Using GPT-4?

[morethanx]

Hi,
I'm having so much fun using your chat plugin to search my notes better and get insights. I was just wondering if I was able to change the app for myself to use gpt-4-0314, and when I went through your code to see if I can adapt it for myself without having much experience in coding I found this part:

const DEFAULT_SETTINGS = {
api_key: "",
file_exclusions: "",
folder_exclusions: "",
header_exclusions: "",
path_only: "",
show_full_path: false,
expanded_view: true,
log_render: false,
log_render_files: false,
skip_sections: false,
smart_chat_model: =="gpt-3.5-turbo",==
results_count: 30,
view_open: true,
version: "",

At smart-chat-model, it says 3.5. Doesn't that mean, it's always 3.5, even when I have a GPT-4 API? Sorry if this is a stupid question, just a newbie here. :) and again, thanks!

[brianpetro]

I'm having so much fun using your chat plugin to search my notes better and get insights.

I love to hear that. Thanks for letting me know!

You can change the model to GPT-4 in the Smart Connections settings.

If you don't see the option, make sure the app is updated using the "Check for updates" button in the "Community plugins" section of Obsidian.

I hope that helps!

Brian🌴

[morethanx]

Thanks for such a quick response. I already changed it to gpt-4, but it says it doesn't know about any gpt-4 because it only has data up until March 21.

[brianpetro]

@morethanx can you send a screenshot?

[morethanx]

Im not at my computer now, but can certainly do that later.

[morethanx]

Hi there, sorry having taken so long. I attached some screenshot. Somehow it is not working at all anymore now. I asked it which GPT it is and it's been working on responding to this for ages (and still didn't come up with a response. The first time this happened was March 17, you see this in my usage history. I closed obsidian at one point because it just didn't stop, but since then it hasn't been working properly anymore. Would you delete and reinstall?

[morethanx]

hi, I reinstalled it and asked it which version it is. It works again, but it still says GPT-4? Is that just a glitch? I double checked, I put it on GPT-4 in the settings.

[brianpetro]

@morethanx I'm glad that we seem to be making progress 😊

If you're trying to verify which model is being used, asking the chat isn't a reliable method. Instead, look at your OpenAI Account Usage for a more accurate picture.

Additionally, I received some errors from the GPT-4 API this morning. This led me to improve the error handling in the latest version, 1.4.1.

Let me know if that helps clear some things up for you.

Thanks for following up,
Brian🌴

[morethanx]

Thank you! It still tells me that it's only using data until Sep 21.

[morethanx]

Without having looked at the code. I saw that one has to write gpt-4-0314 in the code instead of gpt-4 (you probably did this, I wrote it, just in case) :)
https://platform.openai.com/docs/models/overview

[brianpetro]

@morethanx asking an LLM model about itself is not effective. The only things it knows about itself are what it has been told via prompts. The September 2021 date simply comes from a prompt that is included behind the scenes by OpenAI.

[morethanx]

alright. I'll stop grilling it then. :) Thank you!

[gillchristian]

Hi, amazing plugin! 🔥

Reusing this issue since it's related, I just installed it on a few Vaults, enabled the billing on the OpenAI account and set the model to GPT4 but I'm still getting errors when trying to chat.

The request to this endpoint https://api.openai.com/v1/chat/completions is producing the following error:

{
    "error": {
        "message": "The model: `gpt-4` does not exist",
        "type": "invalid_request_error",
        "param": null,
        "code": "model_not_found"
    }
}

Here's the usage in OpenAI

I tried the hard refresh and it didn't work.

What did work is to switch to the GPT3 model. Could I be missing something on OpenAI side? Maybe GPT4 is not yet enabled there? Do I have to join here https://openai.com/waitlist/gpt-4-api ?

[brianpetro]

Do I have to join here https://openai.com/waitlist/gpt-4-api ?

Yes, the GPT-4 API access is still limited.

[gillchristian]

Thanks for the clarification. I assumed that limited access meant the paid account was needed 😬