-
Notifications
You must be signed in to change notification settings - Fork 694
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Regenerate respone #436
Comments
As it is quite useful for my application, I'll give $100 to the person who implements this feature first ;) |
Awesome! I'll check it out later today. Thank you 🙏 |
I checked it out in Windows. I see two issues. P.S. I made the payment here. |
Alright,
|
out of curiosity, what are you building? is the source published yet? It seems really neat! |
I'm using it to tutor my students in Disocrd (so it helps as a TA). We can't reset the memory in a channel since the bot needs to know the context so we can refer to previous points. The problem I'm trying to solve is to remove the output from the context when the bot is giving a wrong or vague response. We just want to replace the last output. this is exactly how the chatGPT interface works, each time you get an output, you get the chance to replace it at that point (but not later when the next output is generated). |
Hello, |
Right I agree, but until bugs fixed and become a general-purpose button like openAI interface, it may be better to have a /slash command that only users who know what they are doing be able to use it. |
alright, sounds good |
I moved the regeneration into a slash command, it does not work for images as of right now. |
I wanted to get a reaction based counter on the message to show the number of regenerations but I cannot figure out how discord.Interaction works with reactions. But let me know if this is what you were looking for |
No worries about that, it's not really useful for my application anyway. The priority should be to remove the previous generation from context memory since currently it's not regenerating, but rather redoing given previous attempts. But I need a sampling function that can reattempt without memory of the previous attempts. Since when it keeps the previous attempts in memory, the new attempts look very similar so it quickly gets stuck in the same frame of mind, but having the ability to regenerate from scratch gives much more brainstorming power |
I finally understand what you're saying! I'll take a look at the OpenAI API docs and figure how to do this correctly. Sorry for the confusion |
Thaanks. Sorry for the misunderstanding |
I don't think what you want is possible due to limitations in OpenAI API: api reference |
Looks like possible if you use chatCompletion API which allows you to store and edit the entire conversation, basically it makes the API memoryless https://platform.openai.com/docs/guides/gpt/chat-completions-api |
also, it allows you to edit any part of the conversation in the past which would be a pretty cool feature to have as well |
oh, awesome! |
Hey @shinysocks, any update or plan on this? |
I haven't yet taken a look at rewriting the API requests. I'll check it out sometime this upcoming week hopefully |
Awesome. Looking forward to it |
Where did you find information in the API to regenerate? I can't seem to find it. |
|
@Zero6992 Does this bot even use the official openai api? |
@shinysocks it's using this https://github.com/acheong08/ChatGPT, which inside is using the official openai api |
Looking that this api's documentation, it doesn't look like they include a regeneration. If you want to utilize this specific feature you may need to open a pull request on that api. |
Yea looks like using that API was a mistake, as the project is dead. I found a better discord chatGPT, which also has the regenerate button. I'll give it a try: https://github.com/Kav-K/GPTDiscord. |
Awesome, that one looks great. |
Hii,
![image](https://private-user-images.githubusercontent.com/15813444/252073132-308ab80b-1c74-433e-bc38-9fa47bf3a255.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.eEuCwHNLYvqWQkiszmWYlsTZPcOYFoAunZMsbUxxlPc)
Thank you so much for this project.
Is it possible to regenerate the response using a command similar to the ChatGPT's interface here?
Cheers,
Arash
The text was updated successfully, but these errors were encountered: