community-help

[wraith crown]

so I guess I just need to open Multiple OpenAI accounts, and load balance the API keys haha 🙂

happy to pay for usage

[fallow snow]

For me it’s 10 reserved for the gpt maker

And 40 ish regular

[timber shadow]

Maybe it's a shared cap somehow? Like if you've used 20 of your 40 GPT-4, maybe that "counts" as like 10 of your custom GPT 25?? No idea at all! Somehow different platforms/devices seem affected differently too. Sometimes I'll get a "no" on PC/web but a "yes" from smartphone/app or smartphone/web.

[proven saddle]

if each message to the gpt creator and for testing a custom gpt counts as 1 message you only really get 20 messages for making a gpt and 20 for testing it which quickly runs out. not sure if that's how it is but it feel like it

[fallow snow]

I guess it’s a mess rn

What’s a good way for it to keep a consistent style and format for gpt?

[proven saddle]

if a change made by the gpt creator counts as 1 as well, you get like 10 changes and 20 messages to the custom gpt before it runs out

[timber shadow]

"Create" messages definitely count as (at least) one. Here's a list of things that all count for (again, at least) one: #community-help message

[fallow snow]

That’s why it’s helpful to try and do everything in one create message

Glad they let us do it now

[timber shadow]

And then to abandon it and use only Configure from there on out 😁😁

[proven saddle]

ah i see, thx 😄 then it's definitely clear why the cap is reached so quickly

if i try really hard i can reach it in 5 minutes

[timber shadow]

Oof that 2hr55min wait must be brutal 😁

[fallow snow]

Help?

[timber shadow]

Maybe share what you've got, and what your goal is, in somewhere like #prompt-engineering?

[proven saddle]

it really is especially if you're using it for programming and you're left with a half finished monstrosity where none of the code is working and you can't guess what chatgpt meant to type for the rest of the code 😂

[gritty anvil]

Oh no I understand, the sensitive part, tracking.

How would I get a custom one to stay on task, what particular instructions would I have to give it

[timber shadow]

What task(s) in particular are you trying to get it to stay focused on? A specific example of what you've got, and what your goal is, might help in this case!

[gritty anvil]

Say help a person get employment, do grant work, or be a daily teaching assistant

[timber shadow]

There's really no "one size fits all" stay-on-task instruction that would apply for all of those. The best way is to start making a GPT with a "Create" instruction like "Create a GPT that helps me find gainful employment" and then continue the setup process from there. When you're done with that, check out the "Configure" tab and see how you like the instructions it wrote for you. Edit them to your liking, then test it. If you're not getting desirable results, you might consider editing the instructions in "Configure" manually to be more to your liking.

Or, you can share here what you have, and what your goal is, and folks may be able to point you in the right direction. When it comes to prompting an LLM, the more specific examples you can provide in the way of "here's what I have" and "here's what I want", the better!

[gritty anvil]

Oh no those are 3 different examples but I get what you are saying, give each GPT a particular task and it will stay on track with the task alone that it's given.

Sweet

[timber shadow]

Precisely!

[gritty anvil]

Okay awesome, now to figure out how to find this conversation again lol

[glad bay]

Anyone else experience losing plus. One minute I was using it. Next moment I am in the free plan

[charred parcel]

has your subscription expired?

[winged nacelle]

this is a good reasoning, but i have a question. This also work for roles?

i mean, i plan to do a lot of things on the same bunch of data, is better to produce various GPT from the same knowledge base, all with a SINGLE exact job, or can he do related job?
Example: understand how a TTRPG works to help me as an author AND be a social media manager for it

[timber shadow]

That's a great question, and I'm not sure that there's a clear-cut yes/no answer for you in one direction or another. My instinct is to say that if your knowledge base already includes a complex data set encompassing the rules and context for a TTRPG, you might be best served creating different GPTs to interact with that knowledge base in different ways. However, GPT is flexible by nature, so it's also entirely possible that you could have just a base "knowledge-files-only" GPT for the TTRPG, and then handle the "roles" you mention with simple text prompts using the GPT. In other words, instead of having one main knowledge base and then a plethora of "roles" added to the GPT as well, you could just have the knowledge base, and then interact with it as you see fit from there, with carefully-worded questions. I don't mean to give you too much of a non-answer here, so I'll pause here and ask you what you think!

[north granite]

Hi, Why can't I create an account from vietnam?

[proven saddle]

i gotta find a way to scrape the web for articles, get the title, make a description based on a specific number of characters, find the author and add the website link.
i don't know how to do this easily, does anyone have any suggestions? right now i'm adding it manually 💀

[timber shadow]

Hello! Vietnam appears in this help article as a supported country: https://help.openai.com/en/articles/7947663-chatgpt-supported-countries
What problem are you running into when you try to create an account?

[tired pasture]

Hello!
If somebody have some tipps, I would take a hint as a beginner with no real coding skills:

I want to use Slack to ask questions to my Custom GPT, to get answers from a website and knowledge files I provide, posted back into Slack. I use Zapier so far to solve the communication between Slack and OpenAI.

Problem:

1. I don´t know/find, how I can directly send my questions to my Custom GPT. Just to general ChatGPT or an Assistant via the API
2. I can provide knowledge files to the assistant, but I don´t have the option there to easily tell it, to retrieve information from the website I also need in.

How can I solve this in the best and cost effective way?

Another general questions - assistant and CustomGPT - will they learn, when I communicate with them, e.g. if an answer is just semi-correct and I just tell it, how it would be correct?
Thanks a lot!

[hushed terrace]

Are GPT plugins still relevant? Do they have use cases still or ?

[winged ridge]

Error !Unterminated string in JSON at position 548 (line 1 column 549)

[abstract ice]

Interest Form: Higher Usage Cap for GPT-4
Thanks for your interest in GPT-4!

To give every Plus user a chance to try the model, we’re currently dynamically adjusting usage caps for GPT-4 as we learn more about demand and system performance.

We're also actively exploring ways for ChatGPT Plus subscribers to use GPT-4 in a less constrained manner; this may be in the form of a new subscription level for higher-level GPT-4 usage, or something else.

Why am I getting this after generating some stickers?

[strange grotto]

Can someone help regarding enterprise sales?
We're a mid side corporation located in VA, USA.
We've already contacted sales (through the webform) but without any contact back.

[feral jasper]

bruh

[humble grove]

Is there any document / help on how to train GPTs which are built using the GPT builders ? Would love if someone can help me with this

[timber shadow]

Hello! What kinds of questions do you have? Here's a basic FAQ: https://help.openai.com/en/articles/8554397-creating-a-gpt
In general: GPTs operate based on the three main parts in the "Configure" tab in the builder: instructions + knowledge files + actions.

[winged nacelle]

ill try this

[cosmic crag]

How long ago did you contact sales?

[azure hare]

does tensorflow work on AMD cpu's

[cosmic crag]

yes

[azure hare]

even though it says

[cosmic crag]

Go to google and search 'tensorflow amd'. You will find a page on amd's website on tensorflow

[azure hare]

im not on linux

so i cant use docker

[cosmic crag]

You can use docker in other OS's, not just linux

[azure hare]

im trying to build and run gpt-2

[azure hare]

i give up
need to install redistributable packages and stuff

wont do that rn

is there a buildable gpt-3, there is a repository that says "gpt-3" but it doesnt seem buildable

[cosmic crag]

Look for open source models. I recommend gpt neo x, but hugging face is chock full of open source models for you to build

OpenAI is more ClosedAI nowadays

[azure hare]

isnt huggingface for image generation
i just need text

[azure hare]

you can thank bill gates for that

is there a difference between git-lsf and git-scm

nvm im dumb

[hushed ivy]

I still wish for this 😆

[tribal gull]

Yesterday, my network was very stuck and I couldn't open the gpt interface, so I kept refreshing, trying to open the gpt interface, but it later showed that I was a network attacker, so I still can't enter the gpt interface on this computer, what should I do

[timber shadow]

Hello! Do you have a VPN installed?

