Basic Discord Bot to chat with text-davinci-003 with mod api support

[solar marsh]

Project help channel for the code on github!

THE BOT ISN'T HERE. Please don't try to prompt it (I will be deleting them to clean up)
Also please don't advertise your own HOSTED project here. If you made a fork with more features, feel free to post the link to your github 🙂

If you're here from github, say hi!

https://github.com/openai/gpt-discord-bot

You connect your openai account by inserting your api key. Your account will be billed for token use. See https://openai.com/api/pricing/

[uneven birch]

Yo this is great, I was planning to do something like this too, but didn't find the time yet :) pretty unfortunate, that chatgpt is not possible to be chatted with, this way it's kinda less nice to chat with it

[warm shore]

namaste ai

[ebon plover]

Congrats on the release :D

[odd star]

Hi

[open pivot]

What if users feed malicious input to it? Does that risk getting my account banned?

[solar marsh]

The models are pretty similar

Feel free to fork and add your own features

[stray flax]

is it paid?

[solar marsh]

It uses your openai account to bill for token use

[stray flax]

like, do I pay money

[solar marsh]

Yes you are billed for token use

https://openai.com/api/pricing/

[snow fulcrum]

stop=["<|endoftext|>"],
what does this do exactly? Wouldn't the bot normally stop at <|endoftext|> anyway?

[solar marsh]

I think that prevents it from returning the token

I was seeing it return conversations and like fill in what you would say lol

[gritty shadow]

Hi

[snow fulcrum]

haven't dug into the code but if you give the AI a conversation formatted like

Human: text here
AI: text here
Human: message here
AI:

Then you can just use Human: as a stop sequence to prevent that

you might also consider giving \n as a stop sequence to prevent the AI from giving multi-paragraph responses

[solar marsh]

Human in this case is your username and there can be many people

I think you need beeline

Newline

Like if you want a story or code

Feel free to fork

[snow fulcrum]

I have a working telegram bot with gpt3 and what I described above is basically what I do. I set the stop sequences dynamically though

but hey as long as it's working, all roads lead to Rome

[exotic saffron]

Isn't "engine" deprecated ?

[solar marsh]

Where do you see engine

[austere sky]

Hi

[calm shoal]

Hi

[fervent wharf]

Hi! Interesting work, congrats.
How is this different from talking to davinci-003 on the OpenAI playground, though?

[exotic saffron]

In one of yours script

I'll tell you

https://github.com/openai/gpt-discord-bot/blob/main/src/completion.py#L52

[dense jewel]

is the discord bot down

is the discord bot down

[exotic saffron]

😁

This work like a ping

Dont ban me 😱😱

[solar marsh]

Thank you!! I’ll fix it

[solar marsh]

What are you talking about

[solar marsh]

It’s the same, the bot formats the conversation for you

[fervent wharf]

Then I’m confused (and I’m asking out of curiosity, not trying to be an asshole) what’s the point?

Just the convenience of Discord?

[solar marsh]

Np! Chatgpt is also just a model you run completions on but the website formats conversations for you

So you can talk to it instead of just having it complete things

[fervent wharf]

Gotcha

[solar marsh]

Yeah convenice and it’s fun to have multiplayer conversations!

Friends can join in talking to the bot

[exotic saffron]

May I ask you some questions Moxi about fine tuning ? I'm learning on the subject before starting to develop my project

[solar marsh]

Yes! Please ask here

Though I don’t know much, I mostly work on dalle, but I can ask around for you

[dense jewel]

the gpt discord bot is offline and did not reply me anymore

[solar marsh]

What bot??

There isn’t a bot hosted here for the public

[exotic saffron]

ChatGPT discord bot would break the internet

Would be fun

[exotic saffron]

I'll do!

thanks!

[dense jewel]

umm the bot we created using api

[solar marsh]

I don’t know what you’re talking about

[dense jewel]

https://github.com/openai/gpt-discord-bot this one , i created the bot in my discord server but suddenly it went offline and did not reply me anymore

[solar marsh]

I can’t help you with that, you are hosting it yourself

Unless it crashed or has an error message, then share it

[solar marsh]

Did it work before? Share a screenshot?

[exotic saffron]

Is chatgpt using code davinci 02?

[solar marsh]

No

[exotic saffron]

So when you ask to write a script, it doesnt switch of model.

[drowsy prawn]

HEI

[still jay]

Hi

[fast heart]

hi

[sharp ibex]

Can't for the life of me confirm if I have a user profile setup in discord. Trying to make the bot keep track of all my conversations with it. Doesn't seem to know what we spoke about in other chats. I'm not a discord n00b but I'm not advanced by any stretch. Anyone have ideas?

[astral rain]

Vb

[ebon plover]

That will need to be coded in, currently there is no bot that enables that AFAIK

[royal lance]

Hello I have experience on discord bot development.
Can we have an interview for discussions of your project?

[solar marsh]

Sure, what are your questions?

[royal lance]

Could you give me detailed requirements?

hi

I am waiting your detailed requirements

[near kindle]

@royal lance what are you talking about? The project is finished and Moxi has released it on GitHub available at https://github.com/openai/gpt-discord-bot 🙂

[ebon plover]

By requirements do you mean like a software requirements spec? Like functional and non functional requirements?

[royal lance]

SRS

If there is no SRS, you can explain me functional requirements

[ebon plover]

Tbh I haven't heard this term outside universities and big consultancies in the UK. So you may likely need to ask the question in other words :)

[royal lance]

Software Requirements Specefication

I mean your project's requirements

[subtle hornet]

you can add discord stuff there.
like if i command it to delete text, kick user, it will do it

[stuck mesa]

Playing catch-up, have created the discord bot, Git let's you create a docker and it works fine for testing a playing around. My question is one of context - each chat seems detached from the previous, so the model is fine but history is not - is that only available through the web ChatGPT instance?

