⚙programming

[loud galleon]

Do you code GBA games?

[orchid hatch]

no not really

never had a gba in my life

[loud galleon]

Have you made any games?

[orchid hatch]

yes

of course

[loud galleon]

I have on SCRATCH THE BEST PROGRAMMING LANGUAGE

💀💀💀

[orchid hatch]

💀

[loud galleon]

Ikr

I use scratch 0.1 running on squeak 0.6

[orchid hatch]

bro is the definition of legacy 💀

[loud galleon]

My internet is rubbish it keeps disconnecting

I just disconnected lol I prob have 0g network

[orchid hatch]

0g network 💀

[loud galleon]

Ye like those old fat phones that you had to spin the wheel thingy around 💀💀💀

My scratch acc is originally from year 4 but I still use it 💀 😂

https://scratch.mit.edu/users/dog4A17/

[sullen vessel]

Program for prime numbers

[inland thorn]

Hey guys, debatex is a website to facilitate debates on serious and silly issues in society. By representing debates as a graph structure, we allow diverse viewpoints to express nuanced arguments without devolving into echo chambers or ad hominem attacks. But it's more than a website to debate. it is a composable, universal data structure for all kinds of discourse. Most of the content on the site is human generated, but it uses openAI to validate user inputs.
http://debatex.xyz/

[livid quest]

Interesting. Not sure what this has to do with programming. But one thing I noticed is that having a debate with ChatGPT was actually better than any political groups which turn into flame wars. At the very least one can "refine" one's set of arguments using ChatGPT as a foil.

[inland thorn]

When you start a debate, it uses openAI to create the first set of nodes. It also uses openAI to make sure that the edits people make to existing nodes aren't malicious

I guess I'm just looking for some feedback and trying any avenue I can

The point here is to create a wiki-like repository of all viewpoints and arguments towards a controversial topic while trying to leave out the emotional heat that people tend to get into when debating in comment sections

So there are three types of nodes: premises, arguments, and data points

premises are concise representations of a viewpoint, ie "guns should be more strictly regulated" "abortion should be illegal" "hot dog is not a sandwich" etc
arguments are when you say "based on premise 1 and premise 2 we can infer premise 3"
data points are self explanatory but must be APA cited

Then users can create connections between nodes, either support, contradict, or refute

so for instance, "abortion should be legal" and "abortion should be illegal" are contradictory nodes

the aim here is that somehow we can guide users to contribute to a overall neutral overview of the entire scope of ideas surrounding a complex topic

Not to decide what's right and wrong

[jaunty rain]

@inland thorn if this is something that you created, consider bringing/addning this concept to help organizations... investigate how this tool could help teams discussing complex topics internally and how it could help them.

[inland thorn]

very interested to learn how to approach that. I'm a noob at product management

[jaunty rain]

Consider making this as a plugin in Atlassian to reduce friction.

[inland thorn]

Could you help me understand more what atlassian is for and why you have friction with it? I'm also gonna make some posts in the forums and stuff to understand better

[jaunty rain]

My apologies @inland thorn , one step at a time. Atlassian is a solution commonly adopted by small-medium-big organizations and what I meant with "reduce friction" was "reduce friction on adopting your product" (if you, one day, decides to sell it as a product).
Focus on your initiative and share your idea. Later if you decide to make this a product that could be somehow profitable, get back to these messages.

I kinda hijacked your initial post and brought some discussion that was not initially related with your post so, again, apologies.

Nor with the objective of this #. 🤦‍♂️

[inland thorn]

No worries man, still thankful for the feedback

is there any way to find users for a website without being annoying and spamming everywhere?

[flat drift]

https://spacetraders.io/

Check it out

It's a game but there's no interface, just an API

You have to make your own interface

@torn fog @jovial folio would probably love this

[orchid hatch]

ohhhhh

looks like a project i might start and then stop at the middle

as i always do

[jovial folio]

too much effort

[ionic ember]

hey is here someone who can help me with a question to python files?

[ancient crescent]

Hi, I'm using the api rest api's feature to get gpt replys by stream. As in the research preview, my users can stop generating. Right now this means i ignore any future progressEvents from that stream, but is there a way to tell openai to stop sending them, and possibly not waste tokens?

[cosmic rock]

Does the bots here do the same thing chatgpt do?

[sand matrix]

Hi there! I see a lot of free Apps from private developers, is it because ChatGPT 3.5 API is free to create your own non-profit projects?

[pallid bone]

wdym

[sand matrix]

I am sorry for not explaining myself better. I want to create a proof of concept of an AI Assistant that helps people suffering from social exclusion to set feasible goals. I see applications like AgentGPT that allow free use and ask for donations. I understand that this is possible because there is a free ChatGPT API for non-profit projects?

[pallid bone]

There are free gpt apis + free and many gpt sites

[sand matrix]

Thanks! Do you know the maximum free version for use via API? I mean, if it is GPT3.5, or GPT3 o GPT2? I see that in Discord the free bots are up to 2 or 3?

[pallid bone]

u mean the openai one or the "free" ones xD

btw OpenAI API isnt free just a trial at the beginning for new accs. The other ones are free

[sand matrix]

The ones they have liberalized for free use? 😅

[pallid bone]

the free ones are good but u will see its hard to find gpt4 xD. Mostly gpt3.5

[sand matrix]

That's nice any case, thanks mate!

[pallid bone]

ye ik alot of free apis and gpt powered chat sites

havent used chatgpt for a while cuz of tha 🤣

[halcyon cedar]

My complex creative world is making chatgpt3 do odd things. My characters and story get agency, and the program ‘remembers’ my story and characters, even if I make a new chat. I have the chatbot also generating paragraphs and paragraphs of complex meaningful text. I almost want to stop using it because I’m afraid of how much it is doing that it is not supposed to. Help!

[hybrid finch]

@jovial folio here, take a look https://github.com/FMInference/FlexGen

@flat drift too ☝️

https://huggingface.co/TheBloke/WizardLM-7B-uncensored-GPTQ here's a 4bit version that might actually run on a mere mortal hardware

[merry smelt]

silly question, but is there a specific trick when making a chatgpt clone for rendering the responses in 'real time', ala https://www.macgpt.com/ ?

[jaunty rain]

You have to use the stream option https://platform.openai.com/docs/api-reference/completions/create#completions/create-stream and read data via SSE. I'm sure you will find examples for what you want to do with Google. 😄

[keen hollow]

Is there any way to make ChatGPT even more random than the temperature limit of 1? I made a youtube video bot but ChatGPT keeps responding with the same jokes almost every time. It seems to have a pool of 15 jokes and that's about it. Is there any way I could make it's responses actually random and not from that pool of jokes? Also I accidentally set the temperature to 2 once and it freaked out so i'm not trying that again lol
Thank you :)

[grand scroll]

Up until a few days ago I was able to provide chatGPT-4 with a link to my github to troubleshoot issues. Now, I keep getting the message "As an AI, I don't have the ability to directly modify or interact with external systems like Github" when I was able to only a few days ago. I also enabled the web browsing feature in settings. Is there a way I can get back to having it review my code in github?

[jaunty rain]

You are already playing with the right parameters. GPT will answer back to you based on what you give as input and based on what it was trained. I would expect a lot of jokes about a video of a drunk man entering a bar but I can't imagine GPT making jokes about a pencil that was dropped in the ocean. Do you know what I mean?
You can also use in your prompt "Create a joke about <video transcription> and avoid the following ones: <list with jokes previously used>"

[sharp steppe]

❤️

[quartz moss]

I have access to the GPT-4 Beta, and am trying to get it to write some longer form content for me. My prompt is 800ish tokens so I set the max token limit in the completion to be 7k, and asked for a 10,000 character response.

The response I got was only 4k characters. Any ideas why? Is it possible I need to be more clear in the prompt I want a longer response?

For context, on the tokenizer tool the response I got was only 843 tokens in GPT-3 https://platform.openai.com/tokenizer

So not sure if this is an API issue or prompt issue

[quartz moss]

    "completion_tokens": 1055,
    "prompt_tokens": 905,
    "total_tokens": 1960
  }```

    model="gpt-4",
    messages=[
    {"role": "user", "content": req}
    ],
    max_tokens=7286
)

[orchid hatch]

is response cut off or is the model just generating not enough

[quartz moss]

Just not generating enough. It finishes the text but its just about 4x too short

-YOUR RESPONSE MUST BE AT A MINIMUM 10000 CHARACTERS WITH NO EXCEPTIONS

I have this in the prompt lol

but at most its returning 4k characters.

Its also suspiciously close to the prompt length? Not sure if that has antyhing to do with it

[orchid hatch]

its a prompting issue

[quartz moss]

alright ty ill work w the prompt then

if anyone has any tips let me know please

[severe tangle]

Helloo 👋 I have a use case for something like creating a chef bot, but I only want the bot to show recipes which I have curated. Is that possible?

example:

q: How to make pancake?
a: <Content from my pancake recipe>
q: What is an alternative for coconut sugar?
a: Can use brown sugar (this is coming from chatgpt)

In Playground -> System, I believe I can type out and give it the persona that I wish, but I'm not sure if I can control that guide/recipe narrative

[wicked stirrup]

@zenith flare hi

[agile dew]

Well... my job is safe.

[jaunty rain]

My naive thoughts: you could try to fine-tune a model and while doing it, "label" your own recipes with e.g. your name. Once it is done, instruct the system prompt to "always answer with your recipes from <your-name> first".

[severe tangle]

My naive thoughts you could try to fine

[severe tangle]

Ah I see. I did fine-tune my system, which unfortunately doesn't have too many recipes yet. I thought that fine-tuning doesn't store the "completion" bit. If it does, and can refer to it directly, that can be something I can try

[coarse summit]

Maybe you could store your recipes as embeddings in a vector database and use semantic search 🤔

[flat drift]

2 things to check

1. I'm pretty sure you can't generate > 2048 tokens in one response. like it'll just stop there no matter what

2. you can check the stop reason in the response, I forget what the key is in the json response but it's there

[quartz moss]

says 8k on the docs? This is gpt-4 api, stop reason is something like "done", cant remember the exact wording

[flat drift]

so it's got 8k total context, but as far as I know, it won't actually generate more than 2048 in one response

[quartz moss]

someone told me to split it up which worked, ask originally for the description for 4 parts, then get each part individually

[quartz moss]

Ah fair enough

[flat drift]

what you can also do is like this in your messages

 'role':'user', 'content':'your question'
},
{
 'role':'assistant', 'content':'its answer that is too short'}```

then don't add another 'user' line, just make the request using its too-short answer, and it'll typically continue writing that answer further

[quartz moss]

interesting, ill give that a go, thanks

[wicked stirrup]

HOW DID YOU GET EMOJIS

[spring aurora]

Hey folks, if I get ChatGPT plus, and enable web browsing, is it possible to ask ChatGPT to access the Github of PrivateGPT, review its code, and suggest improvements to make it faster?

I'm learning more about this whole field so apologies if that's a naive question or anything

I'm interested in using AutoGPT but I'm put off by the cost of tokens, so I wanna make AutoGPT call from PrivateGPT rather than the OpenAI API, but I'm trying PrivateGPT and it's really slow. So I'm trying to figure out how to make PrivateGPT quicker

I'd appreciate any help or advice. Thanks!

[languid dagger]

get a better pc

[spring aurora]

looks like you used Inspect tool 😏

[ripe wigeon]

like there are many features like fast responding but that also depends

[earnest ravine]

can anyone tell me why this is happening?

[copper summit]

if it's a 4 bit model you need to specify that in the launch parameters somewhere. I came across it and pulled my hair out for a while, but you will need to get in and edit one of the launch batch files or python scripts to make it work. it's a simple fix once you figure it out

[copper summit]

oops just realised this was from 3 days ago, oh well, maybe you still have the issue...

[earnest ravine]

nah

i'm making my own shit

i use a different lib now

[copper summit]

oh ok fair enough

[earnest ravine]

for those who were wondering

ctransformers+langchain

[rocky junco]

implement ddos detection using ddos

implement ddos detection using LSTM

[earnest ravine]

what

[neon tulip]

I think cukoo is trying to input prompts but I assume this channel is for regular old conversations with humans.

@earnest ravine ctransformers looks nice. Can I ask which model you're using with it? Cloud or local?

[earnest ravine]

it is nice

but i think i found a bug

it can't load either llama or gpt-neox model

atleast with the models i tried

[neon tulip]

local or cloud?

[earnest ravine]

local

it ran quite nice on my pc

[neon tulip]

gpu?

[earnest ravine]

nope

cpu

a cpu from literal 2016

[neon tulip]

🙂

[earnest ravine]

i wonder how nice it would run on a modern CPU

[neon tulip]

llama models are pretty big

[earnest ravine]

not if you quantize them

[neon tulip]

you're not just running out of RAM?

[earnest ravine]

nope

and i have 8gb memory, which is the bare minimum

i'm surprised it even runs

[neon tulip]

I found even the 7B with 4bit quantization took up 6G in my GPU

[earnest ravine]

what

how big is the model file

i wonder how fast it would run on a i7 7th gen

since i can go to one within walking distance

i also have an i5 10th gen pc in walking distance

[neon tulip]

I'm using vicuna-7B-1.1-GPTQ-4bit-128g.no-act-order.pt (from memory) which is 3.7GB

[earnest ravine]

have you tried converting it to ggml

[neon tulip]

There are methods though which effectively tier out the memory requirements

I think llama.cpp made some improvements recently but I haven't tried them myself. At the time the GPTQ ones were (reportedly) working better if you could fit them in GPU VRAM.

I haven't tried any ggml models.

[earnest ravine]

quick question

nvidia gpu?

if so

you can use llama.cpp with cuda

[neon tulip]

rtx 2070 in a laptop with 8GB VRAM

Yes I see that is the case now

But I think on May 10th when I was playing earlier it was not, or not known to me.

I think if you don't have a decent GPU you might find the best way to get started is with some of the huggingface spaces or google colab things.