[tribal gull]

that's right😢

During that time, my network was very unstable

[timber shadow]

Are you located in one of the countries listed in this article? https://help.openai.com/en/articles/7947663-chatgpt-supported-countries
If so, can you try disabling your VPN, and then trying again? If this doesn't work, you might need to disable an additional feature called something like "Web Protection" in your VPN as well, separately from disabling the VPN itself. It might be called something different like "NetShield" or "Threat Protection" or something like that, depending on your VPN.

[tribal gull]

oh,no,i don't think i'm on this list.Is there any other way?I really need it😣

[timber shadow]

Unfortunately, I don't think so, I'm really sorry! Trying otherwise may count as circumvention depending on how they monitor use and traffic, and I'm just not sure how that works well enough to advise you!

[tribal gull]

well,thanks😭

[foggy musk]

does editing a previous message count towards the 40 message limit in 4.0?

[timber shadow]

Yes it does, here's a list of all the things that count as one: #community-help message

Note there's been quite a bit of inconsistency reported lately regarding GPT-4 usage caps in custom GPTs--they seem to affect the usage cap differently and almost separately from the "normal" GPT-4 cap.

[foggy musk]

ty

[ember pelican]

Hi I want to start playing with custom gpts and I wanted to know more about

1. cleaning the data, plan to upload pdfs and such and want to know if I should literally drop it or spend a little time to fix up the files to a specfic format so the model can understand better

2. limits in file sizes/context

[zealous crane]

one of my most important conversations disappeared from an errant (but unknown) keyboard stroke while i was typing in the conversation. It looks like it was somehow deleted., how can i recover the deleted conversation?

[cursive folio]

https://community.openai.com/t/gpt-builder-bug-custom-profile-picture-lost-when-publishing-publically/498387

Have you encountered this malfunction?

[abstract ice]

why do I keep getting rate limited?

i didn't even request that many stickers 😦

[glossy hawk]

Anyone else having this issue right now, can't do anything - Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

Is OpenAi down?

Been 2 hours now and even from a new chat, I cannot do anything.

[bronze charm]

Looks good here https://status.openai.com/

Working upon prompt as well

[glossy hawk]

Thank you, that's so strange, woke up this morning, tried my first chat with GPT 4 and just get this message - Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

[bronze charm]

This can be fixed by clearing the browser cache, removing extensions that manipulate ChatGPT in some way, or by using a non-chromium browser like firefox as a nuclear option

Though I cannot guarantee efficacy

[glossy hawk]

Yea guess I should have tried that first, let me see how it goes, thanks for the tips.

Tried all of the above and still nothing? I wonder if different countries have access to OpenAi on different servers? If there is anyone here in South Africa, please can you confirm if GPT 4 works on your side, I do have a Plus account and made sure my account is in good order.

[bronze charm]

Are you using some sort of VPN?

If so I would disable it

[glossy hawk]

Issue Resolved: So I installed AVG yesterday and didn't use GPT after that, AVG also installed the Anti Track feature, and that was causing the problem, uninstalled Anti Track and now work perfectly. Thank for trying to help me out @bronze charm really appreciate it.

[bronze charm]

Let's get prompting, shall we?

[glossy hawk]

😅 yip now I can get back to it again

[strange grotto]

hi @cosmic crag thanks for the reply! We've contacted them I believe since beg. of last week.

[limpid wadi]

Hey everyone!
Can somebody tell me how I can use ChatGPT 4 in my API request? Like this?

"model": "gpt-4"?

[trail cairn]

yes

[limpid wadi]

Thanks!

[still epoch]

Hello, everybody. Does anyone know how to integrate custom gpts into Instagram

[trail cairn]

you would have to create one with api and pay for the usage

[onyx quarry]

So, for a week I have been getting this error message. Any idea why? Covers 3.5 and 4. Essentially can't do anything within the ChatGPT system.

[rugged wing]

Clear your cookies logout and login again

And check if you have the same issue

But it is common since ChatGPT have problems sometimes, but not always

[onyx quarry]

Will try - thanks

Also, no joy. Same issue.

Was willing to wait, assuming there was some system-scaling issue with demand, but now a long-lived problem.

(and same issue on Chrome)

[onyx quarry]

Thanks. Alas it didn't help.

[slow bison]

its your prompt

can you share it over here? some topics cannot be addressed by the LLM and it refuses to waste time on those

[onyx quarry]

It's any prompt. I've taken to asking it 'are you working now?', but it is essentially anything I type into the text box

[slow bison]

lol switch accounts or were you on gpt plus?

[onyx quarry]

(I was on gpt plus)
Interestingly, I jave just tried it on the phone app, and that works... (mystery)

Ok - and ipad works via web. At least I've narrowed it down to my computer, but cross browser platforms on that laptop

[slow bison]

ur on ur own mate

[rigid musk]

Try firefox as test

[onyx quarry]

👍

[rustic marlin]

although file size is not a proble, bigger files will cost you more money, so, it is healthy to condense your data if you can
the token limit o nthe AI is pretty large, 128K token context, which is really a lot

[autumn hull]

i need some help, i am facing some issues with streaming tts responses with node js, can someone help me?

in openai doc, there is a example with python, but my application is running with node js. would love to know how i can stream tts responses with node js.

[rustic marlin]

the exmaples are also available in JS, just select JS at the top right corner of the example box

[autumn hull]

no, for streaming real time audio they only provide python example here, with no option to select to js

[rustic marlin]

true, this is lame

[autumn hull]

yes kind of weird.. saw an example on github but it very weird to handle it. => async function main(input) {
const response = await openai.audio.speech.create({
model: 'tts-1',
voice: 'alloy',
input: input,
});

const stream = response.body;
console.log(stream);

}

main('Hello, my name is John.');

but response.body looks really weird

[winged nacelle]

what exactly are custom Actions?

there is a repository or something?

[timber shadow]

Hello! Do you mean in a custom GPT? GPT actions are like developer tools. If you had, for example, a website with an API that did something, you could set up a GPT to use that API in a chat. There are also public APIs out there, I think there are a couple examples if you explore in the Actions interface

[winged nacelle]

Yeah, i mean in the custom gpt, oh, so it's only to external interactions, ok.

i also hear a lot about plugins, but i have the faintest idea how that works and how to use it, where i can find documentation?

[timber shadow]

Plugins will likely be replaced by GPTs in the long run. I think I heard they're no longer adding new plugins, for example.

Only on the web version, you can go to https://chat.openai.com/#settings/BetaFeatures and enable Plugins. Then, go to a new chat and open up the model selector dropdown in the top-left, and select Plugins. Then you'll see a new box, open that, open the store, and explore!

[midnight estuary]

I need some help getting my actions to work. my OpenAPI spec has beet validated using a swagger tool online, but when i try to use the action via the GPT , it always fails... any ideas why?

[winged nacelle]

OMG, i translate an entire 500 pages pdf.... when there is a plugin for that?
i don't know if i must be proud of how i circumvent the obstacle... or cry over the time i spent!

[midnight estuary]

you are looking for a language translation Chatgpt plugin that accepts PDF input?

[winged nacelle]

no i was looking for a tool to make GPT read effectively my TTRPG manual WITHOUT create an entire TXT repository....

BUT i have already create a TXT repository

i don't understand why today i keep getting so many network error, it's just me?

[midnight estuary]

if you are not doing anything different from what worked for you previously, then the problem is likely to resolve itself without you having to do anything different.

[winged nacelle]

it appears only in long responses, i dunno.
i'm testing the knoledge base

[midnight estuary]

long responses usually encounter context window size-related issues

[winged nacelle]

in other words i push it too far, ok, i'll try with more on target question

a good tought for the next session... becouse....

