Add turboGPT.js

[14-3dgar]

TurboGPT Extension

Adds an extension that lets interact with chatbots of gpt-3.5

• No api key required
• Lets make simple prompts and interact with multiple chat sessions

• You can test it here

[LilyMakesThings]

You thumbnail's SVG is using a text element, which isn't allowed, which is why it's failing

[CubesterYT]

lolemo, let me give you a heads up, if this does get merged, I want to tell you that the extension can't be named with scratch in it's name, as GarboMuffin doesn't want TurboWarp to be associated with Scratch in any way. He told me the same thing with my extension, called TurboHook, which was called ScratchHook originally. So try renaming the extension to TurboGPT.

[14-3dgar]

lolemo, let me give you a heads up, if this does get merged, I want to tell you that the extension can't be named with scratch in it's name, as GarboMuffin doesn't want TurboWarp to be associated with Scratch in any way. He told me the same thing with my extension, called TurboHook, which was called ScratchHook originally. So try renaming the extension to TurboGPT.

thanks, I just changed the name

[XeroName]

this is really interesting
love this

[Flappy25]

I wonder if you can use GPT-4.

You can, if you know python

You'd need an API key but i think it's expensive
This extension is gpt-3.5 only, but free and faster than gpt-4

Use the GPT-4 free project

[ghbbeep]

This is a really good extension! Something that could be useful is a block that prints the last thing generated by the AI. Otherwise, this should really be on the extensions website!

[BlueDome77]

I wonder if you can use GPT-4.

People say ChatGPT 4 is getting dumber

[Flappy25]

I wonder if you can use GPT-4.

People say ChatGPT 4 is getting dumber

Maybe cuz the internet is full of fake news or openai is trying to suppress it's knowledge

[ddededodediamante]

so cool

[14-3dgar]

It seems the API is bugged right now, is it down? If so, how often does said API go down?

It works well for me. If the extension doesn't work it's because the OpenAI API is down, probably due to maintenance. Could you send any error messages you get in the browser console?
You can see the openai status page to see how often this happens here
Also, it the console displays an error message with code 429, that is because you are sending requests too fast

[14-3dgar]

Added new blocks

Last generated thing and prompt, suggested by @ghbbeep :

Exporting and importing specific chats:

Reporter to list active chats:

Export everything:

Import and merge, or remove and import:

You can test it here

[True-Fantom]

Hello! I really like your extension, but I also want to point out the shortcomings that can be easily fixed.

Block icon is black, which does not correspond to the scratch style, repaint it in white.

It is not clear from the extension what some blocks are responsible for, it can be solved by adding documentation.

Also, it is not clear from the extension which GPT model is used, and even if you add it to the documentation, it will not be convenient. So you should probably change the name of the extension to "GPT-3.5", it will be more concise and without the "Turbo" prefix, which is useless in this case.

[Drago-Cuven]

@Lolemo this ext is currently broken.

[ghbbeep]

@Lolemo this ext is currently broken.

This has happened in the past, presumably because it's a free API designed for a different service. Perhaps add a label warning the user that the API may go down often.

Another idea is maybe you can either select from a dropdown of known/reputable APIs, or type one in yourself. But as it stands right now, this extension isn't really useable.

[GalacticCoderReturns]

This is cool! I have it in my PenguinMod Library right now.

[GarboMuffin]

The problem with these ends up being that LLMs are expensive to run; someone has to be footing the bill. It happens that the free API of choice shut down since this pull request was created.

[GarboMuffin]

#497 is still open

[zmh-program]

The problem with these ends up being that LLMs are expensive to run; someone has to be footing the bill. It happens that the free API of choice shut down since this pull request was created.

Please don't have that mentality with ai. You have no idea how cheap chat completions are for llm nowadays (including cf's free stable diffusion, etc).

[LilyMakesThings]

The problem with these ends up being that LLMs are expensive to run; someone has to be footing the bill. It happens that the free API of choice shut down since this pull request was created.

Please don't have that mentality with ai. You have no idea how cheap chat completions are for llm nowadays (including cf's free stable diffusion, etc).

Well who's gonna pay for the extension's API, then? Because TurboWarp would have to offer it for free and then what happens when other websites start using it for free?

[zmh-program]

The problem with these ends up being that LLMs are expensive to run; someone has to be footing the bill. It happens that the free API of choice shut down since this pull request was created.

Please don't have that mentality with ai. You have no idea how cheap chat completions are for llm nowadays (including cf's free stable diffusion, etc).

Well who's gonna pay for the extension's API, then? Because TurboWarp would have to offer it for free and then what happens when other websites start using it for free?

U DON'T EVEN know cf free llm.

[GarboMuffin]

There is nothing stopping you from using any of the HTTP/Fetch/Network extensions to use any API you want

But when something is submitted here, the idea is that it's going to keep working basically forever. Free AI services demonstrably are not reliable; again see how the original website that this PR used unceremoniously shut down one day.

[zmh-program]

So you think the merge was stopped because cloudflare would be shut down? As well as it might be good to know about the website I wrote (chatnio.net). Also, my recommended practice is to understand the llm circle rather than just assume so.

[GarboMuffin]

if you think you have a way to solve my concerns then you are welcome to submit a new pull request

[zmh-program]

I think you and the developers in the turbowarp community could tap into it if you're interested via chatnio api (supports openai calls in the same format)

(My non-profit site chatnio.net has 50,000 people registered, just don't think it's a small site)