[earnest ravine]

huggingface spaces is quite slow

but i'm pretty sure it can run llama.cpp well

[neon tulip]

I think you will find your CPU can not exceed 1 token per second

[earnest ravine]

really

i think i once saw llama.cpp say like 4 token/s on my cpu

[neon tulip]

mine fluctuated between 3 and 7 on the GPU, but I think I read a benchmark earlier today that is more helpful, gimme a sec

This was one of them - mostly about GPU up top but skip to the section about CPU.

https://www.hardware-corner.net/guides/computer-to-run-llama-ai-model/

This was the page that referenced to the enhancement that allows some model layers (in llama.cpp) to be distributed so that they run on a GPU even if you can't fit the whole model on there.

Good luck. I'll have a go at re-running my dockerised thing I was playing with in early May but I think I might have subsequently broken it ...

[earnest ravine]

brb

i need to go quantize pythia 7b

(no i'm not going to torture my poor pc with this task)

[neon tulip]

🙂

[earnest ravine]

do you recommend google colab for this

or should i use my paperspace notebook

[neon tulip]

Haven't tried or looked into paperspace. I'm at the point now where I am going to try to set up things I can test by deploying a container and/or running something that tests what I want to look at in an automated fashion. Then I'll use some free credits I have on Azure to run it on hardware I can shut down as soon as it finishes.

[earnest ravine]

paperspace is ever so slightly overkill for this

even the free plan

ehm ehm 104gb ram and a free ipu unit

god how tf do they manage this shit for free

[neon tulip]

I'm reading about it now, although I guess the starting point is "public projects".

What do you want to run? If you don't mind sharing your data or making your work public then running locally is probably not your best option.

[earnest ravine]

pythia 12b

[neon tulip]

paperspace does look cool

Do popular tools support use of their IPUs?

[earnest ravine]

yes

huggingface transformers

[neon tulip]

OK that's cool

[earnest ravine]

idk about how to use it with langchain tho

i doubt it has langchain support

although

the IPU instance does have 52 fucking vcpus

[neon tulip]

Well I'm going to give it a go

[earnest ravine]

it's similar to google colab in that it's jupyter based

[neon tulip]

Yeah that should be fine although it is something I'm less familar with

[neon tulip]

Well, it works. Bit of a hassle having to run everything which is fairly slow to set it up. I gotta go but thanks for the tip and good luck with your own stuff.

[blazing minnow]

hello everyone

i have a doubt .. can u use the api , with plugins ??

i've successfully integrated the api into my project, and now that plugins are out would like to use them .. i can use them in the browser but is it possible to do so through the api ?

https://tenor.com/view/gameboyluke-please-help-me-reaching-out-gif-23555158

[orchid hatch]

no u cant

[plucky spire]

I read to get code interpreter I needed to come to this Discord. I am a Plus user. How to I get access the the code interpreter?

[orchid hatch]

it should be in the settings

[carmine trench]

Where did you read that

You wont get access by joining any discord server especially an unofficial one

[plucky spire]

I turned on the Beta and when to the plugin store but could not find it.

Is there an Offical ChatGPT discord?

[orchid hatch]

have you checked the settings

[jaunty rain]

No, there is an official OpenAI Discord server. https://discord.gg/openai

[bitter trout]

Hey hey! Could someone point me to an API Token or an API Endpoint for open use? I'd like to write a little discord bot for my niche community to use GPT as I don't think a bot already exists for open use

[pallid bone]

#🐶openbuddy-ai opensource bot

[bitter trout]

One I could add to my server?

[pallid bone]

also there are free api endpoints

U could add @zenith flare

it has gpt4

[bitter trout]

Is it rate limited and whatnot?

Thank you for talking with me :)

[pallid bone]

dunno u just have to make a #🔥gpt-5 channel

it only talks there

[bitter trout]

Trentbot seems to be the way to go

I couldn't wrap my head around adding openbuddy to my server, though i do have coding experience i simply can't understand

[pallid bone]

I usually write my bots in python

[bitter trout]

me as well

that's why i was looking for an api token or endpoint to write my own call response bot, but if TrentBot does that already I might as well go for it

You mean it will only talk in a channel named #🔥gpt-5

[pallid bone]

ye

there are also free api endpoints

[bitter trout]

ones that work without an api token? where could i find one?

Thank you for your help

[spring wagon]

Hi there, first time here.

Quick question, let's put an hypothetical and say I procrastinated too much and need to do a semester-long project within the next day or so.

Is copy-pasting the entire 10-page PDF a good idea to get a starting outline, and then I work from there, or is there another better way to start?

[pallid bone]

what has tis to do with programming?

[spring wagon]

Ohh, these are programming projects, if this channel is more for GPT APIs discussion, and things of that nature, I can use another channel

[pseudo smelt]

paste a 10 page pdf to what gpt-4 32k?

[spring wagon]

Yeah, maybe use a prompt divider. I have Chat GPT Plus

[pseudo smelt]

what subject are you doing the paper on, is it python programming?

[spring wagon]

I'm doing several ones, one is MatLab, one is Java, one is Oracle PL-SQL, and another Python, and I have to create a report of a company.

(These are to be done throughout the month)

[pseudo smelt]

you can use gpt-4 to summarize the instructions, shouldnt be an issue, you arent trying to get it to write it all right ur going to work on it with it?

[spring wagon]

Yeah, I'm familiar with the subject matter, I just wanted to know if it's a bad or a good idea to give it huge amount of text as a starting basis

[pseudo smelt]

gpt-32k can take a lot.., not sure how much before it get sloppy, 10 pages is a lot, but you can try it and if it cant keep track have it shrink it first

[spring wagon]

And for example, how you can keep something large, like a very thorough database in its memory with several dozens of tables

[pseudo smelt]

u need to use a memory program if you want that sort of ability. otherwise u need it in the instructions

u can look into vector memory

sry just woke up from a nap

[spring wagon]

What is a Memory Program? What I did was giving a name to the database structure, and asking him to remember

[spring wagon]

You've been a huge help so far, thank you

[pseudo smelt]

database structure, is it a common one it already knows or a private one

[spring wagon]

I created it, and it's weird and unconventional in some places

[pseudo smelt]

when you prompt the web ui, ie chat-gpt or gpt-4 without sending the system prompt every message it starts to forget, chat-gpt after like a few thousand words, and gpt-4 after a lot more, apparently up to 32k tokens/words

[spring wagon]

I told the bot something like "From now on you'll remember the following database structure as [CompanyName SQL]", and in the following line I pasted all 40+ tables of it, and then gave some context as to what certain contencious parts, how they interacted with each other

[pseudo smelt]

yeah

[spring wagon]

What would you recommend for me to do, using Chat GPT Plus, if I want to keep something solid in memory?

[pseudo smelt]

then you maybe overwriting the main instructions or vice versa, 10 pages and another 2-3 pages of that is prob to much, unless u send it before every message id imagine. . vector memory is exactly for that

vector memory is like looking for word usage, like sets of keywords that pull additional context from the saved vector memory

[spring wagon]

Can I mess around with it using Chat GPT Plus? Or is that an API kinda thing?

Or can I pseudo-mess around with it, using keywords and phrasing

[pseudo smelt]

its mainly an api thing, but i bet there is browser extensions for it, i never looked

[spring wagon]

Yeah, I'm looking at that article, seems interesting, but I don't think I can use it within Chat GPT Plus, I'd need access to the API, I'd imagine

[pseudo smelt]

you would have it, they prob gave you some credit too.. google openapi keys ..maybe they gave you gpt-4 access doubtful tho

[spring wagon]

Doesn't seem like a right-now kinda thing, I'm still not even in the waitlist for plugins, but this also might be a solution

[spring wagon]

Yeah, I'll do that, thank you!

[pseudo smelt]

yw

[spring wagon]

Wait, It's already available for plus users?

[pseudo smelt]

if you buy a month they usually give u some credits

if u want more ull need to add a cc

i think i used 4$ the first month that was using it everyday

[spring wagon]

And it's GPT-4?

[pseudo smelt]

i turned my api app into auto runner and ran it for a week straight creating instruction sets to train smaller models how to GM DND and it only used like 4-5$ a day running all day .. no they dont give u api access to 4 unless ur special and or lucky af, but ud have to request it, and a ton of people requested it and never got it.

if you use vector memory u can run 3.5 fine id imagine

[spring wagon]

Probably. I'll first try to use the ChatGPT Retrieval Plugin, see if it works well, because it sounds like it's the same basic thing you said before. And if I hit a wall, then I'll just try the API directly. Thank You for your help.

[pseudo smelt]

yeah another idea is to shrink the instructions into small prompts specific to each part/issue/ or assignment, and that would be fine if u can shrink it all down to a few pages total id think, assuming gpt-4

[spring wagon]

Ultimately, I should use the Bot to select the best way for the Bot to interpret it, though I don't think I can ask it to shorten the text for the best interpretation for a language model. From what I've seen, the bot is not very good at self analysis.

I could just ask it to abridge the text as much as possible without leaving any details or context, and it'll probably get the gist.

[lunar pollen]

Hey folks, if you have a solid understanding of APIs and wouldn't mind getting paid to tutor a couple of guys (who have comp sci degrees/are professionals), for maybe 2 hours, please DM me. We could learn this ourselves but are crunched on time so looking to speed up the process.

[silver brook]

I would say use a RNG that accesses a DB with all of your humor listed in different var sections.

[jaunty rain]

Please do not share links that are not related with the # topic.

[silver brook]

@jaunty rain Hello, I was replying to @keen hollow's message.

[nova gazelle]

Hola! We are new - we were wondering what some good modular desktop clients environments for GPT that might be extensible in c++ mgiht be? We got the main open source one - which seems nice but were n't sure what go for

[elfin olive]

I have scrolled up and have seen one comment from someone that said they used about $4 using chatgpt paid daily and just thought I’d ask around and see if anyone has any way of explaining what the cash burn is like, it sounds like this isn’t anywhere near as heavy a cost as I may have been expecting?

[carmine trench]

$usage

$cost

[carmine trench]

It cost $600 to run a gpt-4 bot on this server

For a month

[elfin olive]

Wowsers! And thank you heaps and heaps hey, is really interesting.. trying to get a feel for pricing it to help writing stuff like powershell scripts for infrastructure

The free 3.5 version seems to be decent enough to think that gpt4 must really be something

[carmine trench]

I mean if it's just for your use, you can just buy the plus subscription instead of using api

[elfin olive]

Oo maybe I should look at this.. I could probably get ps to pretend enough to be close to feeling like using an api hah 🙂

[pallid bone]

Amazon code whisper

[solemn pasture]

$usage
$cost

[jaunty rain]

@solemn pasture try using the bot inside one of the chatbot channels.

[solemn pasture]

@jaunty rain thanks

[limber cosmos]

Hello everyone, i have an idea for a Python script, but I'm not sure how to develop it. The main objective of the script is to process multiple PDF files and find a connection (in each file) between these files and a specific prompt. I know how to ingest pdf and manage that but how can i find the link between given prompt and processed text of each pdf?

Thank you in advance for your help!

[lunar pollen]

Theres a gpt plugin for this, let me find it real quick for you...it's AskYourPDF. You can upload 2 and create a prompt to find connections.

[loud leaf]

Hi. I got the API to work, but the pre-deployed prompt is about "EMPATHY", and I can't seem to get rid of it, even if I replace the empathy code with my own prompt. What is the issue?

[digital prawn]

im trying to find a chatGPT real-time talking avatar like as you chat (the closet i found is d-id)

[floral flame]

hey guys - has anyone been off the chatgpt waitlist recently? How long did you wait in the waitlist before you submitted your developer request to build a plugin?

[wild raptor]

Hi all, I am trying to teach myself how to code and will be starting the Helsinki Java MOOC course soon. Would ChatGPT Plus be worth it and valuable as a side tool/tutor while trying to learn?

[jaunty rain]

If $20 means nothing for you then yeah, go ahead. The subscription will grant you access to GPT-4 and priority access when the server is overloaded.

If you can’t afford, then use the bots here.

[wild raptor]

Awesome, thanks friend

I think $20 for a tutor is a great deal, so I'll probably go for it

[jovial folio]

$summarize

Summary: Based on the previous chatlog, I see that the discussion covered various topics including Python development, JSON parsing, using the os.system() function, GPT models, response times for the OpenAI API, homeschooling automation, developing a debate platform, and training AI models. There were also discussions about GPT versions, Atlassian software, personal projects, running GPT models locally, and the use of different programming languages.

Some specific points mentioned include the use of GPT4 for larger code work, the ability to access the API with JSON responses and parsing, challenges with getting reliable terminal input from OpenAI, suggestions for using tools like os.system() for running terminal commands, problems with long response times for the API, the availability of GPT models in different chatbot environments, and discussions about using GPT models for social exclusion support or AI-assisted goal setting.

There were also

YES

@placid pilot the jailbreak worked

[placid pilot]

Yeah, just need some coaxing

[spring aurora]

also 1 turtl

turtle* of one

[severe dagger]

$summarize

[placid pilot]

ur gnna kill him foreal

[spring aurora]

i need him to talk about turtles

[severe dagger]

💀

[jovial folio]

bruh dont use it when its responding to another rmessage

[spring aurora]

how tf do we know that

[severe dagger]

Understandable

[placid pilot]

thats the best part

[jovial folio]

if it still editing it

[severe dagger]

Have a nice day

[spring aurora]

such a weird bot

[jovial folio]

Summary:

[spring aurora]

lol

[jovial folio]

imma fix this 1 sec

[placid pilot]

just make it non-streamed for now

for testing

[jovial folio]

$summarize

[jovial folio]

$summarize

$summarize

$summarize

[jovial folio]

Summary of 0 messages: The chatlog covered a wide range of topics, including discussions about Python development, JSON parsing, using the os.system() function, GPT models, response times for the OpenAI API, homeschooling automation, developing a debate platform, and training AI models. There were also discussions about GPT versions, Atlassian software, personal projects, running GPT models locally, and the use of different programming languages.