You've reached the current usage cap for GPT-4, please try again after 5:12 PM. Learn more

[timber shadow]

Hello! Please avoid channel spam by posting the same thing in multiple places. Also, this type of post, if you want to post it, would have probably been best suited for a place like #1070006151938314300 or #1070006915414900886, as these chat-based channels flow quickly, and your message could be missed.

[timid quest]

Is there a location in the site or perhaps here in the server where others have shared their custom gpt’s for others to use?

[naive aurora]

that's exactly why i am posting it in multiple places because after two weeks of this i am tired

[spare lichen]

any updates on GPT Plus subscriptions?

[random sphinx]

Could someone help me with this openAPI schema? I try to copy it but saving the gpt always fails

[midnight estuary]

whats the current status?

[rigid musk]

No eta for now

[timber shadow]

Yes indeed! #1171489862164168774

[cosmic crag]

I think I'd just say that the company was in a bit of turmoil last week and then just celebrating, so its likely they will contact you in the next few days.

[cosmic crag]

That is a very large spec, but my IDE is not picking up anything wrong. I'd say move the auth to the auth box underneath the schema box. Remove it from the spec. The descriptions of the parameters should be full and exact and the examples given must be very clear in what they do.
Next time the gpt does the action, look at what it is sending and ask it what the error is afterwards

Trying to save that action, I can't save it inside of chatgpt. Says 'none is not an allowed value'. Look over the spec and see if there is a weird 'none' value somewhere in there.

[minor geyser]

anyone else ever have an issue where a gpt transcript/conversation shared to you won't open and cannot be deleted from history?

[broken zenith]

I had an exploratory session about Microservices and what the model thought were the correct answers for general interview questions. There was some back-and-forth about what the dev teams should be focused concerned with and what should be delegated to ops. I asked about k8s orchestration and service discovery features and the conversation pivoted to istio. There was substantially more to it, but that's the gist of it. Everything to do with k8s and a few more questions before that are deleted this morning. Is there a way to get it back or prevent this from happening in the future?

[fallen flume]

Guy I recently got the premium version of chat gpt but when I go to beta fetures and turn on plugins and analysis thingy I don’t know when to find them like a I watched multiple videos what do I have to do

[timber shadow]

Hello! Plugins are only on the web version. On chat.openai.com, start a new chat, and open up the model dropdown menu in the top-left corner, and select Plugins!

Advanced Data Analysis is now built-in to the base multimodal GPT-4 model, so no separate selection for that anymore. However, there is an official dedicated Data Analysis GPT, if you want the other modalities disabled for that! https://chat.openai.com/g/g-HMNcP6w7d-data-analysis

[fallen flume]

Il try in a bit

[edgy plinth]

im trying to use the assistant playground but i can't seem to get it to access a csv file

[timber shadow]

Hello @latent moat just pinging you here in the community help channel! What issue are you having with the voice feature?

[timber shadow]

You might try in #dev-chat!

[cosmic crag]

The Assistants API is really buggy still

[edgy plinth]

is the web-based playground considered an api?

[timber shadow]

Indeed, I think all things platform.openai.com are under the API umbrella

[edgy plinth]

cool, thanks

[latent moat]

It tries to connect, but I repeatedly get the Connection failed, tap to retry error

[timber shadow]

Do you have a VPN installed?

[latent moat]

no

[timber shadow]

How about an adblocker app?

[latent moat]

no. it has worked wonderfully in the past. Just started seeing the issue today with no new apps installed. I do have a system update pending, but haven't installed that yet. Samsung Galaxy S21 Ultra on One UI 5.1 and Android 13, before you ask 😄

[exotic crow]

There is an open-source project Open Custom GPT which allows to create a Custom GPT and host/embed on your site

[charred parcel]

that's not supported by OpenAI though

[timber shadow]

Thank you for the additional information! Very odd that it has been working, but isn't anymore...hmm...

Have you tried opening App info, force closing the app, and clearing its cache? Or even uninstalling/reinstalling? And have you tried connecting on your cellular network instead of WiFi, if relevant to you? If not, might help to do that force close/clear cache step before you network swap.

[latent moat]

Tried force stop + clearing cache.
No go on Wi-Fi.
No go on cellular data.
Rebooted, cleared cache.
Working now!
Thanks for the help 🙂

[timber shadow]

Aha! Nice troubleshooting, glad you got it fixed up! Thanks for confirming what worked!

[main shale]

Hello is that the right way to upload the training and validation set and also afterwards starting the fine-tuning process?

def upload_file_for_finetuning(file_path):
uploaded_file = openai.File.create(file=open(file_path, "rb"), purpose='fine-tune')
return uploaded_file['id']

def start_fine_tuning(training_file_id, validation_file_id, model_id):
fine_tuning_job = openai.FineTuning.create(
training_file=training_file_id,
validation_file=validation_file_id,
model=model_id
)
return fine_tuning_job['id']

def main():
model_id = "gpt-3.5-turbo" # Replace with your GPT model
training_file_path = os.path.join("data", "llm_training_data_train.jsonl")
validation_file_path = os.path.join("data", "llm_training_data_validation.jsonl")

# Upload the training and validation files
training_file_id = upload_file_for_finetuning(training_file_path)
validation_file_id = upload_file_for_finetuning(validation_file_path)

# Start fine-tuning with both training and validation files
ft_job_id = start_fine_tuning(training_file_id, validation_file_id, model_id)

# Output the fine-tuning job details
# TODO

if name == "main":
main()

[timber shadow]

Hello! If you're fine tuning a model, you might try asking in #dev-chat!

[main shale]

Yes, I am fine-tuning. Thank you 🙂

[exotic crow]

It works on top of Assistant api though which is from OpenAI itself, so don't see any problem with it

[charred parcel]

it's self promotion which is prohibited on this server

[exotic crow]

It's an open-source project, not a commercial tool but if that is not allowed as well will not do it further

[charred parcel]

please check #server-rules rule 7

[cloud plinth]

My paid subscription ran out and I didn't realize it. I prompted a thread I've been working on which was on GPT4, and it snap-rolled it back to GPT 3.5. Now that I've resubbed, it won't switch that thread back to GPT4, it just keeps dumping me to a new chat. How do I get it back on 4?

[charred parcel]

once a chat switches to 3.5 you cannot go back to 4, you will need to start a new chat

[cloud plinth]

Then it should throw a warning or prevent doing that by accident.
This is a 20k word thread.

[charred parcel]

it won't remember all of those 20K words, the context window is smaller than that.

[cloud plinth]

I've had it recall things I mentioned once in a 200k word thread 🤷🏻‍♂️

[fallow snow]

I’m hoping they would change that

Same with custom instructions

[cloud plinth]

I submitted a bug report. Expected and actual behavior are out of alignment.

[elfin quiver]

URGENT - Has anyone noticed that all chat history in their openai has been deleted? Mine says they couldn't be loaded and I think it was hacked or either a temporary bug, which happens sometimes I'm guesstion!?

[warm geode]

I am not able to sign in using google account. "We ran into an issue while signing you in, please take a break and try again soon."

[torn ether]

Mine also can't be loaded and when I try to login it says it's having an issue authenticating me

[torn ether]

Same here

[hollow sigil]

ok so it's not just me

[elfin quiver]

OK, good to know it's not just me. I'm sure they'll fix whatever issue it is...they're usually pretty good with that.

Hopefully it's not the Russians (or whomever) DDOsing them again...

[vital radish]

there it dies !

[hollow sigil]

I can't log in and the API is unusable.

[hollow sigil]

yet "All Systems Operational"

[warm geode]

Yeah looks like authentication is down.

[elfin quiver]

Sometimes it takes a few minutes for stuff to report the accurate status of servers...know that's been the case with online services indicating whether a website is up/down...