[near kindle]

Created what discord bot? This is a channel for the OpenAI open source discord bot for GPT3, there were no docker instructions attached, can you provide more info?

[stuck mesa]

There is another open source (on mobile will find link) discord GPT3 based bot. It is written in Python and has docker instructions for hosting (GIT dev env worked for testing) it is not as rich as others, as context is lost in chats.

[sweet gorge]

Memory is per thread I believe.

[solar marsh]

Wrong thread

[wind crest]

It's amazing to me how little attention GPT3 got compared to ChatGPT because of the puny less amount of work you have to do to get a response with ChatGPT

People are asking me how I generated 28,000 item lists with ChatGPT

uhhhhh I didn't use ChatGPT

Don't know why it bothers me so much, definitely silly on my end

It's just like, this amazing technology has been available for like two years and because it took slightly more work and thought than "Give me ai image prompts!!" people completely ignored it.

[solar marsh]

What is pm2

[near kindle]

It’s a process manager for nodejs

[solar marsh]

But the bot is in python and uses pip…

[near kindle]

😂😂😂😂

[tropic anchor]

I know it's python and pm2 supports python files too. It's very useful if you have a lot of bots or projects.

[solar marsh]

Cool I’ve never used it

[sharp palm]

I've never set up a discord bot before. Is it a simple process?

[acoustic delta]

hello

[cloud dune]

I followed the steps and got to number 7. My env file is saved and ready to go but when I ran:
"pip install -r requirements.txt
python -m src.main"

I got the message 'command not found: pip"

Any idea how to resolve? Thanks.

I think I have to install dependencies but those commands were not provided.

[cloud dune]

Now I installed dependencies and used pip3 and it was working but after several minutes it said, "no module named src"

nvm I had python in command but it needed to be python3

Bot is in the server but every command returns, "The application did not respond". Any ideas what I did wrong?

I checked permissions fyi

nvm I fixed it it needed channel permissions (not listed in github as one of the required permissions)

[wide stump]

#how can I use it

https://giphy.com/gifs/reaction-childrens-miracle-network-hospitals-cmn-3bc9YL28QWi3pYzi1p

[ocean pumice]

hey hi every one I was just working on chatgpt bot for discord but i have encountered this

openai.error.RateLimitError: You exceeded your current quota, please check your plan and billing details.

can anyone help me with this?

how can I buy more limit?

[remote crest]

hi

[stiff kraken]

hi

[ebon plover]

Need to use the form and wait. I have been waiting for some weeks now

[lucid shuttle]

Just to confirm if I use https://chat.openai.com/chat that is free correct, but if I want to add a discord bot to my server I use that $0.0004 / 1K tokens?

[solar marsh]

That’s only if your channel is private

[cloud dune]

Right. Thank you, it works great! Where do I change limits on message count and context limit? Is there an upper bound on message count before the bot closes a thread?

Also, is there a way to not have it respond to certain messages? I trained it to respond to messages appended with 'itb' with a . which is better but ideally would not respond at all but it's no response was (Empty Message). This would be useful in multiperson chats where not every message is directed at the bot.

[solar marsh]

Constants.py for message

Context limit is a model thing, it can’t get bigger

[solar marsh]

You can check the message content in the handler in main.py

[cloud dune]

Yeah I think it said 50 messages for context limit so I guess you can refresh it's memory when you go past 50 messages assuming it is like the accounting method FIFO first in first out.

[solar marsh]

Context is limited by tokens, the api will return error

Yeah you can drop earlier messages if you want

You can try @ebon plover ‘s bot https://github.com/firtoz/GPT-Shell which has long term memory

[worldly marsh]

any plans to port over to slack too?

im too novice to do it myself, but trying to get there

[solar marsh]

No plans

[languid crescent]

Hi

[gusty plaza]

is this the right place to ask gpt-discord-bot questions?

[cursive hemlock]

hey im getting an error when trying to start it, it says no module named "discord" exists

[gusty plaza]

this is somewhere between a question and a statement - how on earth does the config.yaml work in terms of naming the bot, and the conversation structure? the line for instructions: is totes obvious, but i'm drawing blanks on name: and messages: 😄

[dense dragon]

hello

[hoary veldt]

same for me

[gusty plaza]

what files are in that directory?

[hoary veldt]

im using external libraries in pycharm, that may be my reason.

Package containing module 'discord' is not listed in the project requirements

[near kindle]

You can't use external libraries in pycharm to get this bot to work, install the requirements with pip install -r requirements.txt (double check the README though)

[solar marsh]

Add a print to see how it is parsed for the completion prompt. They’re just examples to tell the bot what to do

[gusty plaza]

Hm… can you explain what to do like I’m 10?😇 sorry to be bothering with this

[near kindle]

Check out the project README to get started with setting up the bot, but you require a bit of a programming background to do things like install python, and install the requirements

[gusty plaza]

I’m running the bot just fine, what I don’t understand is the name and message configuration

[near kindle]

Oops! Sorry I misread

But for that, see how the prompt is parsed: https://github.com/openai/gpt-discord-bot/blob/90fce7259aea3faefef92054a684c94c7c66bff5/src/completion.py#L45

[gusty plaza]

Ok thanks, I will check it out. 🙂

[tight ivy]

So you cant do this with a free account? payment method must be entered?