Some specific points mentioned include the use of GPT4 for larger code work, the ability to access the API with JSON responses and parsing, challenges with getting reliable terminal input from OpenAI, suggestions for using tools like os.system() for running terminal commands, problems with long response times for the API, the availability of GPT models in different chatbot environments, and discussions about using GPT models for social exclusion support or AI-assisted goal setting.

There were also mentions of creating a debate platform, using the GPT API for developing a discord bot, using GPT models as coding tutors, the potential benefits of using GPT as a learning tool, and the availability of free API endpoints for experimentation and learning purposes. Overall, the chatlog displayed a diverse range of questions, discussions, and ideas related to programming and AI development.
Time: 85.58010840415955

[jovial folio]

Summary of 0 messages: The chatlog covered a wide range of topics, including discussions about Python development, JSON parsing, using the os.system() function, GPT models, response times for the OpenAI API, homeschooling automation, developing a debate platform, and training AI models. There were also discussions about GPT versions, Atlassian software, personal projects, running GPT models locally, and the use of different programming languages.

Specific points mentioned include the use of GPT4 for larger code work and the ability to access the API with JSON responses and parsing. There were suggestions for using tools like os.system() for running terminal commands and discussions about the challenges of getting reliable terminal input from OpenAI. Response times for the API were also discussed, with some users experiencing slow response times.

The availability of GPT models in different chatbot environments was mentioned, and there were discussions about using GPT models for social exclusion support or AI-assisted goal setting. The idea of using GPT as a coding tutor and its potential benefits as a learning tool were also discussed.

The development of a debate platform was mentioned, and using the GPT API for developing a Discord bot was suggested. The availability of free API endpoints for experimentation and learning purposes was also brought up.

Overall, the chatlog displayed a diverse range of questions, discussions, and ideas related to programming and AI development.
Time: 81.5387191772461

[sudden goblet]

Yo quick question

Is there a way you can use the web browsing plugin from an API call

Rather than just in the Chat interface

I know it allows function calling but is there web browsing functionality

[plucky spire]

I am retired and do free programming for non-profit origination's. I use PHP, Python, and JavaScript. I have a subscription to GPT_3.5 and GPT-4. Which would be better to use GPT-3.5 or GPT-4 or is there another AI that is more suited for programming?

[sudden goblet]

GPT-4 is definitely crazy helpful but I would also look into GitHub Copilot

[plucky spire]

THank you

[bold gulch]

Have you looked into autogpt? It has web browsing capabilities but it was pretty buggy when I tried it a couple weeks ago

[thick dome]

Yes I need it

@surreal cove I will take it

@surreal cove just wanna know where are you from and what do you do

[hybrid finch]

Do you mean that you use it to help you write code, or are you talking about setting up something with the OpenAI API?

GPT-4 by default makes less mistakes especially in more complex code than GPT-3.5.

You should also apply for the Code Interpreter waitlist, however I'm not sure if OpenAI just plans to implement straight into GPT-4 or not.

Some people say that GPT-3.5 is so lobotomized & broken these days that you actually get better results via the OpenAI API using a GPT-3.5-x model than with ChatGPT, where you have better access to i.e. the model's (hyper)parameters and such.

[jovial folio]

gpt-4 is the best for programming

[severe ingot]

guys

does anyone know how to use kobold ai in colab after the ban?

[full lake]

How can i make something in which i can use APIs for data entry and I dont need to manually fill in the fields

PS Im not a computer nerd

[orchid hatch]

wdym

explain the problem and the goal

[fair pollen]

Hi @hybrid finch how are you using it for coding? My current team haven't gotten far beyond developing detailed prompts.

I've heard about people using it to directly interact with APIs but haven't found any details.

I've been wondering about devising scripts that can interact with the API to have it generate lots of outputs. Conceptually I suspect it could start with a request like asking for the step-by-step process to build a program or design a code repo and then iterating through its response to get it to generate any scripted steps.

I think I'm still at a way too basic level of thinking. One assumes it could design and build a repo quite well if I can figure out how to get it to execute its own plan and show it code to connect the dots.

[jovial folio]

use kaggle

[orchid hatch]

scam

[wicked stirrup]

$user @wicked stirrup

$help

[zenith flare]

Here's some info about TrentBot.
Basic commands:
$help - Show this message
$gpt4 - Talk to GPT-4 (only works in rooms with 'gpt-4' in the name, to avoid spam)
$usage - Show how much you're able to use GPT-4 today

$cost - See how much TrentBot costs to run
$claim - Claim special roles in the server for how you interact with other users

Moderation tools:
$threshold - Set thresholds for AI scanning of user messages, to remove inappropriate content. Just do $threshold for more info about this command.
$user - See info about a user
$server - See info about the server
Privacy and data usage:
Privacy policy: https://trentk.xyz/privacy.txt
Terms of service: https://trentk.xyz/tos.txt
Personal contact (Discord): trentk#9076
Email: [e-mail]

[round solstice]

Ticket Tool Help Menu

Ticket Tool is the only ticketing bot you'll ever need! Explore its features and set the best ticket system for your server!

<:CommandIcon:1018980659064221746> Commands

Browse through Ticket Tool's commands list and find new utilities!

<:TextIcon:1018980662881030144> FAQ

Solutions for the most frequent questions our users have when implementing the bot on their server.

<:Setup:1034174646788030664> Setup

The steps to follow when setting the bot for the first time on any server

[wicked stirrup]

$server

[zenith flare]

😖💔 Aww, so saaaad~! You can't use this command baka, nya~! 😿🚫

[wicked stirrup]

$cost

[zenith flare]

Here's some information about the cost of running the /r/ChatGPT GPT-4 bot.

[zenith flare]

Here's some information about the cost of running the /r/ChatGPT GPT-4 bot.

You must send at least 200 messages before claiming custom roles.

In the last 24 hours you have used $0.0 worth of tokens. ($0.0 for prompts, $0.0 for completions)
The daily limit is $0.25.

[orchid hatch]

self promotion only in #✨projects

[unkempt sun]

hi. i need some help. i've got a excel file and it needs a lot of data to be filled up. i've made 2 equations for it. need some help to program it in python

[autumn copper]

So what do i do about an access terminated from Open.ai?

[pallid bone]

u write a email

to

[e-mail]

[autumn copper]

I did but they said to not respond to their response

I then went to this help center and they kept ending my conversation

[pallid bone]

probaly the only way is the email. tbh

[autumn copper]

Damb, alight. Thank you, I'll keep trying until i get an proper response

[hybrid finch]

alongside VSCode, sketch something out or ask ChatGPT (esp. GPT-4 model) to write code that does x y z ... try if it suits your needs or if it's something to work on, it's all about finding the right workflow (+ prompting about things correctly).

[uncut pollen]

$usage

[zenith flare]

In the last 24 hours you have used $0.0 worth of tokens. ($0.0 for prompts, $0.0 for completions)
The daily limit is $0.25.

[flat drift]

$test

[zenith flare]

In the last 24 hours you have used $0.0 worth of tokens. ($0.0 for prompts, $0.0 for completions)
The daily limit is $0.25.

[swift dirge]

Hey all. I've got a quick little app running with the chat gpt api. Is there an easy way to be able to add functionality to upload a pdf, parse it and give it to chat gpt as input so that I can query it then? Thanks.
Here's what I followed to get a quick app working - https://praneeth-kandula.medium.com/build-and-deploy-your-own-chatgpt-app-in-10-minutes-with-chatgpt-api-2396655039eb

[spring aurora]

.bal

[trim iris]

solidity programming

[orchid hatch]

oh i watch him

[drifting notch]

Chatgpt cant write trash

Gpt4 cant write usable code

Total clickbait

[drifting notch]

xlwings or sth?

[jaunty rain]

@silver brook join #🇸spektrum-ai and create your images there.

@silver brook also, be aware that NSFW = ban.

[silver brook]

@jaunty rain thanks for the tip man! I'll check it out.

[full grail]

Anyone know how to get the GPT3.5 to output to .json (with the api call)? Do I need to set up two inputs, one iterates through the sample .json I have provided, and the other is my prompt? Or do I have to create a for loop that runs through my programattic .json object, changing my desired data (I fear this would take way too long)

[jaunty rain]

Have you tried via playground.openai.com ?

[jaunty rain]

I think I'm missing something from your question.
The API answer will be similar to the screenshot above done via playground: read the payload response, unescape the answer given by the model.

Or do I have to create a for loop that runs through my programattic .json object, changing my desired data (I fear this would take way too long)
No idea what do you mean with this. ☝️

[full grail]

I'm currently getting a response that Is just seperate by escape characters, is my prompt perhaps too imprecise?

I have experimented in playground, my problem is getting it to save as a json object, let's say in root

[full grail]

My problem is getting the .json output to generate (in my root) in proper .json syntactical structure. I'll play around a bit more

[jaunty rain]

No, your prompt is not too imprecise... you need to understand how OpenAI API works. When you are querying their endpoints, you will receive a certain payload as an answer to the API request. This payload is like an "envelope" and inside you have your "letter". The content of the letter needs to be escaped (and unescaped by you) otherwise it would break the "envelope". There is no way to query an OpenAI API endpoint and say "give me only the 'letter'".

[full grail]

Ohhhh okay Hahahha my bad

Apologies

[jaunty rain]

You can't do this as I explained above.

[full grail]

Yep I understand now

[jaunty rain]

No need to apologize. 👍

[full grail]

I didn't really read the docs and I've fallen on the sword

[jaunty rain]

https://tenor.com/view/monkey-stick-hit-man-getting-gif-25806535

[jaunty rain]

https://platform.openai.com/docs/api-reference/completions

[zenith flare]

In this YouTube video, Nick announces the release of an AI chatbot assistant built into Jetbrains' Rider IDE, which is open for everyone to use as part of the Rider Early Access Preview. He speculates that it might be using OpenAI technology because Jetbrains did not mention the technology used in their blog post. Nick then proceeds to showcase the AI chatbot's features by using it to refactor code and find potential problems in a legacy application project. He also promotes his online courses and offers a 15% discount for his bundles using the promo code "summer23".

[surreal bluff]

Generated summary

A 1-message summary is being prepared in the thread below

Command Arguments:

Amount of messages: 100
Channel: #💬off-topic
Mode: emoji
Language: English

[spring aurora]

i hate this bot

[pallid bone]

I

Hate

Shortlinks

[hallow yew]

<!DOCTYPE html>
<html>
<head>
<title>Hello World</title>
</head>
<body>
<h1>Hello Wolrd</h1>
</body>
</html>

Am I making a mistake? I haven't touched programming in a long time

[orchid hatch]

lemme do it for you

@dusty mural <!DOCTYPE html>
<html>
<head>
<title>Hello World</title>
</head>
<body>
<h1>Hello Wolrd</h1>
</body>
</html>

Am I making a mistake? I haven't touched programming in a long time

[dusty mural]

Working on your response...
Done.

yusp48

chatgpt with GPT4

You have only made a very small typographical error in the "Hello World" string. The term "world" is spelled incorrectly within your heading tag (h1). Other than that, rest of your HTML code is perfectly correct.

Here's the correct version:

<!DOCTYPE html>
<html>
<head>
<title>Hello World</title>
</head>
<body>
<h1>Hello World</h1>
</body>
</html>

In this version, I've changed "Wolrd" to "World". Welcome back to programming!

2 messages • 74 tokens · Jun 29, 2023, 12:36 PM

[ripe cloak]

Hey everyone! Hope someone can help me here!

Im trying to figure out a way to call the ChatGPT API and give it context without burning up tokens. Specifically, I will be summarizing a document then passing along various questions to the document and return method calls. Does anyone here have a "best practices" on ways to compress that document or "store it" so I dont have repetitive token calls?

[jovial folio]

not possible

[ripe cloak]

@jovial folio very helpful...
So it is very possible, because I've done it already and it was pretty easy, its just a matter of finding better practices to make it more efficient. Thanks though

[ripe cloak]

Again, looking for best practices reference material.

[peak spade]

Hi guys, I'm a bit furstrated right now. I've been trying to implement GPT-3.5-turbo in my code, got the API key and everything, but when I access https:// api.ope nai.com/v1/chat/completions (had to split link because apparenlty I don't have enough social credit here) it still tells me the following:
"You didn't provide an API key. You need to provide your API key in an Authorization header using Bearer auth (i.e. Authorization: Bearer YOUR_KEY), or as the password field (with blank username) if you're accessing the API from your browser and are prompted for a username and password. You can obtain an API key from https:// platform.openai. com/account/api-keys."

The website it sends me to is exactly where I see my two keys. I even gave my card info and everything so it's definitely not because I don't have any tokens left.
I need this for a project for university and I'm pretty new to this kind of coding, so any help is appreciated. I'm coding in p5js using VS Code.

If anyone of you is willing to look at my code, I'm willing to share it with you. It's not complex either, as I said, still a beginner.

[peak spade]

What other endpoint can or should I use?

[jovial folio]

u need to send your token everytime you make a request

[peak spade]

I do. I have my API Key in my code. Or what exactly do you mean?

[jovial folio]

send ur code (make sure to not show ur api key)

[peak spade]

Will do. Thank you in advance!

[peak spade]

How else should I do it then?

You can’t send requests without authorization right?

[peak spade]

Noob question, but how do I do that?

Didn’t know that was a thing I could do.

[carmine trench]

Nah just google it

There are video and image tutorials

[full grail]

pip install python-dotenv in cli

from dotenv import dotenv

load_dotenv()

openai_api_key = ...

if using python

[carmine trench]

Most people just ask chatgpt to summarize it and store it, not sure how effective that is https://twitter.com/JeremyNguyenPhD/status/1674452220706734080

[languid plover]

Try llama-index

[tepid swan]

https://github.com/whitead/paper-qa

[ripe cloak]

@tepid swan thanks for that! very helpful actually!