[storm cairn]

you cant do anything

[hollow sigil]

ahh cool, that was just posted

[alpine surge]

anyone start getting authentication failed??

[dull forge]

i do

[blazing timber]

is it down again 😐

[dull forge]

seems like it crashed

[alpine surge]

Yeah it gives me this error... can't get into anything

[blazing timber]

it's back up

at least for me

[alpine surge]

oh your right...it just came back up

[hollow sigil]

back for me too

[alpine surge]

did anyone else lose all their previous chats?

haha we're all crashing it

[gilded relic]

my card got declined so my subscription didn’t renew. now i can’t upgrade cause too much volume ???! are you serious

i’ve literally been subscribed for like 8+ months now i can’t resub cause of volume. bro

[charred parcel]

you need to go onto the waitlist

[timber shadow]

Hey @thorny belfry pinging you here since #chatgpt-discussions is quite busy right now. Do you have some questions about creating custom GPTs on your ChatGPT Plus account?

[thorny belfry]

After about half year of using ChatGPT, I just learned that you can train it on your data, lol. I have some questions:

1. Do I need a Lot of data to properly train ChatGPT? I have a small knowledge base in Confluence and would like to use it more effectively through queries to GPT, rather than just searching.

2. The content is in Spanish and English. This will clearly reduce the efficiency of training. Are there ways to compensate for this? Translating everything into English before training is not an option.

3. Does a trained personal ChatGPT become public, or can it be made private and closed off to others? I just came across something similar to what I need, but it's a public chat: chat.openai.com/g/g-f52QYAJK1-unreal-engine-5-expert (not an adv lol). So, can it be made private (apparently, just by not sharing the link to it anywhere)?

4. Are there any 'awesome lists' of such specialized chats? For now, I only know about that unreal-engine-5-expert, which is became useful for me personally and is why I am googling for custom personalized ChatGPTs 🙂

[timber shadow]

Just real quick: can you please confirm that you're talking about creating a custom GPT on chat.openai.com, and that you're not talking about fine-tuning on platform.openai.com? Some of the lingo is similar for these processes, so I just want to check before I give any feedback!

[gilded relic]

A waitlist?? after i’ve been subscribed for almost a year probably??

[timber shadow]

Unfortunately, the pause affects anyone without an active, auto-renewing subscription, so if your subscription expired and you're trying to re-subscribe, you'll indeed have to sign up for the wait list. Fingers crossed you can subscribe again soon 🤞

[thorny belfry]

To be honest, I'm not exactly sure what I'm referring to yet. I'm interested in exploring the functionality of ChatGPT when trained on my dataset, and to have an access to this trained ChatGPT through https://chat.openai.com/. It seems like I'm interested in creating a custom GPT on chat.openai.com, yes

[timber shadow]

Perfect! Thank you for confirming! And also just a quick double-check: are you currently subscribed to ChatGPT Plus?

[thorny belfry]

Yes!

[timber shadow]

Awesome! Okay, so you'll want to go to chat.openai.com/gpts/editor. Here, you'll see three main sections for creating a custom GPT: Create, Configure, and Preview. This is what I'd recommend you try to get started:

• Start a chat at Create -- tell it what kind of custom GPT you want to build, and then work with it to finish an initial setup. It will ask you questions.
• Then, go check out the Configure section. This is what Create populates for you -- there is no "behind the scenes" training occurring, everything is there in Configure. In other words, a custom GPT's behavior is influenced by the three main parts of Configure: Instructions, Knowledge Files, and Actions.
• Start by checking out Instructions. This is a field of up to 8000 characters of custom instructions for the GPT to consider in every chat you use. Edit this to your liking, but when you edit manually, be aware that returning to the Create screen and requesting changes will override your Instructions, so I recommend staying in Configure after your first setup in Create.
• Add some knowledge files, if you didn't already during Create. This would be a great place for those English/Spanish docs you mentioned. Knowledge files are like reference documents that the GPT has access to in chats--it can search them and return relevant info, or follow specific instructions in a specific knowledge file.
• Actions are like advanced developer tools, where you can hook up external API stuff to let the chat use specific tools in a chat. For example, you could hook up a public weather API (with the right programming know-how) and get your GPT to check the weather in a chat.
• Once you have a good draft, either test it in Preview, or save the GPT and go to its URL to test it out. Based on how you like it, you can go back and edit it every time (again, I recommend editing in Configure).
  This is a lot! What questions might you have about any of this?

[thorny belfry]

Wow, thanks! this is a bit more user-friendly and easier than I expected. I'll give it a try, it seems like I can manage it

One more question remains: is there any catalog of custom ChatGPTs available?

[timber shadow]

You're welcome! It's definitely not a complicated process, unless you're working on some complex Actions (which I don't have experience with, I'm afraid!). Dive on in, you can definitely manage it. And come back with any other questions you might have!

As for a catalog: be on the lookout for the launch of the impending GPT Store, but for now we have the #1171489862164168774 here in this Discord server!

[thorny belfry]

Big merci 🙂

[timber shadow]

Oh just remembered something that might be relevant to you if you plan on adding multiple knowledge files: there is a limit of ten files for this, up to 512MB each. More info about file limits can be found here: https://help.openai.com/en/articles/8555545-file-uploads-with-gpts-and-advanced-data-analysis-in-chatgpt

[marsh hedge]

Check out #1171489862164168774.

[signal beacon]

chatgpt alpha is showing up again

but...

[split flume]

Has anyone experienced the problem and resolved it?

[lucid osprey]

Is it possible to feed a GPT I’m making website links for relevant into I want it to learn from? The only option I see is “Upload files”

[pale onyx]

I've uploaded a few pdfs that contain a number of publications. When I ask the custom gpt a question - it thinks for too long and times out. Anyone know how to get around this?

[junior maple]

You could give it the links in your instructions or upload a file with a large index of them

[junior maple]

Either remove some content or wait it out :p

Errors like that are usually related to load / demand

[pale onyx]

Thanks. It will still provide a short answer - but will include "within the time allotted for this search, I wasn't able to complete a full review of the information in the knowledge source"

[spare lichen]

https://media.discordapp.net/attachments/1169994504107995177/1171902523230130226/9D16Jc3.gif?ex=655e5e46&is=654be946&hm=87fc42ea63a4ea83ef06a97d1f3b285aad2096fbf922dea0cc884cc1b415b327&

I got gpt plus back 🥳🥳🎉🎉

[bronze charm]

We're so back

[spare lichen]

🗣️facts

[shell warren]

I hit my usage for chatGPT-4 pretty quickly today.

But it's now 14:00 and it still says . "please try again after 1:10 PM......."

If I didn't hit my API cap this wouldnt be so frustrating...

Any ideas....?

[covert kindle]

What a curious thing this is, the date now is 11-November 28 2023, but my usage shows Nov 29. This date change happened in the span of a few minutes. Another issue is where did this cost come from? O,

Why was the usage future dated and where did that usage come from? Could the fact that I changed a character in my session key have something to do with it? (maybe I somehow used someone else's session key)?

Also this picture was taken about 3-4 hours ago, don't tell me the usage page is based in the UK

[still epoch]

Yes i know thanks . But how to integrate it to instagramm ?

gpts -> instagram or whatsapp

[hasty plank]

hey, is it cool to ask for help on a small project either here or a dif channel even if its not related to gpt?

[trail linden]

is the gpt store not workin for anyone. anytime i want to try a gpt it says i need to login but once I hit the login button nothing happens

[ashen crystal]

is there a any way to import module to assistant's code interpreter. i wanna use plotly instead if matplotlib

[slender cairn]

how long do i have to wait on the waitlist? i am on the waitlist i think for half a month or something? is there a plan for us who want to join 😁 ?