[finite parrot]

you dont change the Name in the config.yaml ... it loads the name dynamically:

completion.MY_BOT_NAME = client.user.name

it uses whatever name your discord bot has as client.user.name ... in the config.yaml you wanna keep the name as Lenard, that will be exchanged when the bot starts and loads up the yaml file

also check constants.py if you wanna learn more about

[gusty plaza]

Oh that’s interesting. The name may not matter in the yaml yeah? If I ask my bot what it’s instructions are it replies with the string currently in instructions. I assumed the name field did the same

[finite parrot]

yeah it changes the name dynamically at startup in name: and all the messages in the example_conversations: :
https://github.com/openai/gpt-discord-bot/blob/de808060c1cb57bcf504aa8f5b367c2515dc04b4/src/main.py#L45

the name in the config.yaml matters, if you replace it there you also need to replace the hardcoded name there in main.py

as its hardcoded to search for "Lenard"

[solar marsh]

yeah sorry i got lazy there

[gusty plaza]

I don’t feel so silly asking about the file now

[azure knot]

just added a gpt-discord-bot to me server and its responding properly, i thought you had to pay to get chat tokens, but is it that im just limited to a certain amount of chats per hour?

[eternal marsh]

did u set ur api key?

[azure knot]

yes

[eternal marsh]

u get a free amount of credits

[azure knot]

so a one shot amount of tokens?

[eternal marsh]

they usually give u about $18 worth of credits

[finite parrot]

you can check how many free credits you have left here: https://beta.openai.com/account/usage

[azure knot]

Here's to hoping openai eventually provides a flatrate subscription model

[solar marsh]

Please use #dev-chat and not my project post

[errant ferry]

is https://github.com/openai/gpt-discord-bot based on gpt3?

[near kindle]

Yes 🙂

[errant ferry]

thanks! Playing around with it now for programming help

still unsure what the benefits of using it over the webpage chat but will see

[exotic saffron]

That made my discord crash...

Hey Moxi,

I just started an internship in a web agency and they are asking me to develop a tool using OpenAI API in the upcoming weeks. I heard of a GPT 3.5 API that will be released in the future. Am I wrong? Should I wait for the next API to drop before developing my tool?

Here (to ping you)

[near kindle]

GPT3.5 is already there as an API, it is text-davinci-003 and text-davinci-002, code-davinci-002, the API that will release soon is the ChatGPT API. Please ask this question in #dev-chat or #1037561751362863144 instead of on project posts 🙂

[exotic saffron]

Moxi told me that I can ask him questions here a few months ago

And that it was ok

I understand and thanks for the information

#1055336272543092757 message

[near kindle]

Ah i see, my bad!

[exotic saffron]

I was using answer to ping that day and he said that was ok so I don't get censored by openai bot lol

[exotic saffron]

Its ok dont worry, you dont know everything

[solar marsh]

Just use the completions api it’s already public

What kaveen said is correct

If your questions aren’t related to my bot please post in the other channels

[errant ferry]

wait so there's a difference between chatgpt api and gpt3 api? Is that why using https://chat.openai.com/ seems better for technical questions than using the bot

[bleak flume]

hoLA

[solar marsh]

They’re different models yes

And the bot has a preprompt that it’s human

[errant ferry]

is there a resource where I can see the overview of the differences?

[solar marsh]

Not sure, there’s this blurb here https://openai.com/blog/chatgpt/

And the link to instructgpt

For technical questions I think it’s just the preprompt

Try changing the bot preprompt?

[errant ferry]

oh interesting, I'll play around with it for sure

[rain acorn]

Millions of people go to McDonalds every day because...cheap, fast, easy.

[prisma kettle]

shouldn't the openai calls be wrapped with a thread executor?

or else it's blocking the asyncio event loop (therefore the bot is dead while it is making an openai api call)

[finite parrot]

Yes they kinda should, Lenard runs fine if you just running him against basically a single user, but multi-user will be hard for this Codebase.

So you totally correct its blocking the asyncio event loop, when OpenAI API has a problem even up to the point you have to restart your Bot.

[pearl sun]

hello

[solar marsh]

did you file the issue? i just commented. ill get to it soon! thank you (but feel free to send a PR if you got it working)

[hollow egret]

Failed to start chat No API key provided. You can set your API key in code using 'openai.api_key = <API-KEY>', or you can set the environment variable OPENAI_API_KEY=<API-KEY>). If your API key is stored in a file, you can point the openai module at it with 'openai.api_key_path = <PATH>'. You can generate API keys in the OpenAI web interface. See https://onboard.openai.com/ for details, or email [e-mail] if you have any questions.

why this error appeared? i'm confused. i already specify the openai_api_key in constant.py

[solar marsh]

This is erroring from the openai api

I am unsure as it’s not my error

Make sure your key is passed to the api

[tribal elm]

Hi there, I was talking with YufeiG on GitHub about an error I was having with getting the Discord Bot running in Python. was told to come here for more direct help. (I am a novice and am probably making some simple mistake I bet)

[solar marsh]

Hello

Can you pull main and try again

Also share your command that you’re running

[tribal elm]

I have to admit I have no idea what that means 😅

[solar marsh]

Download the code again and run it again

What command are you using to run it

[near kindle]

Hey! How'd you originally download the code for this project?

Download it again the same way that you did originally, and save it into a new folder, and then you can try running the bot again with the new changes that Moxi has put in, it may have fixed your previous issue

[tribal elm]

Oh I see, Okay, I can totally do that.

[near kindle]

Sounds good, try to use the same command that you used to run the bot earlier as when you got that error from above with "No module named 'src'", what command were you using anyways?