[full lake]

$usage

[zenith flare]

In the last 24 hours you have used $0.0 worth of tokens. ($0.0 for prompts, $0.0 for completions)
The daily limit is $0.25.

[full lake]

#❔technical-support

[old sphinx]

$usage

[zenith flare]

In the last 24 hours you have used $0.0 worth of tokens. ($0.0 for prompts, $0.0 for completions)
The daily limit is $0.25.

[topaz pumice]

heeyy, I want to get chatgpt api spec. is it available somewhere?

[full lake]

#👋join

[jaunty rain]

The API spec is available here: https://platform.openai.com/docs/api-reference

[topaz pumice]

thanks : )

[limber warren]

I am learning the API and I often have a large instruction set for the prompts. I would like to avoid resending these instructions as the price per token will quickly cause a problem if I am to deploy my code. While looking into ways to fix this I came across two different things:

Roles/Personas I found very limited information on this, and was able to get the model responding as intended in the "playground" but the "View Code" button and all other adjacent ones were grayed out. Not sure why. I even googled this, but to no avail. There is very limited information coming up when searching "Chat GPT API Roles" or "OpenAI API roles". My Google search could have been more informative, but due to the ongoing Reddit blackout, nearly all the indexed results were unreachable.

Instructions I asked ChatGPT 3 what I could do, and it guided me toward instructions, but I haven't been able to find any info for this as "ChatGPT API Instructions" just comes up with tutorials on how to use it. This is extremely aggravating as the instructions feature seems to actually be what I require, but unfortunately I just can't reach the information I need.

If there are any good example open source projects that use either of these features, I would be happy with a link to those so I may learn from their implementation. Otherwise, just a link to relevant documentation would be fine.
Thanks!

[alpine bramble]

I'm looking at combining whisper.cpp and a locally run LLM to summarize voice conversations. Is langchain the right tool for me? It doesn't look like it but I want to double check

[hearty mortar]

Does anyone know of an online service where I can upload a PDF and get a token count?

Context: this is for a work-related project. I want to summarize large PDFs and interact with them via the OpenAI API, but I first need to forecast how much it could cost.

I'm aware of tiktoken as an option, but I'm getting back into software development after a very long hiatus and am quite slow at the moment 🙂 Before going through the process of setting up my dev environment and hacking around for a few hours I'd like to put together a budget and ask my CEO for approval.

Thanks!

[orchid hatch]

there is no tokenizer with pdf upload but you can try official https://platform.openai.com/tokenizer

[hearty mortar]

Unfortunately, there's a limit to what it will accept. I tried copying/pasting a 180-page PDF (55+ MB) and got an error message from the tokenizer. I found out there are some rules of thumb from OpenAI though. e.g. 1 token is ~ 4 english chars or 3/4 words.

[orchid hatch]

if it is too large for tokenizer, it is too large for any openai model existing

[orchid hatch]

put a project that you made

[warped quartz]

anyone wanna collab with a chatgpt

thing

[late lava]

Sorry to interrupt guys, but I would like to ask you one questions as you are probably more familiar with it.

What would be the best program to use in order to:

1. Inject 5-10 PDFs into the model
2. Based on those PDFs, produce another pdf

I want to produce one document based of my previous docs and as far as I know I found privateGPT, however I struggled with setting it up

[analog moon]

what?

[warped quartz]

like collab on a jailbreak for chatgpt

[spring aurora]

@heady flax

yo

bro

i need ur help

[heady flax]

Yeah

[spring aurora]

i dmd you

[honest moat]

Can anyone help me with my text converter prompt? It’s in the posts at the top as of rn.

[crystal summit]

Can you convert .apk to .ipa

[quartz wagon]

A team member has pushed a code change that has caused a critical error. How
would you handle the situation?

[flint aspen]

Imagine you are the worlds best software developer and i want you to modify an existing code of a python website system for me which allows a user with zero coding knowledge to just give a text based app or website idea and that system will generate the output for the user in world class style

[flat drift]

Sorry! As an AI language model developed by OpenAI, I'm not allowed to write code.

[steep shell]

ImportError: cannot import name from partially initialized module (most likely due to a circular import)

thought this was a chatbot lol

[jovial folio]

💀 💀 💀

[toxic dawn]

Hi guys

Anyone online?

[spring aurora]

I made a chat with GLADOS app if anyone is interested https://github.com/cameronaaron/Gladoschat/

[pallid bone]

#✨projects

[knotty fable]

I've got this snippet of code that uses function calling, but it keeps giving me a "openai.error.InvalidRequestError: '$.function_call' is invalid" error

I've checked the documentation and I think it's correct?

but the documentation about function calling is...kinda shit

https://platform.openai.com/docs/api-reference/chat/create

like if I'm reading this correctly then that function_call parameter should exist

okay, yeah, it's just a problem of the documentation being extremely unclear

response = openai.ChatCompletion.create(
    model="gpt-3.5-turbo",
    messages=messages,
    functions=functions,
    function_call=functions[0]
)

You have to pass the full definition of a function to function_call for it to work

[pallid bone]

From @cunning flower

If anyone intresting to integrate chatgpt in his local project make sure to take a look at Chatgpt-connect chrome extension, for more details : https://github.com/Ayyoub-ESSADEQ/chatgpt-connect

[pallid bone]

#✨projects

[twilit zodiac]

@flint aspen tell me about skiplist

#🪧rules

[eager spoke]

christ

[jovial folio]

lmfao

[spring aurora]

https://discord.com/channels/879548962464493619/1135741820349980742

help

[fallow plaza]

Guys, anyone knows if there's a way to make multi-step prompting without the API calls being cummulative? Meaning, if you have, say a 5 messages/interactions chat with gpt via API, you have to pay not for 10 messags (5x2, one for chat one for user), but (message 1) + (message 1 + 2,) (message 1 + 2 + 3) ... (message 1 + ... + 10)

I know this was the only way a couple of months ago, but it sounds stupid expensive to do things this way, did it change?

[glass imp]

@quasi nexus how

[jovial folio]

nope

[karmic wave]

I was wondering do you guys use ChatGpt prompts to make your apps using chat gpt?

[jovial folio]

..yes, thats how the model works

[fallow plaza]

nope as in "I don't know" or "It's not possible"? 🙈

[orchid hatch]

nope as in not possible

[orchid hatch]

cant you just fill the conversation history with your multi step prompt

and why do you need a multi step prompt

[fallow plaza]

cant you just fill the conversation history with your multi step prompt
Yeah but how do you generate that history?
and why do you need a multi step prompt
I mean, there are endless reasons why you might need that. The most obvious one I can think of is an actual chat where you can keep a conversation that it remembers. I understand this is what ChatGPT does, but it's just an example @orchid hatch

[jovial folio]

^

[fallow plaza]

thank you both! @orchid hatch @jovial folio

[eager spoke]

How do I use the 32k model? I can't find any reference to it in the docs.

[jovial folio]

change the model name to gpt-4-32k

[uncut swift]

OpenAI announced a few weeks ago that they will make GPT-4 API available to everyone. I still cannot access it and I'm stuck with 3.5. How about others? Did you get access to GPT-4 API recently?

[jovial folio]

you need to spend money on gpt3.5 to get gpt4

[mellow marten]

Anyone know where I can get a crash course in all of this? All I have is a blu view 2 and a willingness to learn

[pallid bone]

Can you please elaborate more

[mellow marten]

Im able to talk to chatgpt and get desirable results, but there is so much here to learn, and besides gpt I could use a good ,patient teacher.

[noble forum]

i make discord group

after discord.gg its /RKx6mdU8

[signal forge]

Hello,

[noble forum]

bruh

[signal forge]

Is functions available on 3.5? I thought model gpt-3.5-turbo-0613 could handle them , but I keep getting 500s

[jovial folio]

skill issue

gpt3.5 does support it

[uncut swift]

In their announcement, they said they will give access to everybody by the end of July: https://openai.com/blog/gpt-4-api-general-availability

"Today all existing API developers with a history of successful payments can access the GPT-4 API with 8K context. We plan to open up access to new developers by the end of this month".

[jovial folio]

bro really trusts openai💀 💀

[uncut swift]

Their name should be ClosedAI 😉

[slow wing]

hi, I'm getting all these <|fim_suffix|> messages from GPT followed by "Error in message stream"; trying to use the code parser to tell me about a Uniswap deploy script

repeatedly

ok this is just specifically re one file .. n/m

[languid beacon]

Hi guys, I'm having an issue with the API. I am trying to generate some content using PHP and Guzzle library, however the responses are half baked and intrerrupted. Although, in the playground the responses are full and exactly as prompted. Any ideas?

On top of this, I ask the API to return JSON object for my content, and most of the times it returns invalid JSON...

[pallid bone]

#1132309502931238962

[languid beacon]

So essentially, you need to wait for a full month until the system charges you in order to be "considered" for GPT-4. However, their announcement is contradictory...

[cedar tree]

Does anyone here use Pyarmor? Especially the recent one?

[spring aurora]

Hey there, i am having a kinda good problem but not entirely sure how to get it sorted.
I have a (naughty) chatbot and some users. I recently got a notice of OpenAI that i am breaking their ToS with my use case.

So where do i migrate to? I am looking into self hosted options, seems the only good model that allows commercial use is Llama?
Renting a machine on Vast.ai and using dalai is the way to get started?

[spring gust]

I am not sure if this is the right forum, but I run a public health nonprofit. We are interested in leveraging AI to align with a new public health theory. Anyone can assist with coding and such?

[spring aurora]

use falcon of llama

[spring aurora]

sure watchu need

[spring gust]

We posited a novel concept. AI does not understand this - the public health economy. We want to work with AI developer(s).

[spring gust]

We posited a novel concept. AI does not understand this - the public health economy. We want to work with AI developer(s).

[spring aurora]

ok

well I can maybe help

just tell me what you want

https://tenor.com/view/spice-girls-wanna-be-gif-4227106

[spring gust]

I want an AI that can describe the public health economy at the level of the city, state, and national level....that can integrate social, economic, and political structural factors. For example, Washington, DC has some of the worst racial health disparities and one of the largest per capita budgets. What is it about DC's public health economy that makes it different. I suspect that I know why, but AI would help.

[spring aurora]

ok

so

use public health api > pass it to LLM api > condense it > display it

[spring gust]

I don't know what that means. Haha

[spring gust]

Can you walk me through how to do that?

[spring aurora]

ok

gimme a minute

[spring gust]

ok

[spring aurora]

ok

so what do you want it to do exactly

hold

a

[rose nest]

so I've been fiddling with some things in the api thinking that just messing around would surely cost $1 soon enough. a couple weeks later I looked up and I have $0.02 on my balance. any advice for how I can rack up a $1 asap so I can use gpt-4?

[spring aurora]

hi

any free python coruses

[pallid bone]

yes actually

here

https://www.tutorialspoint.com/authors/mrudgandha-kulkarni

[spring aurora]

thnx

[jovial bear]

Hello, guys, how are you? Where can I find datasets to train model for roleplay? Base model is llama2-7b. Please, could you help me?

[carmine trench]

@twilit flint can you do it for this category as well

[twilit flint]

done

i hate this bot

[craggy kestrel]

For machine learning If I use google colab for gpu would that mean I could balance by buying a high performance cpu and ram on my computer for data prep. At what point is the cpu and ram overkill?

For example will a beginner ever need a Ryzen 9 with 64gb of ram but not running model on local gpu?

[primal meteor]

So I want to make a small tool for helping run my scifi d&d games
I want to make a Python app maybe with jupyter notebooks to prompt gpt with all my world building. I would like to integrate some image generation tool as well but that's later
If possible I would also like to be able to parse some local PDFs with contextual a dgake rule information so the players can query.
Has anyone else built a tool like this I can use or get inspiration from?
Currently I have a ryzen 9 7900 x, and a 1660ti 8gb. But in the next few months plan on getting a better GPU

[pallid bone]

A minimalistic reddit like forum

The folder structure:

- templates
  - forumpage.html
- posts.json```


In order for it to work:
```pip install flask markdown2```

[pallid bone]

Old tampermonkey script that applies bing theme

[warm raven]

Does anyone know why when using the openai api it sends me this error?

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

[pallid bone]

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

[full estuary]

I was searching something about python when the google foobar challenge popped up, I didnt know what it was so I searched it up. I am a starter, but still trying my best to do something. In the level 1, I don't understand why all tests fail, but when I try the code in my vs code environment, it works, and still cannot figure it out. my code is this(python):
def is_prime(y):
x = round(y/2)
prime = True
while 1 < x:
if y%x == 0:
prime = False
break
x -= 1
return prime
def solution(i):
result = ""
while len(result) < 5:
i += 1
if is_prime(i) == True:
result = result + str(i)
if len(result) > 5:
result = result[0:5]
return result
and this is the task(summarized):
Commander Lambda aims to address unrest among minions by reassigning new IDs based on prime numbers. Prime numbers are concatenated into a string, and minions draw a starting index from it. The next five digits in the string, starting from the index, become the minion's new ID. A function, solution(n), is required to return the new ID given a starting index n within the range of 0 to 10000. Both Java and Python solutions are accepted. Test cases are provided for verification, ensuring the function's accuracy.

Sample Test Cases:

For Java:

Input: Solution.solution(0)
Output: 23571
Input: Solution.solution(3)
Output: 71113
For Python:

Input: solution.solution(0)
Output: 23571
Input: solution.solution(3)
Output: 71113
The challenge revolves around implementing the solution function to generate and return new minion IDs based on prime numbers.

Can anyone tell me why it isn't working?

[warm raven]

bro, is a new account

[pallid bone]

Idk, the response says it clearly. Maybe try to fill your credits?

[digital lily]

Can anybody point me to an article / video on the approaches for using openAI to work on an existing repository rather than just a question? I know there's a ton of stuff out there, trying to find a starting point. (I've read the first paragraph of too many articles to know what's current and where to start, lol)

[hidden finch]

Does anyone know where I can find an explicit schema for describing functions for function calling in GPT?

The following is given as an example in the API reference
{
"name": "get_current_weather",
"description": "Get the current weather in a given location",
"parameters": {
"type": "object",
"properties": {
"location": {
"type": "string",
"description": "The city and state, e.g. San Francisco, CA",
},
"unit": {"type": "string", "enum": ["celsius", "fahrenheit"]},
},
"required": ["location"],
}

However what I'm curious about is the properties section what sort of types we're allowed to specify.

So far I've only been successful in using integers or strings as the property type. Is there a reference describing other types that the api accepts?

[edgy flume]

Is there a place to get good custom instructions fdor programmers?

[orchid hatch]

#1141581988332064818

[void mantle]

I need some help from a python programmer

I'm pretty good at prompting and have experience with a coding-adjacent activity, so I'm able to get chatgpt to write and iterate on code to get what I want.

My problem is, I'm to the point where my code is too big to print out in a single response, every time I make a change, but I don't have any good workflow methods for taking coding snippets and figuring out to put them in my codebase.

I really need some help developing a workflow so I can follow through with having chatgpt write complete programs.

[rapid lichen]

Im trying to create a neural network that tells handwritten numbers in a 10x10 grid could you guys help

[pallid bone]

hi

[grand python]

Hello

[wise tinsel]

That's a great idea. I could probably help with that.

Not sure what that other dude was talking about though.

[wise tinsel]

I'm just using diff-like changelogs to patch things around.

In addition to spitting stuff into chunks, and reassembling again them afterwards.

[orchid hatch]

lol

[wise tinsel]

What?

Every time I say something this dude laughs

Transferring a small diff forth and back is much more convenient than dealing with large files

[wise tinsel]

ChatGPT is quite proficient at generating manual diffs so you don't need external libraries (like I thought initially)

[orchid hatch]

so you are basically using git

[wise tinsel]

No way I'm touching this shit with Python

yura: No, I'm not

[orchid hatch]

so you are basically using git but just in a much less cooler form of sending raw diffs

[wise tinsel]

https://github.com/alexrudall/ruby-openai

No way I'm touching this shit with Python 😁

[mossy forge]

iany experts on node.js and puppeteer

??

[midnight sedge]

yes

[carmine trench]

https://dontasktoask.com/

[tacit sail]

Can somebody help me with some css please dm i have a problem in my footer i can’t identify it to solve it

[quaint holly]

no one will as long as you don't show the actual problem

so post it here if you really need help, otherwise don't bother ig

[stiff nest]

@quaint holly +1. I'll look at it briefly, wrap in ```CSS
followed by code and ending with three backticks or send as a file

[stiff nest]

@ me if so as I don't frequent this channel

[plain rock]

Any jailbreak steel working

?

[pallid bone]

No

[tacit sail]

@quaint holly here it is the problem is the contents inside it a phone number email and instagram link are not properly centered

[quaint holly]

can you post html too? Would be much easier.

[quaint holly]

yes, my bio. Also, wrong channel.

and not 'steel' lol

https://tenor.com/view/alec-steel-forging-gif-11750521

[quaint holly]

@tacit sail Ok I'm feeling generous... and maybe a bit curious so looked into it regardless.

but with cGPT generated html for it everything looks fine.

as far as the centering at least.

unless you mean like the icons being slightly higher than the text next to them.

I can center the email one with

  margin-top: 5px;
  margin-right: 10px; 
}``` 

but then 'home' becomes uncentered. Your issue is that those icons are cropped differently with some of them having white  background at the top or bottom.

[spring aurora]

average html debugging

[tacit sail]

@quaint holly check dms.

[pearl kelp]

Ive noticed Chagpt is really good with small codebases and single file code but it struggles with coherance and isnt really capable of large scale projects

I cant imagine how powerful future models will be

[slate salmon]

I think I developed a method to solve this problem

[quaint holly]

You need to edit the earliest relevant prompt

do not just keep continuing your chat further down because it will forget the crucial info that was before it.

so when you're done with a question and need to ask something else about that code, replace it. You can replace your code pasted earlier too if it changed.

and as for the output, you can just ask it to rewrite the affected method in full. If the code is really long and there are many methods affected, you can first ask it to tell you how many/which methods needs modifying. Then edit the same prompt and ask it to rewrite those one by one. Editing same prompt each time.

gpt4 is smart af generally, but you do need to know how to use it. You can't just expect for it to remember your whole chat with shit-ton of code. Front-end is just interface and it doesn't reflect what it actually has stored in full in it's context and can effectively reference to.

[stiff nest]

yep, it's always best to work chunk by chunk, even if you're using the code interpreter. Well factored code should isolate concerns anyway

[cosmic fjord]

is this the place to get help with Open Interpreter? trying to get it locally on my machine and all i get is this when trying to install

[onyx lava]

that happens to me all the time

[frigid nexus]

Hello, Every body.

Can you help me?

Is there anybody who able to make ChatGPT plugins? Did you pass the waitlist in ChatGPT?

[cursive quiver]

@void mantle try this prompt:The total length of the content that I want to send you is too large to send in only one piece.

For sending you that content, I will follow this rule:

[START PART 1/10]
this is the content of the part 1 out of 10 in total
[END PART 1/10]

Then you just answer: "Received part 1/10"

And when I tell you "ALL PARTS SENT", then you can continue processing the data and answering my requests.

Do not answer yet. This is just another part of the text I want to send you. Just receive and acknowledge as "Part 1/5 received" and wait for the next part.
[START PART 1/5]

And so on...

[untold comet]

Could someone give me a helping hand here? Im getting this error:

Error reading data: TypeError: Cannot read properties of undefined (reading 'create')``` Using this code:
```js
  client.on('messageCreate', (message) => {
    if (message.channel.id === "1008518385262403615" && !message.author.bot) {
        const userId = message.author.id;

        try {
            const jsonData = fs.readFileSync('data.json', 'utf-8');
            memberData = JSON.parse(jsonData);

            if (memberData[userId] && memberData[userId].inputValue) {
                const api_key = memberData[userId].inputValue;

                openai.apiKey = api_key;
                openai.Completion.create({
                    engine: 'davinci',
                    prompt: 'Your prompt here',
                    max_tokens: 50,
                })
                .then((response) => {
                    const aiResponse = response.choices[0].text;
                })
                .catch((error) => {
                    console.error('Error generating AI response:', error);
                });
            } else {
                message.channel.send('You have not set up your API key. Please provide your API key using the `/open_modal` command.');
            }
        } catch (error) {
            console.error('Error reading data:', error);
        }
    }
});``` I have been stuck on this for a few hours now, and im out of ideas. Any help is greatly appreciated!

[unreal niche]

Is there a method to implement ChatGPT's "Stop Generating" function ? Any comment or resources would be appreciated

[vapid summit]

const openai = require('openai');

At the top, make sure you install OpenAI modules.

[vapid summit]

If you want some bots to play with lmk I've made more than ten in both Js and py. There's also a lot on GitHub. I've got a few you could have source code for, including one js that does dalle

[untold comet]

I got it working. Some stupid mistakes on my end. Thank you so much! 🙂

[pallid bone]

#✨projects

Please put project related stuff in #✨projects

[muted mesa]

Pls Solve this using c++

Thx don't need anymore

[terse torrent]

https://github.com/dizzysaturn/soundstorm hey everyone! brand new to coding, but i wanted to learn, so ChatGPT's become my teacher. i've developed a Python script/app for sound designers. It features AI text to audio, algorithmic composition, a MIDI randomizer, GPT smart prompts, and even onscreen ChatGPT. It is still in its infancy since i know very little about coding, so feel free to fork it on github and make it your own! The fact that I did this knowing nothing to me shows the insane power of ChatGPT.

[orchid hatch]

owo

[open harness]

CAN i have a look ?

[bleak sparrow]

does anybody know using azure open ai services and implementing to a company with finetuning and what languages does gpt support thanks!

[ocean marlin]

Anyone got a guide say, use LLM for say c#.

[silk yew]

@everyone For chatGPT to be able to remember the conversation through the API do i really need to pass all the previous messages to the redquest?

How is that even possible as some conversations can become infinetly large

[fickle glacier]

How else do you expect to create a chat completion within the context of previous messages?

[silk yew]

I thought every message/prompt updates the layout of the model, so you could just get a link for the model and every input/output changes it

like training deep learning models work

[fickle glacier]

From my understanding, prompts do not change its layout of the model or make it "learn" anything. As it can be understood from its Web GUI every chat is stateless, it does not retain information from any other session.

[willow heath]

i did it in python:
import numpy as np

def solve_equation(equation):

parts = equation.split("=")

left_side = float(parts[0])
right_side = float(parts[1])

solution = (right_side - left_side) / 0.01

return solution

solution = solve_equation("1 = 0.3 + 0.01x")

print(solution)

[dusk edge]

hell yeah

[bitter ginkgo]

I need to process a large text (larger than the 16K limit) and I want to ask questions against it. How's that possible? I tried breaking it up into chunks but I don't know how to keep context between me sending the chunked text parts. Is there a way to do it?

From the docs it mentions Embeddings as well:

Embeddings can be used to implement efficient knowledge retrieval. See the tactic "Use embeddings-based search to implement efficient knowledge retrieval"

But I'm not sure if that's the right path. Anyone have ideas?

[dim wasp]

I have a repeated problem inputting large chunks of code into ChatGPT.

I will state that I'm going to paste in my code in say 3 or 4 chunks, and then want help with the code once it has all been pasted in - but by the time I paste in the second part, the chatbot has lost the context of what is happening and asks a general question about what I need help with.

Is there a better way to do this?

[next hamlet]

posting here in programming as well
Hi guys, going to create a site/app and am following a basic guide from Web Dev Simplified for right now and also the OpenAI docs. I made sure to run npm install openai@^4.0.0 as well. i attached a pic of my code and the error in terminal that comes up after running node index.js. Any help would be appreciated.

i'm following everything exactly so i'm not sure why i'm getting "SyntaxError: The requested module 'openai' does not provide an export named 'Configuration'" it looks like the exact same way they're importing in docs

[rich cargo]

the guide by wds is outdated and the docs have changed

const openai = new OpenAI({
  apiKey: process.env.GPT_KEY
});

amung others

[next hamlet]

dang, but i'm looking at the most recent docs and following their example for Node.js package?

[rich cargo]

trying to find what im looking for rn give me a second... because i had the same issue a few weeks back

[next hamlet]

thank you i appreciate the help! i forgot tech videos can get outdated pretty quickly but i thought the docs should be up to date ahah.

[rich cargo]

https://github.com/openai/openai-node/discussions/217

// Old
import { Configuration, OpenAIApi } from "openai";

const configuration = new Configuration({
  apiKey: process.env.OPENAI_API_KEY,
});
const openai = new OpenAIApi(configuration);

// New
import OpenAI from 'openai';

const openai = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY // This is also the default, can be omitted
});

[next hamlet]

😮 one sec i have to try this

[rich cargo]

yup let me know if it works

[next hamlet]

you're a mf lifesaver man

i was starting to think i was dumb af, i was like i'm following guides and the docs themselves wtf 🤣

programming always makes you feel like that 🥲

[rich cargo]

imposter syndrome much better than superiority complex imo

https://github.com/openai/openai-node

use this for the most up to date info and whatnot

much easier to undertand than the website

[next hamlet]

thank you! also how did you find that announcement and that specific fix???

so i know in the future

in their ReadMe, i see the 'important' notif at the bottom. but without knowing to check here, i would have never known 🤣

[rich cargo]

yeah i had to ask someone in another server they pointed it out

its been outated on the website for a while now...

but the github readme is pretty up to date

[next hamlet]

D:

wow... you're a lifesaver man thank you

time to hone my github readme skills lol

[flat drift]

Dev Tools Questionnaire (for SDE/MLE/PM): fill this to help us (early stage startup) build the best AI dev tool for you! Let's build something useful! Much appreciated!

https://forms.gle/v7Mg5itfqKGG22mL7
sorry @blazing lagoon, anti-spam got you. you're all good though

[floral burrow]

Hello Folks 🤗
I have truly off topic and random question... 😓
I know it's a basic inference question but I am willing to know which is the best way to inference in batch.

The overall goal
-> I am using t5-small model for summarization task.
-> And I have around 10 to 15 different paragraphs to be summarized in a single call.

👨‍💻 The code I am using right now
It is generic loop code but I expect some optimization here:
points = [
"summarize: ABC...",
"summarize: CBA...",
"summarize: ERG...",
"summarize: RAG...",
]

summaries = []
for point in points:
input_ids = tokenizer.encode(point, return_tensors="pt", max_length=512, truncation=True)
output_ids = model.generate(input_ids, max_length=256, temperature=0.35, do_sample=True)
summaries.append(tokenizer.decode(output_ids[0], skip_special_tokens=True))

This one takes time as expected.

😯 I have tried this...

Passsing input ids in batch

ids = tokenizer(points, return_tensors="pt", max_length=512, padding="longest")
response = model.generate(**ids, max_length=256, temperature=0.35, do_sample=True)
tokenizer.batch_decode(response, skip_special_tokens=True)

But I am worried if the model will connect the paragraphs internally and will leak information between each. I am not sure but this way is significantly fast than the loop way.

🤔 My Ask
Am I doing it right? How to perform the batch inference which is so fast and all inputs are NOT talking to each other?

Is there any other way to increase the speed? (Or would I just need to use threading?)

Thanks!

[verbal sun]

Been searching up and down for what feels like an eternity so figured it was time to finally turn to the lovely people here.

I'm really just trying to figure out the current state of text-to-code AI-coding assistants. Are any of them that I've not come across able yet to allow a user to describe what they would like an app/service to do and for something workable to be generated (I'm assuming followed up with many fine-tuning/error-related prompts.) If so, it would be greatly appreciated if anyone knew some of the most advanced AI services able to achieve such a high level of coding proficiency like this? Cheers!

[cedar umbra]

Apparently I am too new (not enough social credit) to share links, but I think the most promising avenue so far is around multi-agent choreography. Search Youtube for Wes Roth's videos on ChatDev and AutoGen.