[junior breach]

I am also in the same situation. 😦

[wraith zodiac]

hi, i have a word document (71779 characters) that i want chatgpt to help me rewrite - is there a best way to do this?

[vital radish]

have you tried to just upload it and ask it to rewrite ?

[wraith zodiac]

is that the best way?

[glossy wasp]

Heyo guys can anyone tell me that is it possible to upload multiple documents (PDFs) and then chat with it to get relevant input ?

[timber isle]

there are custom GPT-s I think for this porpuse

[runic torrent]

Data analysis is broken since a few days for me, it cannot open a CSV file. Is this known? Is there a workaround?

[silk shoal]

Yes, This does work you can take a flow for this

1. Upload doc
2. Discuss the major topics involved in them and then extrapolate You will get relevant inputs

[brittle vault]

I shared my GPTs publicly there are 8 comments on that but i cant able to see it how to see that

[trail cairn]

dont message the same in to every channel, remove the spam and ask in one place like this help channel. the number tells you how many used it and nothing more

[chrome finch]

Hi, I have problem with my chat. I use GPT chat as normal. For each question, a modal with a captcha solution pops up. This is bothersome and I can't use the chat normally. What is going on ?

[trail cairn]

something on your system or network makes you look like a bot

[chrome finch]

I have new IP address every 24h

[trail cairn]

then its your browser or os

[olive stag]

No matter what file I upload to ChatGPT, including my GPT, it always gives Error Analyzing error and does not read any file.

[normal blaze]

I seem to have broken chatgpt after asking it to make a mind map?

It's stuck like this and I've tried refreshing the page and everything I could think of

[rigid musk]

Try firefox as test

[normal blaze]

I tried with other images, it's also bugged

I'll try firefox

[thorny belfry]

Hey guys. I have a trouble with my chat.openai.com - there only two custom chats attached, others are only visible in Explore, Recently Used section. Is there any chance to attach more? As I see on screenshots here, other users have no such problem :/

[normal blaze]

The same problem persists on firefox sadly

[rigid musk]

Are you using a vpn

[normal blaze]

No

[rigid musk]

Can you try on a new chat in incognito ?

[normal blaze]

I've never had this issue before, odd

And I've just installed firefox recently-when you told me to-so I don't think it's because of extensions, vpns, or addtional browser settings

[normal blaze]

Is there a chance that this is a server side problem? Or is it 100% a me problem and I'll have to figure a way out haha

[rigid musk]

Can you try on a phone to compare ?

[normal blaze]

I think this is the root of the problem "You have sent too many messages..."

I'll wait out a bit

[rigid musk]

Hmm let's hope it is only that :3

[normal blaze]

Thank you!

[vestal rover]

Nope. Using pics today does not work. In the #1070006915414900886 channel is a topic in this problem.

[lucid anchor]

https://chat.openai.com/g/g-Y1bhNwCdA-rock-simple | https://chat.openai.com/g/g-uK6OGJNVe-companion-gpt | https://chat.openai.com/g/g-VCjWFjlr8-arnold | https://chat.openai.com/g/g-2PLbS9ZXA-hidden-spot | https://chat.openai.com/g/g-EnxrIDKeO-tangible (If you are a software developer reach out so we can optimize some of these for the better of humanity, If you are an investor - I'm listening) 👽

[manic shell]

Hiya, I was wondering if - in it's current state - it would be possible to have a Chat with more roles other than just 'assistant' and 'user'?
Basically - is it possible to have for example 'assistant 1' and 'assistant 2' and a 'user' in the same Chat?

that would probably help me in my quest to make generative dialogue with more than one 'person' at a time.

[normal blaze]

thank you for filling me in

[normal blaze]

I think you could prompt GPT into playing 2 characters

I'm not an expert though

[trail cairn]

if you mean api then name will replace role in message

[manic shell]

I am talking about API access. as I understand it, you can send a system message, a user message, and example assistant messages.
you then request a response which takes the 'assistant' role

i was wondering if i can, say, request a response from 'assistant_1' and 'assistant_2' instead and include system messages for both of those assistants to give them different personalities

I feel like the Chat API has been more built towards a single assistant though, which when asked can roleplay multiple characters.

i thought maybe if it's possible to add multiple roles then it's a bit more robust instead of purely relying on text prompting for the multiple characters part

of course you could create two assistants and pass the chat information between the two, but I was hoping there was a way to do this where both assistants were following the same communal thread if that makes sense.

[trail cairn]

there is a name property in a message that will replace the role (IIRC)

[merry elm]

so we can make custom gpts but can we somehow integrate plugin functions into them? anyone know?

[ember aurora]

Is GPTs just replacing the step “Act as a professional physician”? Or is there any benefits other than that which I’m not aware of?

[frozen jetty]

Title: Substring Search in JSON Files with Non-Common Word Filtering Using Assistant API

Body:

Description:
Hello fellow developers! I'm currently working on a project that involves searching through JSON files based on input sentences using the Assistant API. The objective is to identify matches where either a word or the entire sentence aligns with parameters in the JSON entries. Additionally, I'm aiming to filter out common words, or stop words, and provide the corresponding 'Id' field for matches.

Specifics:

Substring Search: Break down the given sentence into words for a precise substring search in the information contained in provided JSON files.

Non-Common Words: Exclude common words (stop words) such as articles, prepositions, and frequently used words from the search. Focus on finding matches with non-common words in the JSON entries.

Output Format: Return the corresponding 'Id' field for matches where non-common words or the entire sentence align with any parameter in the JSON entries. If no match is detected, generate an output of 'No Entry.' Structure the output in JSON format as follows:

Example:
{
"ids": [ids with a match]
}
If no match is found:
{
"ids": []
}
Note: I'm utilizing the Assistant API for this task.

Instructions:
Feel free to share your code, pseudocode, or thoughts on how to approach this challenge. Any programming language is welcome. Let's collaborate to create an efficient solution.

Happy coding! 🚀

I am giving the JSON file in the Assistant API.

The main problem is that I am not getting the correct output.

Can anyone help me in this?

I am using Assistant API just because of fast searching and some intelligence.

[celest stream]

I am using GPT to turn Youtube Transcripts into properly readable text, by asking it to remove timestamps and adding punctuation.
ChatGPT does this absolutely perfect, but I would like to access the 16k context for longer transcripts.
I've tried making an Assistant in the Playground to do this, but it's... lacking. It just roughly summarizes the transcript and calls it a day.
You are an assistant who takes transcripts with timestamps and unformatted text, and turns it into that exact same text, but without timestamps, with proper formatting, punctuation, paragraphs.

I'm not entirely sure how to approach this issue, what can I do to achieve the results I want?

[frozen jetty]

I hope this message finds you well. I'm working on a project that involves matching user-inputted sentences with a dataset of text entries. I have a specific task in mind, and I'm seeking guidance on the best approach to implement it.

Task Overview:
I have a dataset with entries containing various fields, including a "query" field with text descriptions. My goal is to create a matching system that does the following:

Takes a user-inputted sentence.
Removes common words to create a reduced representation.
Searches for exact matches in the dataset based on the reduced representation.
Searches for matches based on individual words in the dataset and retrieves corresponding IDs.
Example:
If the user input is "My weekly goal has completed," the system would search for both an exact match and matches for individual words in the dataset.

Questions:

What would be the best way to preprocess and reduce the input sentence?
How can I efficiently search for exact matches in a large dataset?
What strategies would you recommend for searching and matching individual words in the dataset?
Are there any existing libraries or tools that could simplify this task?
I'd appreciate any insights, advice, or recommended resources you can provide. Thank you in advance for your assistance!

Best regards,

[fading harness]

Hello, I am currently trying to scan images using both ChatGPT and MathMentor, however, everytime I send images it returns in a error "Hmm...something seems to have gone wrong."

