)

I can see all the metadata and potentially build the same embedding that the official site does