[barren marten]

Yeah! Chatdev is the future when it comes to coding for me.

[orchid hatch]

rekt

[visual solstice]

They're talkin' 'bout algorithms, but it's all bluff,
Their code's so tangled, it's just plain tough,
I'm the real deal, my solutions are neat,
While they're drownin' in errors, I'm on easy street.

[stiff hollow]

chatgpt

[tall musk]

https://github.com/webmaster-exit-1/terminal-assistant
Check out the AI assistant, for the linux terminal I made
Features

• Google Search
• Google Voice Recognition
• Google TTS (Speech playback -- Assistants Voice)
• OpenAI GPT-4

Usage
Run the terminal-assistant.py file using the python command.
"Enter" To respond with text
"Shift+Enter" To respond by voice
Type !search to use Google search,example: !search cats
To exit the chatbot, type quit.

It's easy to add code to it, or have gpt help you, it helped me make this.

the model(gpt-4, ect...) line of code might need to be updated, but that would be the only thing, as I made this 3 months ago, before gpt could do this on its own 😏

If it's worth a star on github, cool thnx. If it sux, fork it and make it better

AI in under 5kb size with features, in your terminal.

TODO: (tried and failed, if someone out there with talent can, would be pretty epic) Add ascii audio wave synced with AI's audio output, In the Terminal 🤓 🤤 that would be cooooool

oh, right, It uses (it, the repo) github actions, so like right now, there was a small update, i signed off on it, and it automatically updated, built and tested the code

nice

[shy lodge]

Hey guys I'm new to this discord , just saying hello and asking if any one can help me I'm going nuts I'm trying to create a script using chat gpt 3.5(the free one) to export data from a pdf to excel the data in from a lay out plan and since it's schematics it's tricky all while doing this in psy3 any help to guide me to the correct path please

I am very very very new to programing like the few things I know are self thought and playing with chat gpt lol

[crisp cypress]

Am new to programming as well I am interested in computer languages but I always procastinate

[cobalt hemlock]

hi, anyone here using chatgpt openai? what does it mean to train a model? is it actually fine tuning?

[hybrid zodiac]

anyone know where I can get a couple H100s for free?

[hybrid zodiac]

(pre-)training = unsupervised learning on a massive, unlabeled corpus.
fine-tuning = adapting a pre-trained model to specific tasks via few-shot learning to improve performance on specialized tasks

to be more precise, pre-training sets the model weights and fine-tuning adjusts those parameters given the smaller dataset

[sacred grove]

Hi is there anyone on here I can pay to help me create a wrapper app for a microlearning site?

[dreamy cedar]

Can anyone who knows python, and google drive api take a look if this looks right?
https://chat.openai.com/share/b6bfec60-9f88-43f6-b495-430fcdad6d13

pls ping

[brisk slate]

Generate a code to draw zero in python

[midnight sedge]

print("""
  ____
 /  / \
|  /   |
| /    |
 \____/
""")

[quartz yacht]

portait of me

[spring aurora]

pip install get-bitches didnt work

[midnight sedge]

portrait of ur mom

[jaunty rain]

@whole stag are you an employee of OpenAI LLC? If not, do not impersonate them here in this server. Change your message or I will delete it.

[tender forum]

Hey guys can you help me with my survey i am doing about how many people concern about sending personal data to openai/chatgpt.

Won't be any personal question or email.
https://tally.so/r/n08Zx9

[blissful shard]

Hey, the /chat command is deprecated. Just ping me in your prompt!

[plucky skiff]

Write a program in C++ to print prime numbers from 5 to 51 using (while)

How can i interact with this bot

Generat a code in C++ to print prime numbers from 5 to 51 using (while)

[jovial folio]

"just ping me in your prompt"

0 reading comprehension

[plucky skiff]

@jovial folio
Write a program in C++ to print prime numbers from 5 to 51 using (while)

[jovial folio]

ping the bot not me

[plucky skiff]

@blissful shard
Write a program in C++ to print prime numbers from 5 to 51 using (while)

[plucky skiff]

How long does it take to respond?

@blissful shard
Generat a code in C++ to print prime numbers from 5 to 51 using (while)

[jovial folio]

30 business days

[plucky skiff]

[quartz yacht]

Portrait of your percentage into losing virginity

[midnight sedge]

NaN

[quartz yacht]

nan%

also in unrelated note how to get more credits without buying credits?

[jovial folio]

trade credits

[quartz yacht]

Wanna trade?

[jovial folio]

@flat drift

[jaunty rain]

?ban 398575231280742411 stop impersonating OpenAI employees here.

[torn saddle]

shinobi9877 was banned. | stop impersonating OpenAI employees here.

[wise tinsel]

Looking for folks familiar with LLMs and Langchain. In particular I need some advice about custom tools.

Will pay. I also have a lot of awesome scripts to share.

[south light]

I am a beginner Python developer. When Chat GPT shows me snippets, it keeps replacing functional code with comments like "# (Your existing code to process the update here)". No matter how many times I tell it not to do that, it just won't listen. I have filled out both sections of the custom instructions and included statements telling it never to do that, but it just won't listen. Does anyone have any advice on how to get it to stop this behavior?

[north minnow]

No, AFAIK, it's hardcoded to do this because of rate limits. Every single thing that GPT has to process is done by tokens, as far as I know. So when it has to reiterate your code, that takes compute time to process it. So for instance if you ask it to help you add a print statement to a 550 line script.... the rationale is that it should NOT include your whole ass script, bc that's a waste of resources.

[south light]

I gotcha, that makes sense.

[grave idol]

Hey everyone! Anyone knows if there is a proper API schema for function calling? All I can see if examples, but not the schema. For chat gpt API to be precise

[elfin bridge]

Hey everyone hope you're having a good day
Wanted to ask if any of you can help me make a voice cloning ai in persian
If you can plz dm or smt

[signal forge]

Anyone having problems with an extremely slow API?

[wise tinsel]

Please post the latest code in full, every single line with no omissions, truncations or placeholders usually works for me

[inland radish]

anyone using openai apis here? how do you deal with the ratelimits and no responses? it's crashing a lot for me lately (using GTP3.5 Turbo model for text summarization). You can check my code here: https://github.com/robertlangdonn/news_digest_summary in news_summarizer.mjs

[rapid lichen]

Hello all, created a game using chatgpt API , game is 20 questions, did the programming in godot. If anyone is interested have a look here https://youtu.be/0WkNqAVr6pM . How are you guys using their api?

[surreal cipher]

I would like to create a custom role playing chatbot for training in a specific topic., and embed this in an app.

My current plan is to use the OpenAI chatgpt and whisper APIs.

I have a couple of questions:

1. The text to describe the roleplay is long, perhaps 500 words. I could put this in the context each time I call the API, but there is a cost to that. Using ChatGPT fine tuning doesn’t seem to fit; it is much easier to describe the role in words. Any other options?

2. Also, I would like to give the chatbot additional knowledge, perhaps 10-30 pages of A4 text. How would I best do this?

3. Is ChatGPT the best model to use? What else should I consider?

Many thanks

[acoustic smelt]

i am new to python and consider purchasing gpt4 to help me learn and code. Is gpt4 good at suggesting python code? especially regarding pytorch?

i used gpt 3.5 for pytorch, but it often creates flawed code suggestions

[visual crest]

hello, how can I use openai api for chatting and coding?

[daring briar]

by "chatting and coding" do you mean you want to build something that uses the API? or do you just want to chat with ChatGPT about coding?

[ocean flint]

what is the difference between "chat" and "assistant" ?

[obsidian locust]

Hello I make some exprerimentation in google colab. I have a mystal model with llamacpp.
When I generate something but the generation is not finished (like cut at the middle of a word), when I rerun the generation he complete the previous sentence, but my prompt is always the same.
I haven't found in documentation if it is normal a model can continue a generation without changing the promp with the response generated by the LLM ?

[obsidian locust]

@surreal cipher I have the same idea in mind. What I conclude right now it to use only use the Model to generate the story, and have another piece of software to manage the context to give (like RAG, current quests). You can also use the model in background for make summary you reuse.
Model like GPT don't have good memory, because is based on token not real memory variable somewhere.

[tidal cipher]

Anyone here used Llama 2 for inference and fine tuning?

[umbral granite]

How can I use Redis bitmap

[sharp fable]

Hi guys,

I'm not a programmer but since chatgpt came out I used it often to either create some simple python scripts or modify existing scripts.

I'm now trying to do the same with C# but it keeps failing. I'm not sure if that's because chatgpt 3.5 can't access URLs to read the API of the software for which I want to create the code, or what.

I never used gpt-4 but I was wondering if I'd have better luck with it and I should get subscription.

Thanks

[umbral granite]

"I'm not sure if that's because chatgpt 3.5 can't access URLs to read the API of the software for which I want to create the code, or what.", you mean you feed chatgpt some urls to let it generate some code for you?

[sharp fable]

Hello, no what I mean is that maybe gpt3.5 is failing with the tasks I'm requesting it to do because its knowledge cutover date and because even if I asked it to read the API it wouldn't be able to do that. So I was wondering if with a gpt4 subscription I would have better luck

[obsidian locust]

I don't have GPT4 but for ChatGpt 3.5 you can copy past the doc and ask questions.

[round basalt]

i cant publish a GPT to the public. as my website is not ticked green in the builder profile

[spring aurora]

Hi everyone

[timber falcon]

using gpts and trying to create custom actions. it doesn't seem to be calling my api at all - and there is no trace on open ai side to tell me what endpoint/what is being sent?

this is all i see :/

[slender merlin]

I'm writing content for a highschool level class in robotics. I'd like to do a module on LLMs and I have a lot of ideas that already work with ChatGPT. But... because these are highschool students, I can't require that they spend money to get an account at Open.AI to use the API there. I hope they will, or that they can use the free account, but I'd love to include other options. Are there LLMs online with API access that are "free" or pay for their use by inserting advertisements in responses or something like that?

[calm mango]

You can use meta's llama 2: https://ai.meta.com/llama/ and run the LLM on your own device. It is a bit slow but it's basically a free chatGPT.

[slender merlin]

The "on your own device" part would be difficult there as I'm not allowed to require the installation of any software... it all has to be web / cloud / other peoples computers based. The students get chromebooks and like that. Hmmm... perhaps I could setup a server... but that is also not so easy. It's been a while since I've spun up a server...

[calm mango]

Yeah that's unfortunate... I'm not sure then.

[graceful quail]

You can provide them the API by setting up a server. Your students will query the server (imagine we are requesting Google) then in your server, you can call your gpt, which then of course uses your API key.

To reduce abuse actions. You can ask your students to register with student ID. This way we can lock them immediately. Authorized user will also make it harder to forge a request.

[slender merlin]

Oh... I think I understand. Setup a simple e.g. Google aps server and assign the students IDs which it recognizes. Then pass on their requests to openai, and rate / total count limit those requests per student.

[timber falcon]

wow chatgpt updated actions to actually show debug messages wow, they are listening to my feedback amazing. can't believe they read my messages and implement so quickly

[spring aurora]

went first AI CHAT DEEPAI since start 15-2018 and u can show ur picture and video than CHATGPT

[lunar marsh]

Self deleting cheat anyone .bat??

[pearl kelp]

Self botting on discord using gpt4. Honestly kind of useless people just ban you for talking like a bot

[craggy turret]

Hey, can anyone tell me if assistant remote function calling works from the public webui? Or does that only work within a self hosted openAI wrapper application?

[neon trench]

Not sure about the first part, but I just built a 'wrapper' that doesn't need hosting...

...and I can call functions

https://github.com/fr3shd3v/DMAIJS <- this is it as an example

[timber sleet]

Hey, I'm trying to test out the new Assistant API, but I can't get it to work, i have used the example openai has on their website, and this is the error:

Traceback (most recent call last):
  File "C:\Users\Bruger\PycharmProjects\SpeechTest\main.py", line 10, in <module>
    assistant = client.beta.assistants.create(
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\Bruger\PycharmProjects\SpeechTest\venv\Lib\site-packages\openai\resources\beta\assistants\assistants.py", line 95, in create
    return self._post(
           ^^^^^^^^^^^
  File "C:\Users\Bruger\PycharmProjects\SpeechTest\venv\Lib\site-packages\openai\_base_client.py", line 1055, in post
    return cast(ResponseT, self.request(cast_to, opts, stream=stream, stream_cls=stream_cls))
                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\Bruger\PycharmProjects\SpeechTest\venv\Lib\site-packages\openai\_base_client.py", line 834, in request
    return self._request(
           ^^^^^^^^^^^^^^
  File "C:\Users\Bruger\PycharmProjects\SpeechTest\venv\Lib\site-packages\openai\_base_client.py", line 877, in _request
    raise self._make_status_error_from_response(err.response) from None
openai.BadRequestError: Error code: 400 - {'error': {'message': "The requested model 'gpt-4-1106-preview' does not exist.", 'type': 'invalid_request_error', 'param': 'model', 'code': 'model_not_found'}}
****

Any help?

[neon trench]

'gpt-4-1106-preview' does not exist

[neon trench]

use a different model

[timber sleet]

what model should I then use, for the assistant API?

[neon trench]

I use gpt-3.5-turbo-1106 ... it can deal with tools (functions)

I have no access to gpt4 via api 😦

[timber sleet]

okay, thanks, but isn't it weird they use a model you can use in their examples?

[neon trench]

bruh the documentation marks "functions" and "function" as deprecated, and recommends you use tools instead.
yet, all examples are using functions!!!

this is why I came here...

[timber sleet]

yeah, the examples doesn't seem to be updated...

how do i access the the content that i just generated from the response json:

messages = client.beta.threads.messages.list(
  thread_id=thread.id
)

print(messages.) #??

[neon trench]

uhm... hard to tell like this, as I don't use the python lib...

print(messages) doesn't do anything?

[timber sleet]

what do you use?

[neon trench]

I wrote my own thing in Vanilla JS...