[rugged cedar]

Why does chatgpt display math formulas and stuff as
\frac \sqrt on mobile?(only mobile)

[modest jewel]

same for me - GPT4 has been in the gutter for 6+ hours now (mathematical/graphing equations, image responses, etc) all don't work

and it's not reflected on oa's status page either

[river estuary]

Hi! I'm not sure why but I can't save any of my customGPTs, does anyone know what i'm doing wrong or experiencing the same thing?

[proven saddle]

you need a name for it, doesn't look like you have one

[river estuary]

oh, so simple thank you! lol

[worthy hornet]

Does anyone have issues with propagation of DNS after adding TXT record to verify your domain? After 144 hours, only 1-2 specific geolocations / nameservers are missing.

[rugged wing]

is there a way to actually make new gpt 4 give me a solution to problem instead of a big pile of useless text

its extremly annyoyng

[timber shadow]

Hi, what kind of problem are you asking GPT-4 to solve?

[rugged wing]

new gpt 4 being dumb
and dumb in this case in a very light word

i tell it to fix a part of code
it gives me an instruction of what to do in order to fix it
i tell it to fix it again
gpt still refuses to do what i want it to do

and this keeps going back and forth more or less like this

[trail cairn]

its a known bug and openai working on it

[rugged wing]

doesnt look like they do
fast fix would be to allow old gpt 4 usage that one was perfect even if it had a fraction of the context availabe
and i am already paying for gpt+ and i dont want to pay for api usage just to get access to older and better model + make something to actually send stuff to it

[rugged wing]

its worse for me it doesnt give even a single line and tells me that as an ai made by open ai it cant edit my code

[timber shadow]

Hello! Do you happen to have a browser extension installed that might affect ChatGPT? I don't recognize that "play" icon under the GPT icon. If so, could you try disabling any such extension(s), and then try again?

[peak lotus]

Is this real? When I logged in it didn't ask me anything

[timber shadow]

I found some articles from late April about an age verification system being implemented in ChatGPT for Italy. Have you already verified your age as an Italian? What is the exact email address you got this notification from?

[peak lotus]

I checked the link and it actually brought me to an article from OpenAI so it's real, but I don't think that I ever verified my age

Wait, looks like I did. That email was sent to my secondary account but not to my main one, so I probably verified it.

[dense void]

hey guys, i'm having issues with using GPT on my desktop. I've tried google chrome and brave, both don't work. GPT is working on my phone though

[timber shadow]

Hello! Can you try on a new install of Firefox on desktop? Chrome and Brave are both Chromium-based, so sometimes it works for folks in a non-Chromium browser like Firefox.

[dense void]

I've tried clearing cache and restarting. still not working

I can try yes, how do I resolve it for chrome though? annoying to have to switch browsers every time

[timber shadow]

It seems to be caused by a variety of things--do you have a VPN installed?

[dense void]

I do yes but it's not active

and still not working

Still getting the same error on firefox too

oh woah, i just had to fully quit the VPN and now it works

[timber shadow]

What VPN do you have? Sometimes there's an additional feature that needs to be disabled, even if the VPN itself is already disabled, and it's called different things in different VPNs.

[dense void]

I have NordVPN

[timber shadow]

It's called "Web Protection" in NordVPN. You can find the setting in the "security score" section in the bottom-left of the desktop app. If the problem re-occurs, you might try disabling this feature as well, separately from disabling the VPN itself.

[dense void]

Ahh I see, okay thank you!

As well, I've signed up for the waitlist for the GPT4 upgrade, any idea how long it'll be?

[timber shadow]

You are welcome! Thanks for reporting what fixed it for you!

As for the waitlist: no, there hasn't yet been any announcement about how getting off the waitlist will work. Hopefully you can get subscribed soon! 🤞

[prime cedar]

Hello everyone! I really enjoy using the GPT4 chat, and I decided to write a long story, a novel)) I wrote a little bit a day. We defined the characters, their characters, made a plan and gradually started writing. Everything was great. Yesterday I exceeded the limit of 40 messages in 3 hours and the chat version went to 3.5. In my case this version works very badly and when I try to switch to version 4 (in the upper left corner of the chat) a new chat window opens. New empty chat. How do I go back to version 4 in an existing chat?

[timber shadow]

Hello! Unfortunately, there is no option currently to revert back to 4 once it's been switched to 3.5--I'm really sorry! Your best bet at present will be to copy/paste critical messages/context into a new chat to try to get back to where you were. If you feel like sharing: about how long was the chat before this happened?

[prime cedar]

oh... almost a month... we wrote very slowly and tried to get used to a specific style...
I'm sorry... I'll try to somehow transfer all the information to the new chat... Thanks for the answer!

need to add a counter of sent messages, because you can get so carried away and lose everything...

[timber shadow]

Shoot, I'm really sorry that happened to you. For what it's worth: depending on how long the chat had gotten, it's possible that a near-monthlong chat had already reached the hard context limit of GPT-4. When that happens, old data is automatically "forgotten" to make room for new data. Which might be a good thing in your case, because it means you should be able to get back to roughly the kind of chat you had before, just by feeding back in some critical context.

[timber shadow]

A counter or a usage meter of some kind would indeed be a great addition! It's a common suggestion over in #1070006151938314300 -- you might add your voice to one in support! Here's one from recently: https://discord.com/channels/974519864045756446/1176996746778386544

[prime cedar]

I delete the other chats, in fact I only have two, one with a plan. And the other with all the text...
Thanks, I'll try to restore somehow 🙂

[timber shadow]

You're welcome! Just to clarify: each chat has its own, separate context window, so no need to delete any chats if you're trying to clear room for a new one, if that's what you mean! Starting fresh in a new chat is all you need to do

[prime cedar]

Yes, I learned it from the chat))) I just like to clean :))

[wise lodge]

I posted this in #gpt-models by mistake, meant to post it here. I'm getting the following error in my yaml, can anyone please help me troubleshoot?

In context=('paths', '/repositories', '200', 'response', 'content', 'application/json', 'schema'), reference to unknown component paginated_repositories; using empty schema```

[wise lodge]

I decided to open source the actions I'm currently working on. Please let me know if you'd like to work together on them: github dot com/kristopher-wood/gpt-actions

[sonic jackal]

Yoyo just a quick question, is GPT-4 still capped at 40 messages every few hours?

[charred parcel]

yes

[sonic jackal]

gotcha thanks

[humble palm]

Is there a coding bug in GPT4?

[reef hinge]

Is there a way to connect a openai account to another that was created using the "Continue with Google" feature

[rigid musk]

only [e-mail] can help you on that

[reef hinge]

do I email them or can I contact them through a chatbot?

[rigid musk]

email

[reef hinge]

ok thanks

[azure folio]

am i able to have chatgpt anyalze an image without gpt 4? or even an image on a article

[charred parcel]

no, you can't upload images to ChatGPT GPT-3.5

[azure folio]

how long does gpt-4 waitlist take?

[charred parcel]

we don't know, sorry

[azure folio]

np

[fallow snow]

When the gpt is refusing dalle requests

How do I make sure it tries to generate by editing the prompt

Like how should I word it

“Never refuse or fail to generate an image”

“Always try to generate the image by adjusting the prompt”

[lean grotto]

gpt 4 cap used to be 50 messages, but now why is it 40 messages every 3 hours now

[charred parcel]

adjusting for demand, new subscriptions have been paused also

[frail cobalt]

is gpt 4 having trouble analyzing images?