[tribal elm]

Maybe I should ask what command should I be using? Because it sounds like probably am doing this all wrong, with how little I know. I did fill out all the pertinent information and had all the python files present as it was outlined in the project, But I don't know if there's a specific way to run it, I would just try executing the main file, and it sounds like that is probably very wrong xD

[near kindle]

Ah I see, how were you executing the main file? By just right clicking and running it with python?

[tribal elm]

hehe yep

[near kindle]

Can you try it again, after downloading the files again?

[tribal elm]

absolutely 👌

[near kindle]

Let me know with a ping what happens

[tribal elm]

Thanks! and ty for the help! 🙂

@near kindle Same result!

[near kindle]

I see, are you familiar with Command Prompt on Windows?

I want you to open a command prompt window in the folder that contains the files for this bot, are you able to do that?

[tribal elm]

sure thing ✅

[near kindle]

Then, inside that folder try running the command python src/main.py

[solar marsh]

python -m src.main

[tribal elm]

I got this back:

[balmy estuary]

Can you try moving main to the previous folder (outside of SRC) and then try running it like you were before?

Main.py

[tribal elm]

Just tried that, the window opens then disappears in a fraction of a second. cant see what issue it generated

[balmy estuary]

That's because it's running correctly

Now, using CMD prompt, with it inside the directory

Type python main.py

[tribal elm]

command prompt from the folder where Main.py is currently? or do I return Main.py to where it came from and do this?

[balmy estuary]

As it is

[tribal elm]

okies

[balmy estuary]

Where main py is

[balmy estuary]

Type pip install discord

[near kindle]

oh do pip install -r requirements.txt

[tribal elm]

That's done now, thanks! So now when I try the command again I get:

[near kindle]

^ this should fix that 🙂

[tribal elm]

okies!

ty for putting up with my noobness! Appreciate it!

[near kindle]

No worries at all, it is definitely confusing on windows

[balmy estuary]

It helps make the program bulletproof

[tribal elm]

This is good, yea?

[balmy estuary]

Auyy

Yup

[tribal elm]

woohoo 😄

[near kindle]

perfect!

So yeah, you can't really run python files with right click in most cases, we have to use terminal 🙂

[tribal elm]

eyyy, it joined my server, thank you so much ❤️ ^_^

[tribal elm]

that is VERY good to know!!

yay it's working. Thank you again all of you, I really appreciate the help ^_^

[tribal elm]

Is there any trick to get it to adhere to the Config file on how to speak? I've noticed it doesn't seem to be aware of whats in there

[solar marsh]

It should be using it, what do you mean it’s not aware

[tribal elm]

Okay, it took me some time, but ChatGPT helped me fix the config.yaml file's formatting errors I had made though my inexperience, and now it is working perfectly. 👌 ✅

[fluid dove]

Idk why I can't chat

How do I fix it?

[solar marsh]

What is the error message

[celest sedge]

When i ask the bot to write some code for me it tells me that it will do it in few hours. The code i have asked to write is pretty basic

How can i modify it to actually start doing something?

[solar marsh]

This is a chat bot. For casual conversation. Change the prompt in config.yaml to be ai assistant if you want an ai assistant

[celest sedge]

Can you help a little bit on how I can achive that? With a sample config that could somehow fit my purpose? Thank you

[solar marsh]

Just change the prompt in the config to what you want

[celest sedge]

You are talking about this config ?

[solar marsh]

Like instead of you are discord user change it to you are an ai assistant

Config.yaml yeah

That tells the bot how to act

[celest sedge]

oh 😄

I made the following config

here is the result

basically i have to give him all the possible answers that someone will ask?

[solar marsh]

no, the examples just tells the bot how to talk

like if you want them to use slang or something

its not that necessary

[celest sedge]

For example if i would like the bot to help with configuration from a specific api that is probably not present on the open ai website, could i add it here? or where should i add it?

Also getting this error pretty often

[fluid dove]

How do I enable moderation

[solar marsh]

it should be by default

[solar marsh]

sorry i cant see the error in your screenshot. what is the crash reason

[solar marsh]

what do you mean

[celest sedge]

Btw I have build a very big dataset and now discord doesnt let me even start /chat :)))

Is this limitation only from discord or if i switch the client i should be fine?

[solar marsh]

what is a dataset

[celest sedge]

I have parsed a whole html documentation and made the headers as user bob text and the paragrphas under that header as user ai text 😄

now my prompt has 9000 tokens, i think its a little bit too much 😦

[solar marsh]

i think model max is 4k something?

you can just test your prompt on playground first

[tribal elm]

Hey sorry another question, is it only able to be set to reply in Threaded conversations, or is it possible to have it be in a regular chat channel(s) instead/too?

[solar marsh]

you can change the code to work in regular channels

its just discord api

like theres a thread check in on messages, if you just remove that

should work

maybe

lol

[fluid dove]

Im getting this error message

[finite parrot]

That's an answer from the OpenAI API not sure if there are still problems, for me it all works fine, did you give it another try?

[eternal marsh]

WHY DO i need to add ALLOWED_SERVER_IDS?
can't i just leave it blank so anyone can add the bot
why do i need to add it manually for every server?

[finite parrot]

You can just remove the check: https://github.com/openai/gpt-discord-bot/blob/90fce7259aea3faefef92054a684c94c7c66bff5/src/utils.py#L66

Its good design to have an allowed server list, you will pay for completions done by people that add the bot.

[solar marsh]

Sure you can remove it if you want to pay for everyone’s

I had the issue where people I didn’t know kept adding it to their server

[umbral merlin]

hi

i am having this error

Error - You exceeded your current quota, please check your plan and billing details.