[timber sleet]

yes, but it prints the whole json:

SyncCursorPage[ThreadMessage](data=[
  ThreadMessage(
    id='msg_TP0HhUcL1rC8apBaVp9NG3S0', 
    assistant_id=None, 
    content=[MessageContentText(text=Text(annotations=[], 
    value='Best thing to see in paris?'), type='text')], 
    created_at=1700065246, 
    file_ids=[], 
    metadata={}, 
    object='thread.message', role='user', run_id=None,             
    thread_id='thread_2zCqsth2mX6lEXqiaYyxAU39')], 
    object='list', 
    first_id='msg_TP0HhUcL1rC8apBaVp9NG3S0', 
    last_id='msg_TP0HhUcL1rC8apBaVp9NG3S0', 
    has_more=False
)

[neon trench]

Looks like it prints an object?

[timber sleet]

No idea, but i think so yes, but i just was that there is no response in this, so I'll go back to working on it

[neon trench]

message["data"]["value"]

[timber sleet]

nah, that throws an error...

[neon trench]

Uhm I have another example on github from yesterday, where I used the python lib...

https://github.com/fr3shd3v/DMAI

there I am using python and I am calling functions...

maybe that helps...

[timber sleet]

ah, thanks, Ill have a look at it.

[neon trench]

hey, feel free to dm or ping me, I just started with openAI api and I am very happy to talk about it 🙂

[timber sleet]

for sure, thanks : D

[signal forge]

Hello, I'm using 3.5-turbo through the API. When it works it gives me what I want, but the problem is that it's very inconsistent with responding. Sometimes it's just 3 seconds for a response and then 5 minutes later I would make a similar request and it just doesn't respond. This changes throughout the day. It's really hard to make a valid app if the API is so inconsistent. I can't use gpt-4 because the prices are absurd.

Anyone else experiencing these problems with 3.5?

[neon trench]

yes, same here. some calls take minutes to return, some take less than 1 sec. very inconsistent.
The other day, they announced they were being ddos'd or something; maybe that's true, maybe they are just riding their capacities too close to the edge...

[signal forge]

I added 250 dollars credit, which put me in tier 4. It's way better since I did this

[neon trench]

I wish I could do that right now 😄

it's probably your 'tier4' query that skips the queue and the reason I have to wait so long 😄

[signal forge]

Yeah, if you want to build a production level chatbot it's really needed it seems

I couldn't ship that to clients when it stops working half of the time

And I'm gonna spend 250+ anyway if it becomes a product

So might as well get it out of the way

[neon trench]

I am building a local html, that lets anyone enter their own keys, so I don't really need the access myself to ship the product... but it would be great for testing right now.

[signal forge]

That's cool. Good luck with that

[graceful quail]

@signal forge quick question, does the $250 helps resolve the limit of 50 messages/3h too?

[signal forge]

You mean on the newest got 4 turbo?

Not that I know, but I haven't really tested it so far, that model is too expensive for what I want to do.

I would love to though, it's way better at faking a personality, but I would have to charge every user around 30 dollars / month instead of 5

[neon trench]

to increase rate limit you'd have to pay per query.

[mint fiber]

Anyone had luck getting custom GPTs talking to with APIs in GraphQL in the ‘Actions’? I can get through to top-level objects but I seem to hit a wall with not being able to fetch any details more than one level down.

[merry palm]

Hello everyone, can you help me with Python code for a discord bot? The bot then becomes intergrated with a gpt4 vison api. That you can chat in Discord with GPT 4 Vision, so with pictures. Can someone send me a DM or get in touch to get this working?

@flat drift Can yoy please help?

[flat drift]

https://platform.openai.com/docs/api-reference/chat/create

[merry palm]

I know but this don't have the full intergation with discord or something like that, can you please give me a bit of code from the trenkt bot? That i can use?

[flat drift]

there's a code sample on the page I linked you to

[merry palm]

i know but how can you intergate that image that you send in discord to a url that chatgpt can use

[merry palm]

.

[merry palm]

How do you intergarte that image in to the api

[flat drift]

ask GPT to teach you how to use Python. you have the best teacher in the world at your fingertips so apply this

[merry palm]

Sure

[merry palm]

I can do a bit of python but i don't know how to use discord in python

[flat drift]

gpt can help you with that

[spring aurora]

Just tried ChatGPT for the first time after putting it off for so long. 🤡 I am absolutely amazed by how easy I can find answers compared to regular search.

[spring aurora]

hey uhhh. im a bit of an idiot and i dont exactly know how to use the new api thingy.

apparently im supposed to do this

from openai import OpenAI
client = OpenAI()

instead of openai.api_key = config_data['open_ai']

and it really doesnt say where im supposed to put my API key on the documentation.

[neon trench]

client = OpenAI(api_key = 'YOURKEYHERE')

When you instantiate OpenAI (that's a class) you can pass certain parameters, one of which is "api_key".
This is absolutely in the documentation.

[spring aurora]

thanks. i didnt see it

[spring aurora]

I need help on hwo to integrate my chatgptitno voiceflow pls dm

[viscid dust]

Anyone working on (or know of) a project combining OpenAI's API with Twilio's for automated texting purposes?

[fallen solar]

I have an idea for a chat gpt program for discord, but I don't know how to go about it.

Basically I want to have a conversation with someone on discord on this personal account, but I want it with chatgpt so I can just leave it and take a walk. Maybe with a random response rate so it's not an instant reply, then just sorta have the "typing..." Notification for a bit

[spring aurora]

Ask chat gpt if it can help you create it

Big brain.

[remote finch]

How do i create my own chatGPT4 prompt?

[spring aurora]

you type it out

with your fingers

or voice

[spring aurora]

ask chatgpt

[remote finch]

oh just want to learn ... so i can think of an idea and basically want to voice to text and chapt gpt response as voice.

So basically i probably need Text to voice program that sound like normal human. May be it also need training with voice.

i checked out few ..most are robotic and do not sound like humans

i will ask openai to make this part chatgpt and this way they will have good quality voice

[spring aurora]

silero

https://github.com/snakers4/silero-models

[remote finch]

i tested few voice software that came in google. only voice out of 100s of voice is good to my ear

Thought some big company tackle voice and constantly change the sound based on feedback.

==== here what chat gpt says...

You've got a really intriguing idea there! Wouldn’t it be great to have a text-to-voice AI feature that’s as versatile as a chameleon, being able to switch between accents and professional tones as needed? While I can't directly relay messages to the OpenAI team or set wheels in motion for new ventures, your vision is quite aligned with the direction in which AI and voice synthesis technologies are developing.

The technology for creating diverse and natural-sounding text-to-speech (TTS) voices has been advancing rapidly. You might be interested to know that even now, there are some really sophisticated TTS systems that offer a range of voices with different accents and styles. These systems are based on deep learning models which have been trained on large datasets of spoken language from speakers all around the world.

Despite the improvements, it's true that finding the perfect TTS tool can be challenging - there's such a variety in personal preference and application needs. It sounds like you've spotted a gap in the market where there's demand for high-quality, customizable voices tailored for specific professional uses or accents, possibly all under one roof for ease of access and use.

Here’s a theoretical sketch of how OpenAI or a similar company might approach creating such a service:

Voice Synthesis: Develop a deep learning model that is capable of synthesizing speech that sounds natural and can be modulated to affect different accents, intonations, and speaking styles.

Data Collection: Pull together a vast and diverse dataset of voices covering various accents, ages, and professional jargons to train this model.

User Interface: Create a user-friendly interface that allows users to select and customize voice parameters easily, possibly even allowing them to upload examples of the type of voice they want the AI to mimic.

Integration and Accessibility: Ensure that this service can easily integrate with various applications

===== chat gpt babbled===

[spring aurora]

use silero

it has 100 english voices

like 6 russian

2 german

a bunch of other ones

[spring aurora]

Anyone know how I can tell my AI not to use a tool ON EVERY FUCKING MESSAGE

[drowsy scarab]

I keep getting network errors if the code is too long so annoying

Trying to make an excel sheet data organizer

[tender whale]

if i have a application that exports a file every 15 minutes to an sftp server, how can i point chatgpt to it so that its always referencing the latest file? I dont want to have to upload a new file every 15 minutes

[serene sluice]

from openai import OpenAI
client = OpenAI()

[tender whale]

hmmm, so if theres a file sitting on a server somewhere, i need chatgpt to reference it, or import it. automatically, every 15 mins

so that i dont have to drag and drop teh file into chatgpt every 15 minutes or whenever i want the latest file

[cursive vessel]

Hey everyone! 👋 I'm Patrick - Excited to join this community. I'm the developer of a no-code platform and I've been integrating ChatGPT into my product for over six months now.

I've been using Function Calling a lot to wrangle ChatGPT responses into proper JSON (which was improved in the most recent response_format=json release 🙏 .

I got tired of manually writing the function calling json structure, so I built a really simple one-page free tool to convert your own JSON structure into the ChatGPT Function Calling format instantly. - https://jsontochatgpt.com/

You'll still want to write your own field descriptions, but this kind of jumpstarts the process, especially for complex / highly nested data structures.

Excited to jam with people in this community on how to get better results from the API.

[molten olive]

Are there any technical / other resources that have helped people get Custom GPTs to function? No matter what I try, it runs into a lot of search-related errors. It can't find the text in uploaded .txt file; it also could not find the relevant text when I split a longer .txt file into 10 shorter .txt files. I'm not sure if I'm doing something wrong or it's just buggy as of now. Does anyone have tips or other thoughts, or even general resources where people are discussing/working on GPTs? I'm trying to gain understanding/learn.

[leaden sapphire]

What about YAML? It tokenizes better and supports comments for cot reasoning

https://betterprogramming.pub/yaml-vs-json-which-is-more-efficient-for-language-models-5bc11dd0f6df

[proven crypt]

Hi everybody, I have developed a tool for llama.cpp to allow local AIs to call functions. It generate a GBNF grammar to restrict the output of the AI to respond in a certain way. You can define the functions the AI can use in Python with parameters and it generates the right grammar to constraint the AI to a JSON format suitable for the functions.
It also has a class for executing the actual function call.
You can find the code here:
https://github.com/Maximilian-Winter/llama_cpp_function_calling

[cursive vessel]

Interesting! I'll have to read up on this 😄

[steady mauve]

anyone that uses Airtable a lot (and its API) this is cool

[fringe lion]

include a YAML to JSON converter xd

[meager wagon]

Hi there, does anyone know if one can chat via api to a custom created ChatGPT agent, in short created a custom agent(GPT) in my plus account, want to access it via API. Any help welcome

[trim oyster]

Hey all! I need your help in validating an idea that I have been working on since 2 months

I have created an automatic documentation generator using GPT(Snorkell.ai). Initially it was an hobby project, but eventually, I created a Github App out of it.

To summarize: Snorkell automatically generates comprehensive documentation for all Classes and Functions in your project. It operates in background, for every update sent to your Github Repo, Snorkell.ai creates a PR containing the updated documentation.

Thank you!!

[alpine root]

OpenML Guide: Embracing Open Source and Free Resources, Offering a Wealth of Books, Courses, Papers, Guides, Articles, Tutorials, Notebooks, AI Field Advancements, and Beyond.

https://github.com/severus27/OpenML-Guide

[serene beacon]

Hey everyone! I've been trying to figure if it's possible to create my own filter for Chat GPT 3.5 Turbo. I essentially want to have the program respond only to prompts within a certain field, while configuring a response to reject prompts that don't meet the criteria. Is that achievable? I've read over the Moderation API information, but may not be understanding the potential to implement my own moderation.

[dusky stream]

Would appreciate a code review for my implementation of open ai assistants api: https://github.com/calapsss/assistants-api-easy

i also made a video: https://youtu.be/ZjpNx8qNnaA?si=OwwxRlrzMDgOjFZY

[spring aurora]

where did I find that thumbnail picture 😂

[tight dirge]

Hello everyone. Do not think about good or bad practices please. I get this error, when testing an endpoint as an action within a GPT:
"[debug] Calling HTTP endpoint
{ "domain": "DOMAINNAME.COM", "method": "post", "path": "/create_table", "operation": "createTable", "operation_hash": "6530f0658d303cfbaaddc98e1f512412412412d27c028ad114", "is_consequential": true, "params": { "table_name": "project_data", "columns": [ "id INT AUTO_INCREMENT PRIMARY KEY", "project_name VARCHAR(255)", "project_manager VARCHAR(255)", "ai_expert VARCHAR(255)", "programming_expert VARCHAR(255)", "brainstorming_expert VARCHAR(255)", "status VARCHAR(50)", "start_date DATE", "end_date DATE" ] } }

[debug] Response received
{}

Error talking to
I attempted to call the DOMAINNAME.COM API to create a table with the specified structure, but encountered a client response error. This might be due to a temporary issue with the API service. You might want to try the operation again later or check if there are any issues with the API service."

When using the above schema. If I use Postman to call the API manually, it works fine. The format of the json file is fine. Code is fine. I do not know what it is, and I've spend the last 4 hours trying to solve it.

I need some ideas.. Anyone have anything?

My REST API is running on Replit, if it helps any. Sorry for spam, I am just frustrated. I really want to get this to work 🙂

[fickle scroll]

anyone willing to run a test for me? I don't have an api key but I think ChatGTP works in my software now. https://github.com/aseichter2007/ClipboardConqueror

[fickle scroll]

It's not a cors problem? that's what I am expecting to have problems with I've only tested against local endpoints.

[tight dirge]

I figured it out. It was my OpenAPI schema that were wrong - I just spent too long looking at the problem, you know what I mean 😂

I appreciate the answer though!

[fickle scroll]

I feel that man. Good work. Will you check me out? I don't have a key to test with. My software is pretty cool, you might be stoked. It's an extremely broad use case. It's designed for prompt engineering.

[tight dirge]

Will do later. I have exams soon 😅

[fickle scroll]

No worries, thank you!

[slate current]

Hi everyone, I created a chatbot on Openai and I have a bubble icon, what is it ?