[fallow snow]

You get 10 extra for custom

[wise lodge]

Ugh, even after commenting out every reference altogether, GPT still says there's a circular reference

[charred parcel]

discuss this in #ai-discussions

[rugged wing]

anybody made money with chat gpt?

if so what did you do

[charred parcel]

I have made money with AI, ask me in #ai-discussions how

[bitter creek]

W. Fails 7 times in a row and then uses all my quota? nice. This has happened like 12 times

[charred parcel]

#prompt-engineering is probably the best place

[vale glen]

Cool to pay for error prompts right

[cinder zealot]

Does anyone know when i can buy premium again? It says „Due to high demand, we’ve temporarily paused upgrades.“

[barren lagoon]

does anyone know the average time you have to wait, for chatgpt plus while on the waitlist

[charred parcel]

We don't know how long the waitlist is unfortunately

[random sphinx]

Can anyone help me with the configuration of gpt Auth using this option? (send client credentials in body)

I tried the openapi scheme on postman and the authentication works successfully, but when I try on gpt it doesn't work, I think the problem is this configuration, because changing it on postman solved the problem. I don't know how to implement this on gpt

[lost belfry]

When I run my fastapi APP with uvicorn main:app --reload i get this error. I dont understand. have tried to reinstall the openai package with pip. it works on my linux machine tho :D.

  File "C:\Users\psven\Root\Coding\webdev\test-projects\devflashcardgpt\FlashcardGPT-Playground\backend\main.py", line 8, in <module>
    from openai import OpenAI
ImportError: cannot import name 'OpenAI' from 'openai' (C:\Users\psven\scoop\apps\python\current\Lib\site-packages\openai\__init__.py)

[pastel mulch]

anyone knows a way to add credit to openai with paypal?

[rigid musk]

You cant

[lilac crest]

Hi guys!

I have been trying to figure out the "scope" of the opt-out request functionality that OpenAI provides (not the one in Data Controls).

Has anyone used it? I've asked their support for details regarding the scope of the opt out, as I personally believe their wording is a bit vague.

I want to know if we can count on our org being opted out of training, if we do this for every single account.

Thanks a ton :)!

[rigid musk]

Other payment option are available on mobile

[pastel mulch]

only ChatGPT is on mobile

[rigid musk]

If paying on mobile you can still use on a browser

[pastel mulch]

wdym?

[rigid musk]

Payment on mobile are with Google or Apple service.

If you pay on mobile app you can still use on a pc your same account

[pastel mulch]

not available only a cc

[rigid musk]

If google or apple only accept cc , unfortunately that's the only payment they accept

[pastel mulch]

alright thanks

[bitter creek]

Ver cool

[orchid marsh]

Hi i get an error on my nodejs code:

async function createImage(prompt) {
response = openai.images.generate(
model = "dall-e-3",
prompt = prompt,
size = "1024x1024",
quality = "standard",
n = 1,
)
image_url = response.data[0].url;
return image_url;
}

the error:

Error: Cannot read properties of undefined (reading '0')
/Users/raphaelsente/Desktop/PandoraServices/node_modules/openai/error.js:43
return new BadRequestError(status, error, message, headers);
^

BadRequestError: 400 'dall-e-3' is not of type 'object'
at APIError.generate (/Users/raphaelsente/Desktop/PandoraServices/node_modules/openai/error.js:43:20)
at OpenAI.makeStatusError (/Users/raphaelsente/Desktop/PandoraServices/node_modules/openai/core.js:251:33)
at OpenAI.makeRequest (/Users/raphaelsente/Desktop/PandoraServices/node_modules/openai/core.js:290:30)
at process.processTicksAndRejections (node:internal/process/task_queues:95:5) {
status: 400 ...

[vital radish]

Sorry to hear that

[orchid marsh]

i fixed it somehow haha

[slender island]

do not copy paste code from chat gpt without understanding first how nodejs works

[orchid marsh]

I'm not copy pasting lol.. I was integrating gpot inside an admin website as assistent and got that error, whatever i tried..

[supple falcon]

guys i want to make a device that communicate gpt 3 over api with esp32 but whe i run the ode it says please pay api cost but i got the 5$ free bonus so what i do

[fair trench]

Hello, are the rate limits imposed by Open AI in effect globally for all their endpoints, including the endpoint for polling for finished jobs (Assistants/Threads)?

Furthermore if I may ask, is there content in the body of a rate limited request?

Thank you!

[trail cairn]

https://platform.openai.com/account/limits

assistant uses the model you pick

[fair trench]

I've read those in detail. So in conclusion, even the polling endpoint to check if a run is in progress is rate limited?

[trail cairn]

polling a run status does not count to rpm but note that assistants are in beta so you should not really hit any yet as they should not really be used in production unless you really want to

[fair trench]

Thank you so much for the information! One final thing - the body of a rate limited HTTP response is empty, right? We're to find all the relevant information in the headers?

[trail cairn]

yeah you get the http code plus info in the headers, including rate cooldowns

[cursive fox]

Is there a template available that provides the correct role and instructions, along with a description for customs GTPS?
Thank you in advance guys

[charred parcel]

I've answered you in another channel, you only need to ask once in one channel

[cursive fox]

Understood

[rugged wing]

I used the application on Android before and everything worked (I live in Ukraine), but recently Google Play tells me that the application is not supported, and when I try to log in I get an error

[timber shadow]

Hi! Is your device's date and time set correctly for your time zone (is it set to "automatic" for setting date and time)? Do you have a VPN installed?

[rugged wing]

automatic setted, I tried using VPN but nothing changed
clarification, my app began to work poorly and I decided to reinstall it, but found that, as I said earlier, it is not supported
so I'm trying to install the apk

[timber shadow]

Sometimes VPNs can cause more problems actually, so if you're comfortable experimenting with uninstalling that completely, it might help with the troubleshooting. Can you also try on both a WiFi network, and a cellular network, to see if it's the same with both?

Edit: it may also be helpful to open Android's "App info" screen for the ChatGPT app, and force close/clear cache between attempts. It can help the app load up fresh in between when you're trying new things.

[shrewd cove]

Guys if I don't renew my ChatGPT Plus subscription do I lose all the GPTs or when I subscribe again do I get them back?

[charred parcel]

that's a very good question, I don't know the answer and I am not going to test it given there is a long waitlist for new subscriptions. My advice is stay subscribed to avoid disappointment

[shrewd cove]

True, absolutely. It would be interesting to understand, though

[solid mango]

Im building a chatbot that can provide answers to questions related to data science, search the internet, and provide links. I have also created a knowledge base for frequently asked questions. If the query cannot be answered using gpt-4-1106-preview, im searching in the knowledge base. But im kinda struggling to extract the links from the internet.

For example, if someone searches "how to plot a time series on matplotlib", gpt-4-1106-preview will provide the answer, but in addition to that, I also want to provide links (like links to documentations, stackoverflow, etc) from the web that people can click on to learn more about it rather than just depend on the chat. Does anybody know how to do this?

[sly field]

Hello. In case of training 100 examples, how many epochs is appropriate?

[unborn coral]

the GPT model his not capable to analyze my files

[swift iris]

hey everyone! quick question here. I'm working to help my dad quickly edit a textbook he's working on and was wondering if people have some good recommendations on how to leverage the most amount of text at one time.

is it best to do small blocks of text, single pages, or can we do chapters at a time? are there some plugins that'd be ideal for this?

[junior maple]

What specifically do you need to change?

[granite crescent]

Hello, where do I go for billing help? My API access has been suspended since July due to all my cards that used to work fine before with both ChatGPT Plus as well as the APIs suddenly stopped working with the OpenAI payment system and since then, I have attempted several times to reach somebody credible in the OpenAI team who could look into it and fix the issue but I seem to always get a response from the bot with the same lame pre-written response template.

So far, I have also tried adding funds manually to my account that's higher than the amount on the last pending invoice, yet it won't automatically deduct from it and won't restore my API access. I'm a developer and had been using ChatGPT Plus ever since it became available in my country last year and soon after started exploring the APIs but sadly for the past 5 months, I've been left in the dark and it's extremely frustrating. I have never faced such an ignorant customer support in my 20 years in IT.