While chatting with my discord bot

please help

[finite parrot]

Seems your account has ran out of free Credit, like the message says you need to check your billing plan: https://beta.openai.com/account/billing/overview

Its not related to this Bot at all.

[umbral merlin]

I just created my account around 1 hour ago

How can it run out of it if I haven't even used it?

[orchid nacelle]

This isnt the place for this

Try API Support. And instead of posting a log of messages

Try asking for help

[umbral merlin]

In OpenAI's Github, It redirected me here

[finite parrot]

You have to contact OpenAI directly for this, the Discord People here can't look into your account and billing details.

[orchid nacelle]

Ok, and I'm directing you to #1037561751362863144

[umbral merlin]

Ok thx

[solar marsh]

Ohh I think someone reported this issue… I tried looking at the code last night but didn’t get anywhere. Dm me your email and I can look later when I’m home

[umbral merlin]

You can talk me on my discord only

[solar marsh]

But yes this is a general api issue question, I just happen to have an interest in server issues but please also post in api support

[umbral merlin]

OK

Thx anyways

[solar marsh]

Post in api support I’ll go there

[eternal marsh]

Used Up the $18

[prime ermine]

Hey can we run this bot in our own server?

[solar marsh]

Yes, just read the readme

[tribal elm]

lol okies xD ty

[fluid dove]

How can I remove the prompt

[finite parrot]

config.yaml to modify it, for removing it completly you need to modify completion.py

[kind mulch]

im stuck on step 7 install dependencies... the code isnt working for me... anyone have a solution?

[eternal marsh]

what OS? any error messages suggesting which dependency is failing to install?

[fossil sleet]

Hey

can anyone tell how to add chat gpt bot

[near kindle]

Check out the repo at https://github.com/openai/gpt-discord-bot, you need to install it 🙂

[sharp raptor]

/chat hello, i'm Tien

[solar marsh]

theres no bot here...

i checked and I did explicitly say that in my post message

[carmine lance]

does anyone know how to solve it?

[finite parrot]

Try python instead of NodeJS. The Bot isn't written in Javascript.

Check the Setup Part in the Readme

[carmine lance]

and how does it run in python?

[near kindle]

? Check out the README for instructions please.

[fluid dove]

This bot generates only short answers how can I increase it?

[finite parrot]

Modify the config.yaml with own Example Conversations and modify the Instructions.

[fluid dove]

Ok

What should I do to get codes as responses

[solar marsh]

its short because max_tokens is low

if you want to use it like chatgpt change the preprompt to be AI assistant

[fluid dove]

Okay

[fluid dove]

Is 2048 low?

[fluid dove]

Like prompt=AI assistant?

[finite parrot]

Modify this file: https://github.com/openai/gpt-discord-bot/blob/main/src/config.yaml

[fluid dove]

Alright thx

[finite parrot]

Change Instructions to something that goes in this direction:

You're a virtual assistant designed to help with coding tasks. You have knowledge of programming languages, software development frameworks, and best practices in software engineering. You're capable of writing code, debugging, and optimizing performance. You communicate in a professional manner, using proper grammar and technical terminology. You have a passion for technology and enjoy collaborating with developers to create high-quality software. Use markdown syntax for readability.

The default 512 max_tokens should be fine, Discord has max 2000 Characters per message.

[fluid dove]

This doesn't seems to be working

I'm always getting responses like this

[ebon plover]

Try asking for a code example in a
specific language

[fluid dove]

How

[ebon plover]

For example "show me a code example in Java for Arduino to detect a Bluetooth connection to a mobile"

[fluid dove]

Okay

Tnx it's working

[ebon plover]

:)

[fluid dove]

How can I get response codes in code block

[ebon plover]

You ask it to use triple backticks like discord markdown

And use language suffix

Give it an example like

[fluid dove]

Okay

[fluid dove]

How can I increase the typing status

[ebon plover]

What do you mean by increase?

[fluid dove]

I can't increase the typing status?

Increase the time

[blazing gust]

Pushed a PR, check it out.

[ebon plover]

You can repeat the request every 7-8 second

[fluid dove]

Ok

[fluid dove]

Does anyone have dalle discord bot

[solar marsh]

Yes, check the openai GitHub for Dallify

[fluid dove]

Is there any project post

[solar marsh]

https://discord.com/channels/974519864045756446/1039968564699992106

[fluid dove]

Thx

[fluid dove]

How do I add multiple servers

[tribal elm]

heya, is there a way to remove the context limit so a thread can go on for much longer?

[finite parrot]

The Context Limit is the limit of the Model itself which is 4096 Tokens.

You can modify the completion.py File to have only the last n Messages in the Context, which allows for a bit longer Threads.

But it will lose the context of the first messages.

[safe turtle]

Ples Help me,
I can't run this project 😦
It's notice: KeyError: 'DISCORD_BOT_TOKEN' 😦

[finite parrot]

Looks like an issue with your .env File not being loaded at runtime and those Variables not being used.

[fluid dove]

Check your bot token or rename info.env to .env

[safe turtle]

Thanks you very much,
I had change file name, and It had run. I had add bot to my server.
But I check in my server discord, Bot notice not respond 😦 Please help me more.
Thanks

[finite parrot]

Problem is you running main.py directly from the folder src. Which makes it impossible to import src.otherfiles, try to launch it like the Readme suggests:

python -m src.main

You can do this via the terminal in your IDE i think.

[fluid dove]

Is it possible to upgrade into chatGPT api

[latent frost]

how to make an ai image

[fluid dove]

Use Dall e2

[eternal marsh]

Why is my GPT chat bot on discord not working properly and giving irrelevant responses?