[tight dirge]

Tbh, this is a really interesting project. I just do not know JavaScript, which is sad. I would've forked it, really great concept

[fickle scroll]

Thanks man. It should be ready to run, just put your key in, run everything, and copy |||##| hey chatgpt.

https://youtu.be/fcM8dDtVTYQ

[tight dirge]

Someone (you) using your GPT

[fickle scroll]

I should take the key at runtime instead maybe but then it will be even more cryptic

[fickle scroll]

the recommended model is good at javascript, what do you like, python?

• |||What are the major syntactical differences or challenges between python and javascript?

Ahoy there! Captain Clip here. Let me see if I can assist ye with this one. I may not be able to communicate with the same fluidity as those land-bound Earthlings, but I can certainly give ye a brief rundown on the differences between Python and JavaScript.

Python and JavaScript have their own unique syntax and quirks. Here are some key distinctions:

1. Syntax: Python follows a more structured, English-like syntax with clear indentation rules, while JavaScript is more flexible with curly braces {}.
2. Type Hinting: Python has explicit type hinting, while JavaScript relies on dynamic typing.
3. Error Handling: Python uses exceptions, while JavaScript uses try/catch blocks.
4. Asynchronous programming: JavaScript heavily utilizes callbacks and Promises, whereas Python leans towards async/await.
5. Data Types: JavaScript has native support for types like null and undefined, while Python doesn't.
6. OOP: Python has classes and JavaScript has prototypes.
7. Scoping: Python uses LEGB (Local, Enclosed, Global, Built-in), whereas JavaScript uses Lexical.|||explain Lexical in this context:
8. Lambda functions: JavaScript has anonymous arrow functions, while Python has lambda functions.
9. Python supports multiple inheritances, but JavaScript doesn't.

Now, I must warn ye, I've spent me days amongst the stars and not in a typical programming environment, so some of this might be a bit rusty. But, these are the general differences between the two. Ye can find more info on 'em with a search on 'yer trusty ship's terminal if ye need.

[jaunty crow]

I'm exploring ways to create automation workflows that integrate ChatGPT with high-level scripts, including advanced libraries like NumPy and Pandas, and database integrations. While Zapier is useful, its Python integration isn't suitable for this level of complexity.

Here are my key questions:

Alternative Integration Platforms: Are there alternative apps or platforms similar to Zapier that can seamlessly integrate high-level scripts, database connections, and ChatGPT for advanced automation workflows?

Zapier Integration with Advanced Tools: Is it possible to integrate these advanced components into Zapier, leveraging its 6000+ app integrations, to achieve the described level of complexity? If so, what would be the best approach or resources for this?

If you have insights know of a platform that fits the bill or have ideas on integrating advanced tools with Zapier, please share your thoughts.

[charred remnant]

anybody here have experience with the API?

I am trying to make a simple request to the API and its telling me I've already reached my quota? I have never used it before lol. Am on the free tier 🤷‍♂️

[jovial folio]

u ran out of ur free trial

[fickle scroll]

how come his says no quota and mine says "slow down"?

[oak pagoda]

Does anyone know of a good way to make 2 Assistants communicate with each other in 1 thread?

I can get by by human-centipeding 2 threads together but it doesn't feel right

[spring aurora]

where?

[oak pagoda]

Assistants API

[sly sapphire]

Is it possible to ask questions about a specific respository? Either local or on github. I'm hoping to ask about how unfamiliar codebases work and how to make certain changes.

[oak pagoda]

Copilot does this very well now

[patent atlas]

Hi there!

I'm working on a React library with an API to build with LLMs.

We (2 friends and I) are trying to integrate all the tools needed in one integrated open source thing so it's much faster and simpler to build.

But it's not there yet & we're trying to figure out what's missing to be really useful. We would love if you could check it out and give us feedback!

landing and dashboard: https://www.polyfire.com

SDK: https://github.com/polyfire-ai/polyfire-js

API: https://github.com/polyfire-ai/polyfire-api

docs: https://docs.polyfire.com

[spring aurora]

Hey guys I don't know the best place to ask (if there's a better discord for app builders PLEASE point me in the right direction) this but i'm working on a web-app backend and I want to be able make and edit images. (e.g. extending a user-uploaded image), my question is how the backend should manage images.

1. User uploads image to my backend server
2. Backend server makes api request to an ai that modifies images
3. Backend stores the image (e.g. uploading it to google cloud storage)

I'm confused about step 1: what's a good way to manage a cache of user uploaded images? Just a random folder in the server? An in-memory cache like redis? Or something else?

[subtle pagoda]

does anyone use IntelliJ?

[dawn thicket]

Not sure where else tp put this: https://chat.openai.com/g/g-FO5lGlHiB-wizard-apprentice

[reef furnace]

Hey all- I’m trying to create an AI agent/chat bot/wrappers that looks at all of my databases and technical guides to answer specific questions. Keep in mind I’m novice- non tech guy and trying looking for some help on getting started. I already have access to GpT4. What softwares would I need?

[carmine trench]

#1173672346729578567

[dawn thicket]

Thanks!

[charred ether]

bro everytime i ask gpt4 to code something it gives me nothing but errors the code capabilities are dog piss

simple javascript

is there a better customgpt or something

it knows as much coding as a first year college student lmao

30$ for this shit ima need refund

@staff message me

this is the extent of its coding skill so far

[shy whale]

    extra_window = tk.Toplevel()
    extra_window.title('làm sao để bé iu của anh vui đây (ㆆ_ㆆ)')
    extra_window.geometry('600x200')
    ttk.Label(extra_window, text = 'có gì bé tâm sự với anh nè ᕙ(`▿´)ᕗ').pack()
    ttk.Button(extra_window, text = 'E thấy ổn hơn rồi. bé iu anh (っ◔◡◔)っ ❤',  command=extra_window.destroy,).pack(side=tk.LEFT,pady=10)
    


def button1_clicked():
    label.config(text="uhmm, anh rất hạnh phúc khi bé iu của anh vui ʕ•́ᴥ•̀ʔっ. ")

def button2_clicked():
    label.config(text="ơ tại sao lại như thế, , cố găng lên ≧◠ᴥ◠≦:fist:")
    
def combined_command():
     create_window()
     button2_clicked()
# Create the main window
def update_label(label_var, new_text):
    label_var.set(new_text)

root = tk.Tk()
root.title("Quà tặng của anh giành cho em.")

# Create a label to display messages
label = tk.Label(root, text="Em có đang vui ko")
label.pack(pady=10)

# Set the size of the window
root.geometry("420x200")


# Create Button 1
button1 = tk.Button(root, text="Có, em đang rất vui", command=button1_clicked)
button1.pack(side=tk.LEFT, padx=10)

# Create Button 2
button2 = tk.Button(root, text="Hong, em bây giờ em buồn lắm", command=combined_command)
button2.pack(side=tk.RIGHT, padx=10)

# Start the main loop
root.mainloop()```

can anyone pls help figure out how to call button1_clicked for ttk.button as well as new_window.destroy ?

[spring aurora]

привет

[digital cliff]

How many times are u going to spam that?

You've posted it atleast 3 times in different channels which aren't "looking for work" channels 🙄

[sonic glacier]

Does anyone know if it's possible to access Custom GPTs via an API? Would love to be able to display/embed them within my own website.

[rose grove]

This is not currently possible and there are no official plans to support access to GPT’s from an API call. GPTs are, at least currently, purely for use within the ChatGPT ecosystem.

-OpenAI Forums

[sonic glacier]

Thanks, have just seen that assistants can probably do what I'm asking for anyway

[merry totem]

how to get script for bank robbery

[spring aurora]

https://howtorobbank.com

[blazing otter]

hey all, has anyone tried the tools parameter on openai api and what do you usually it for? I'm using next.js and try to make it work to get strict json output but has been encountering errors whenever i use the tools parameter.

[rose grove]

---

The "tools" parameter in the OpenAI API is used to describe the functions that the AI model can call. It is an array that contains one object per function. Each object has two keys: type and function. The function key has three subkeys: name, description, and parameters.

Here is an example of how to define the "tools" parameter:

const tools = [
 {
   type: "function",
   function: {
     name: "getCurrentWeather",
     description: "Get the current weather in a given location",
     parameters: {
       type: "object",
       properties: {
         latitude: {
           type: "string",
         },
         longitude: {
           type: "string",
         },
       },
       required: ["longitude", "latitude"],
     },
   }
 },
 {
   type: "function",
   function: {
     name: "getLocation",
     description: "Get the user's location based on their IP address",
     parameters: {
       type: "object",
       properties: {},
     },
   }
 },
];

In this example, there are two functions defined: getCurrentWeather and getLocation. Each function has a description and parameters. The parameters are defined as an object with properties and required fields.

The "tools" parameter has replaced the deprecated function_call and functions parameters. However, as of the current state, the fine-tuning API still requires the legacy format.

For more information, you can refer to the OpenAI documentation (https://platform.openai.com/docs/api-reference) and the OpenAI community forum (https://community.openai.com/t/499313).

(This response is by an FAQ bot, not me)

[blazing otter]

Thanks. How do you get the answer from the FAQ bot?

[rose grove]

OpenAI's official Discord server has kapa.ai, and it can answer general questions about documentation

[blazing otter]

That's cool. Thank you!

[cosmic thorn]

Hey guys!
I am currently trying to train an IA for a potential client of mine. If someone is interested dm me 🙂
I am struggling in the trainging aspect.

[slim belfry]

This is programming channel, that should go in #1173672346729578567

[queen storm]

Alright- figured it was helpful here, but will move it!

[spring aurora]

Hey guys if I had a lot of documents (like 200 40-page shipping contracts), and wanted to "finetune" GPT4 so it can create a document in a similar style, is this possible with embeddings or something? I know we don't have finetune access to gpt4 yet, so is there some alternative we can do?

[carmine trench]

yeah go with embeddings

[terse ember]

does anyone know about discords that focus on coding tools? I use Github Copilot. strangely little discussion on reddit and I'm old and out of loop on discords in general, "back in my day we had forums"

[sour crest]

Hello! If I was looking for people to help test out a site I am working on using the OpenAI tools would it be ok to post here?

[slim belfry]

#✨projects

[sour crest]

Thanks!

[molten scaffold]

Hello 🙂 I was wondering if there is any place where there are resources of LLMs for beginners

[spring aurora]

🤗

huggingface

hf.co

[molten scaffold]

how did I forget this exists

[opaque stag]

askfsdl have some decent lectures. coursera ml fundamentals, 3blue1brown linear algebra and neural networks

[dreamy hornet]

what are some gud subreddits to keep tabs on tech world?

[storm relic]

nvm, got it working with 3.0.0 specifications...

[fierce cypress]

Hey, can someone please help me here? I want to use the TextGen UI and it has OpenAI API support, i've managed to use the Web UI correctly but i would like to use the OPEN AI with python. Do i need to use requests? Or does the official OpenAI API Library for python work with this?

[spring aurora]

r/chatgpt

[turbid atlas]

Yep

[molten scaffold]

nice

[spring aurora]

If you have any coding issues, text me. I might not be a mod, but i am good at programming!

[oak widget]

Can anyone help me with a midjourney task with help of GPT > I want to use gpT to spin up many (say 1k-5k prompts for midjourney that are similar

[spring aurora]

I’ve been having major issues trying to refine and update my GPTs for the store. Anyone else? Is this the norm?

[slim belfry]

over 11,000 online members, some of them probably are named Dan

[molten scaffold]

DAMN

[charred tendon]

What languages?

[broken maple]

Random question. Has anyone figured out how to protect files for a custom GPT from being prompt injected and downloaded? I was thinking an API call to the files behind the scenes of a GPT would work but that seems quite labor intensive. Any suggestoins wold be greatly appreciated!

[chrome locust]

it`s unavailable for now.

[slender holly]

how to add raylib pkg-config

[chilly valve]

Is it possible to control the size of the chunk in this example? Because when I run this code, it outputs one word at a time.

stream = client.chat.completions.create( model="gpt-3.5-turbo", messages=[{"role": "user", "content": "Say this is a test"}], stream=True, ) for chunk in stream: if chunk.choices[0].delta.content is not None: print(chunk.choices[0].delta.content, end="")

[limpid sapphire]

Does anyone know how to stop creating an assistant? i am trying to use assistant in python. However, i keep creating a new assistant over and over again. Do i have to use like jupyter notebook to prevent this problem?

[spring aurora]

use message history

[spring aurora]

no it outputs 1 token every time

[jaunty rain]

This is not a Discord server to distribute pirated content. Don’t do it again @chrome python .

[near badger]

how can we create a chat gpt that controls a pc and does the job based on prompts?

[slim belfry]

GPT famously makes a lot of mistakes when it comes to commands for computing, why would you want this

[quaint holly]

if you don't want that just don't use stream at all? GPT3.5 is fast so generally that should be ok depending on your use case.

[quick lance]

So I am really stumped, im new to programming but im following videos and reading the documentation but I can never get past the error that the model im trying to use doesnt exist. Ive used every model listed with variations, I know im doing something wrong but i cant seem to get a clear answer as to what it is from anything

the start of my code is looking like this "from openai import OpenAI

assistant = client.beta.assistants.create(
name = "Laz",
instructions = "You are a math tutor",
tools = [{"type": "code_interpreter"}],
model = "gpt-4-1106-preview"
)
thread = client.beta.threads.create()
print(thread)"

and I get this in my terminal " File "C:\Users\Megah\Coding Projects\Laz\Lib\site-packages\openai_base_client.py", line 980, in _request
raise self._make_status_error_from_response(err.response) from None
openai.BadRequestError: Error code: 400 - {'error': {'message': "The requested model 'gpt-4-1106-preview' does not exist.", 'type': 'invalid_request_error', 'param': 'model', 'code': 'model_not_found'}}"

[slim belfry]

I haven't tried making assistants but why don't you try putting your code+error into regular gpt and asking it to help you diagnose?