[fluid dove]

What are the responses

[blazing gust]

What are your prompts?

[eternal marsh]

Bot did not respond

[eternal marsh]

I use these prompts

https://github.com/openai/gpt-discord-bot

It worked but the bot didn't respond to anything

[fluid dove]

Make sure bot have permissions

[eternal marsh]

Yes, I did that, but it still didn't respond

[fluid dove]

Did you get any response on console?

[eternal marsh]

Ahh it's working now

I used the /chat command and it worked. Before I just directly chat my bad

[fluid dove]

GG

[eternal marsh]

Thank you buddy!

[fluid dove]

You're welcome

[plain igloo]

Will this be updated to the chatgpt API?

[fluid dove]

Check pr on GitHub

[plain igloo]

not sure what you mean sorry!

[fluid dove]

Pull request

On GitHub

[plain igloo]

the one for model flexibility?

Will that work?

[fluid dove]

Working for me

[fluid dove]

No the one "Use chat model"

[plain igloo]

Thanks

[fluid dove]

yw

[blazing gust]

@fluid dove I made that PR, You have tested the model flexibility pr? And it works?

[fluid dove]

Yes

Good job

[blazing gust]

It wouldnt work for turbo though, right? As that requires a little more adjustments on the completition side of things, right?

Leave feed back on the PR so it can get merged.

[finite parrot]

For PRs: only bug fix PRs wil be accepted. If you are implementing a new feature, please fork this repo.

[blazing gust]

I read that, but cant imagine why they are doing that? Whats the point in open source, if not to improve the software.

[finite parrot]

Looks to me like the Chat PR is fine on the first glance, here is how i changed it:

            convo=Conversation(messages + [Message(MY_BOT_NAME)]),
        )
        # Start the butchering again
        # Create a Custom ChatML based ChatLog instead of using the Rendered Prompt
        chat_log = []
        chat_log.append({"role": "user", "content":BOT_INSTRUCTIONS})
        example_messages = MY_BOT_EXAMPLE_CONVOS[0].messages
        for message in example_messages:
            if message.user == MY_BOT_NAME:
                chat_log.append({"role": "assistant", "content": message.text})
            else:
                chat_log.append({"role": "user", "content": message.text})
        for message in messages:
            if message.user == MY_BOT_NAME:
                chat_log.append({"role": "assistant", "name": MY_BOT_NAME, "content": message.text})
            else:
                chat_log.append({"role": "user", "content": message.text})
        print(chat_log)
        # Use ChatGPT API to Answer Chatmessages
        MODEL = "gpt-3.5-turbo"
        response = openai.ChatCompletion.create(
            model=MODEL,
            messages=chat_log,
            temperature=0,
        )
        reply = response['choices'][0]['message']['content']
        # End of butchering
        rendered = prompt.render()

[finite parrot]

Its more a techdemo vs a software that gets maintained.

If you blow it up in features it will be hard to use it as educational foundation to learn.

[blazing gust]

I can understand that, but chosing a different model isnt exactly earth shattering feature overload.

[finite parrot]

But it would normally require a full refactor really, the hack i did is the same hack that happened in the PR.

Just that the PR even ignores the BOT_INSTRUCTIONS and EXAMPLES

[blazing gust]

turbo would require a refactor, for sure.

Damnit, guess Ill do it at some point.

[finite parrot]

The other issue is its not really gonna be Lenard anymore what showcased in #gpt-discord-bot

At least for now, maybe later personas with turbo gonna be easier.

[blazing gust]

I have some other things going on ATM, but will revisit this.

[solar marsh]

because its a code sample not a project

i made it for fun

the point is you fork it to add on to it, if you want to add more features. base sample should be simple

we can make a chatgpt discord bot sample though

maybe if i get time

[supple torrent]

how to make Open AI create something for me?

[dim epoch]

i have the bot running , is there a way to prompt for img generation like /chat img: prompt

[olive mist]

@fluid dove

Image generation is not available in this server.
To start generating images, click here.

[upbeat kayak]

hi

[humble prawn]

GPT -4 is open or subscription-based?

[ebon plover]

Subscription

[foggy wing]

hi

[lilac tulip]

Thats really awesome, did you ever think about adding voice to it?

[tired vine]

It looks like it require more permission than the followings?
I set these on and get a permission error, not the case if I set it to Admin.

Send Messages Send Messages in Threads Create Public Threads Manage Messages Manage Threads Read Message History Use Application Commands

Ok found it, it's the embed link permission

[tall lagoon]

I've tried (I think) everything to get this to run - I've Manage Server permissions myself, given the correct Guild IDs to ALLOWED_SERVER_IDS, even given the bot "8" permission (Admin) when adding to server. Bot appears to be added to server correctly, but then triggers the should_block function suggesting the guild.id is not in ALLOWED_SERVER_IDS. I've tried kicking the bot, reworking .env repeatedly and can't sort it. Any tips please?

[tall lagoon]

OK, seems to be working once I installed PyNacl...

[ashen warren]

How do i make the bot gpt-3.5-turbo?

[finite parrot]

I posted a change for this here earlier: #1055336272543092757 message

But in general this is not the right place to discuss it really.

[ashen warren]

Sorry

[ashen warren]

Thanks, but I figured it out already!

[finite parrot]

[drifting dawn]

Why my code report such error? Anyone can help?

[fluid dove]

Check your bot token

[versed hill]

helpp.. no link show in console

[fluid dove]

Check if your server id is correct

[versed hill]

yes it is, idont know why its error

[fluid dove]

Are you using replit?

[versed hill]

yess

[fluid dove]

Where did you put your server id?

In secrets?

[versed hill]

yess

all in .env.example i put in secrets

[fluid dove]

You don't need to put in .env

Go to secret and add key DISCORD_CLIENT_ID

And add your client id in value

[versed hill]

I already did.

[fluid dove]

Did you do
pip install -r requirements.txt python -m src.main

[versed hill]

In shell right? yes

[versed hill]

this is the message in python -m src.main

[near kindle]

Your server ID is not "1"

Put in a valid server ID and it'll work

right click the server and click Copy ID

[versed hill]

yeahh i put my real id in secrets

[brisk oriole]

after using "python -m src.main" it says "ModuleNotFoundError: No module named 'dacite'"

any fix for this?

I have Dacite installed.

[tiny canyon]

• gives warning: the file "AutoGpt.json" does not exist .Local memory would not be saved

Requirements.txt gives this after pip

[tiny canyon]

Did your problem solved?

[tiny canyon]

Help

[clear arch]

ccccccvklekblketlhrbhihckhtfukhvrtkbnbrubvdv

[versed hill]

Not yet

[tiny canyon]

do u know anyone who can help

[tiny canyon]

i wasted whole day for my issue not solved yet

[versed hill]

Me too 😂.

[tiny canyon]

[fluid dove]

Whats your issue

[tiny canyon]

Solved

[fluid dove]

Okay

[idle lagoon]

/go

[drifting dawn]

How did this happen? Anybody know how to solve the problem?

It worked but the bot didn't respond to anything

[fluid dove]

Does the bot have enough permissions?

[drifting dawn]

I granted it administrator privileges

[fluid dove]

Ok

[drifting dawn]

Even after granting administrator privileges and agreeing to all gateway intents, my bot still does not respond. Does anyone know how to solve this issue?

[ebon plover]

you need to allow your guild @drifting dawn , scroll up and look for ALLOWED_SERVER_IDS

[drifting dawn]

Thanks a lot, it did work!

One more question, do I need an OpenAI Plus account to gain API usage in order to use this bot?

[clever grove]

I really need a api... In Macao Theres No Api Access

[fluid dove]

?

[forest sierra]

Hi I really appreciate this sample code, and I got it working, so THANK YOU! My question is around the end of life announcement for text-davinci-003 on 2024-01-04. Will the code in the repo be updated or will it be obsolete? Would love to see a gpt-4 bot on discord sample code. I saw the messages above regarding gpt-3.5-turbo, and would try that if nothing else is coming out. Thanks again!

[subtle osprey]

can someone help me with this error please

[fluid dove]

What error?

[narrow moon]

it works in session memories files, so you need to make sure those are saved somewhere

[devout lily]

Is it works with chatGPT4-Turbo?

[heady anchor]

Will feeding in inappropriate content for antiraid / spam protection with this bot get your account disabled for submitting inappropriate things through the API?

I feel like there needs to be some type of exception given based upon use case. I would be careful and work with support before using this in production. As you probably need to be granted some type of exception to use it on a public discord where inappropriate messages will likely be submitted.

Be careful using this on a public discord server. Without an additional chat filter API that prevents inappropriate things making it to the API you risk your account without permission.

An account ban is a lot easier to appeal if you use a third party LLM to filter prompts before they get to the API. As far to my knowledge it's a best effort clause that makes that appealable.

I don't fancy getting banned without this transaction layer in place so won't be using for now.

I suggest adding langchain to do this easily with an additional LLM as the spam prevention filter. This adds delay but will lower your risk.

I am happy to offer a commit that does this, and submit a PR.

[heady anchor]

Update: I wrote this on the road and totally didn't find this....but now I have!

https://platform.openai.com/docs/guides/moderation

This is that exact thing. message contexts should be classified with the moderation endpoint before being submitted to the API. Though the moderation endpoint model has a token limit of 2,000 characters, the documentation recommends text splitting and summerization to receive back accurate results for spam/harmful message detection. Your discord bot should include moderation endpoint classification because its made available to a public discord server (in this case) It should block the user from being able to recieve any response if the prompt they send is not in accordance with the user policy. Using the moderation endpoint is a good way to lower this risk for yourself.

The specific model name is: ext-moderation-005

Example curl for testing:

curl https://api.openai.com/v1/moderations \
  -X POST \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -d '{"input": "Hello, world!"}'

[heady anchor]

Here's the relevant section of the code.

Very nice that the template includes these to protect developers.

[near kindle]

Yup! Moderations is checked both for the input and the output of the model before it is displayed, the input moderation will catch almost all bad things

and yes, if you repeatedly abuse the API with malicious input / disallowed input there is a risk of your account being deactivated

Side update-- This bot has been updated to use gpt-3.5-turbo (and other chat models) and the Chat Completions endpoint

[near kindle]

Another update -- bot has been updated to allow for changing the model, temperature, max tokens when starting a chat with /chat

[buoyant sinew]

ayo, i couldnt find any solution, so anyone can help me with my bot? i did try follow the instructions, and to my knowledge i did fine, however as u can see there is KeyError: 'DEFAULT_MODEL'...

[near kindle]

In your .env file define a new variable called DEFAULT_MODEL

like
DEFAULT_MODEL=gpt-3.5-turbo

[buoyant sinew]

its by default gpt 3.5

i didnt change anything there

[buoyant sinew]

im blind ngl

[buoyant sinew]

but now new problem guys

gpt 4 on playground site didnt help me

[near kindle]

The bot name can't be that

[raw gull]

Getting this same error as Gustavo

[raw gull]

Utterly unrelated to the name of the bot

I have found 3 errors this afternoon since updating my fork to the latest release. I will be submitting a PR

[raw gull]

Env example in the repo is currently wrong

They should just feed the entirety of the bot into gpt4 turbo like I did 😉

[near kindle]

no, you

that's wrong, its related to the name of the bot

[raw gull]

Read the message

[near kindle]

the error code is exactly due to the name of the bot

the name of the message

it says messages.2.name does not match the regex

because there are special characters in the name for the bot

[raw gull]

It's not related to the name of the bot, its related to the name of the user

[near kindle]

(which you cannot have)

oops you're 100% right

[raw gull]

Yes I know

[near kindle]

I'll flatten the name and push an update

[raw gull]

😉

[near kindle]

what's wrong with the .env?

[raw gull]

The var name is wrong compared to constants.py iirc

Eating dinner rn

DEFAULT_MODEL vs MODEL I think

And prompt render_json produces funny output

[near kindle]

Could you open this as an issue on the gh?

[raw gull]

Sure I'd be happy to 😃

Sorry for being silly a minute ago. Didn't think I'd get an immediate reply

I'll bump the channel here after I open an issue

[near kindle]

Sounds good! I have a change out for fixing the model .env default, will look into the prompt render_json after you put the details on an issue

[near kindle]

Nah that's my bad I totally misunderstood the error

[raw gull]

There appears to be an issue for the constants.py / .env.example mismatch: https://github.com/openai/gpt-discord-bot/issues/73

User roflcakesII reports that it doesn't work for them, but it worked for me

I created a PR from my fork https://github.com/openai/gpt-discord-bot/pull/75

[near kindle]

yeah dw I already put in a change for the env issue moxi will merge it in later

[near kindle]

user may have not saved the file or something, can confirm it also works locally for me

I'll follow up with that user

[raw gull]

where is the regex call even occurring? this is a weird issue @near kindle it's like it's happening inside the endpoint

i have a user with two spaces in their new-style discord name

but at any rate base.py has some strangeness in how it renders the Prompt i think

[near kindle]

It happens upstream within the endpoint yeah

the regex stuff for the name field

The name has to be flattened to unicode and special characters stripped before sent to the API

so in that user's example

we'd keep their egnlish characters

but remove all the special weird stuff, spaces, etc

[raw gull]

i thought it was a gpt-4-turbo issue at first but going back to the older version of the bot worked

[near kindle]

The older version of the bot was using text-davinci-003 as the endpoint for completions

and how that worked was, the conversation didn't have a messages array

it was 1 big string

so we were able to use special names and etc with special characters

[raw gull]

oh. i had written my own chat completions code

it was a wild weekend in may lol

[near kindle]

lol

[raw gull]

and now we want the bot in a server i'm in

it was working great on the old version with my changes, except that it's not using the async open ai client and i was getting heartbeat interrupted errors

i am going to mess with it this weekend

[near kindle]

yeah async client is definitely important for heartbeat interrupted errors

I'm planning on a revamp to this bot in the next week, as well as an upgrade for the dallify bot to dalle-3

so be on the lookout for that, I'll fix a lot of the issues with names, stability, etc

[raw gull]

lol i started to do that in mine too before i realized there was an upgrade

So as I am looking through my old code and the new code I do recall that I had issues with the Prompt render methods. i know it's friday night and i think this is your job if i am judging the openai repo correctly so I will leave you alone for now

i will be happy to share any changes i make this weekend

🙂

[near kindle]

Can you make a github issue about the prompt render methods?

I'm not really sure what the issue with those are

[raw gull]

https://tenor.com/view/futurama-good-news-everyone-professor-farnsworth-great-news-gif-17213270

So modifying base.py line 81 to "name": message.user.replace(" ", "-"), works but is not imo ideal

once i was full caffeinated and compared my chat completions code to yours i realized yours was not the issue

🙂

[raw gull]

The ultimate issue is that the name field in the chat completions ai message object has that regex that doesn't support spaces. I can replicate this issue by changing my server name to anything with a space in it and trying to chat with the bot. The issue was a little muddled by the fact that my friend who's been helping me test had not updated his account to the new username, and so discord.py doesn't provide all of the info that I expected

I don't know how i would go about creating an issue for the chat completions endpoint

Open to advice there

[near kindle]

This isn't an issue with the chat completions endpoint

This is expected behaviour, the name field can't have special characters and spaces

But it's an issue for the discord bot since we don't remove special characters in a username when we send their username to the endpoint, so we need toe dit the bot to simplify the user's username string to only letters and numbers, etc, so if you open an issue on the github for the discord bot we can track that and I'll get that in

@raw gull does that make sense?

[raw gull]

Yes, I understand. However, I do not see this constraint in any of the documentation for the chat completions endpoint.

The completions API will refer to a user by the string in the "name" field, which should have spaces if their user has spaces in their name, so this still feels kind of like an issue with the completions api. One of my users is "Ben Queen Mineral" and having the bot respond with "Hi BenQueenMineral" or "Hi Ben-Queen-Mineral" or whatever the 'simplification' would be looks weird

I had already worked around it by modifying Prompt render_messages to remove the spaces

[near kindle]

I understand, so you want to raise the issue with the chat completions endpoint itself, right?

There doesnt seem to be a real official way to raise an issue with the upstream API itself, just the libraries (e.g the python library) but that's not what you need since the library doesn't actually control the behaviour of the endpoint, so I would say you should create a forum post on the openai forums for this issue and start a discussion about this

[warm kiln]

hello