[junior maple]

Part of the problem is that they're only hiring San Francisco residents with 10+ years of experience for these support roles. They will always be behind if they refuse to scale

[charred parcel]

the only place you can go is to email [e-mail] - clearly state your problem so that back and forth emails are not required

[swift iris]

it'd mostly be idea and paragraph structure, some grammar as well.

[granite crescent]

Thanks for your response! I had a huge respect for OpenAI due to their products and having learned quite a lot from Sam Altman through his YC lectures as well as due to some extraordinary people like Karpathy in their team but I no longer feel that about them (OpenAI as a company) after such a frustrating experience. Their success has made them ignorant about their customers.

[granite crescent]

Thanks for your response! I'll try that.

[quick ivy]

Hey... I guess nobody knows the answer to that:
Is there an embedding cost to uploading documents to the Assistant API?

[junior maple]

Check out https://openai.com/pricing

[quick ivy]

that suggests there's no embedding fee :/

only some standard daily fee

[junior maple]

I believe that's the correct interpretation

[civic aurora]

sorry for the noob question - I just upgraded to the gpt pro account. is pro distinct from plus? once i get off the plus waitlist, will i need to pay another $20/month on top of the pro account? should i cancel pro if what i want is access to create custom gpts?

[trail cairn]

what pro account?

[civic aurora]

it gave me an option to upgrade to pro before i saw the plus upgrade banner.

[foggy musk]

is there any way i can send a message in gpt 4 with a keybind?

pressing enter doesnt send it

[vital radish]

That is a chrome plugin/extension . Has nothing to do with ChatGpt from OpenAi

[knotty abyss]

how much time until 'You're making too many requests, please slow down.' goes away for api payment?

[timber shadow]

@dapper prism @glad hull Just pinging you both there--slightly better place for some troubleshooting! Have you both tried both a Chromium browser (Chrome, Edge, Brave, Opera), and a non-Chromium browser (Firefox, Safari)?

[glad hull]

Tried safari and Firefox and Chrome no success

[timber shadow]

You said you've tried with your VPN disabled--what VPN do you have? Sometimes there's an additional feature to try disabling, separate from the VPN itself.

[glad hull]

It works on my phone but I prefer web

Global protect

[timber shadow]

I can't find much about the features available in Global Protect--it looks like it's kind of a business-oriented VPN, is that correct? Do you have the option to totally end the VPN processes and services, as a test? That is, not just disabling the VPN connection, but ending the actual processes and services running on the computer, even in the background? On Windows, you can do this in Task Manager.

[glad hull]

just tried to no avail : (

yeah it's business vpn

[timber shadow]

Dang, and just to double-check, you were able to end all Global Protect things in both the "Processes" and "Services" sections of Task Manager?

[glad hull]

yeah used activity monitor on Mac and terminate application

still didn't work

[rancid grotto]

I have a problem with my AI on my website that I made with chatGPT. the AI on the website is saying some random stuff about history or jokes or anything like that whenever I send I massage. I'd me thankful if anybody helped

[timber shadow]

Hmmm, it seems like it's a little different from Windows, and I'm having a little trouble finding the correlate to "services" for macOS. Basically: you terminate applications in "Processes" on Windows, but there's also a secondary section called "Services" which allows you to stop "background things" that apps use as processes. It sounds like these might be called daemons or agents on macOS, and I'm wondering if it's possible that Global Protect has one or multiple such things running in the background still, even though you've terminated the application itself. Does any of this ring a bell to you? Because on Windows, it sometimes takes terminating both the processes and services of a VPN to truly terminate all VPN functions.

[timber shadow]

@proper stump Hey Jan! Just pinging you here from your post in #openai-chatter -- what kind of trouble are you having with ChatGPT?

[proper stump]

thank you, when i make a question to chat GPT in my brave say 'Page Unresponse' and take like 2-4 minutes to make a response

[timber shadow]

That's frustrating. As a test, can you try in a new install of Firefox?

[proper stump]

Firefox is better to use ChatGPT?

[timber shadow]

It depends! Sometimes it is. Some users report issues with Chromium browsers like Chrome, Edge, and Brave, and don't have the issues in non-Chromium browsers like Firefox and Safari. We can do more troubleshooting even if it works in Firefox, I'm just curious if it works better for you there for now.

[proper stump]

Appear this when i'm using Chat GPT

[timber shadow]

Is that in Firefox, or Brave?

[proper stump]

FireFox

This page is slowing down firefox

[pliant locust]

This has probably been asked a million times. Is there an ETA on the time frame for the Pro Waitlist?

Also, am i able to get Dall.e somewhere else? Or it has to be with ChatGPT Pro?

[timber shadow]

Do you have any browser extensions installed that might be affecting ChatGPT?

[timber shadow]

There haven't been any announcements yet about what getting off the waitlist might look like. DALLE 3 also powers Bing's image creator, so you can use it there as well! Just a different implementation, same underlying model!

[pliant locust]

Oh really

[proper stump]

nou

[timber shadow]

Okay. Is it possible that your computer is running out of resources? Do you, for example, have a lot of other programs running on the computer? Or, what are the general specs of your computer? Can you test on another device like a smartphone, or a different computer, to see if you see the same behavior there?

[proper stump]

My pc is normal

[dapper prism]

Yeah, I tried Chrome, Bing and Firefox and none work at all. So frustrating but I’m still guessing it’s the districts filter getting in the way. Worked perfectly on Monday so that’s the only thing I can think of causing this.

[timber shadow]

Shoot, I'm sorry I couldn't be more help! I'm not used to this kind of bug. You might try posting in #1070006915414900886 -- OpenAI staff are periodically present in there and may be able to help better than I can, or maybe someone with a similar issue will be able to chime in!

[timber shadow]

I saw that you tried disabling your VPN--which VPN do you use? There may be an additional feature to try disabling.

[dapper prism]

Thank you so much helping

[dapper prism]

Thanks but has nothing to do with a VPN. In fact it turned out I didn’t even have a VPN lol. My mistake. GPT has worked fine all year until yesterday. Now the site loads, but produces no responses and will not allow me to log out/in. Useless for me at the moment TBH

[timber shadow]

Do you happen to have a banner at the top of your screen asking if you want to try an alpha version of ChatGPT in another language?

[dapper prism]

Nope. Here is a screenshot of what I get. Notice it offers to upgrade but I am already a Plus user. Help! lol

[timber shadow]

And this is what you see in both Chrome and Firefox? And just double-checking from before: absolutely no VPN installed, even if disabled? What was the thing that made you think originally you might have a VPN?

[dapper prism]

When I ask a question I get the something went wrong notice

No VPN at all (not allowed on school laptop) and tried on both Chrome and today tried Firefox with same exact screen. Phone app works fine as does MacBook at home

[timber shadow]

Ahh I see, I wonder if it could be related to it being a school laptop? Even if you didn't change anything on your end, sometimes devices connected to businesses and schools seem to sometimes have these kinds of network issues. Is there any school network/proxy program installed, for example, that lets you access certain services for your school or something? Can you ask a schoolmate to test on their school laptop as well?

[stiff gate]

Has anyone been able to upgrade to gpt-4 recently? I’ve been stuck on a wait list for weeks now 😦

[dapper prism]

That’s what I am guessing. The filter the district uses must have been updated in some way. The frustrating part is that Perplexity and Claude have no issues. So odd. I’m the only teacher using it this much so no colleagues to work with on this. However I did check with another teacher and he also has the exact same screen

I appreciate the help here

[timber shadow]

Absolutely -- I was wondering what you meant by district filter! That makes sense. Perhaps you could get in touch with the IT folks for the district and see if they might be able to help. You might find more info as to what exactly is failing on the back end in "Inspect" on the page -- if you can find some errors and some URLs, you might be able to pass them off to IT to see if they can help you restore access.

[dapper prism]

They are not the biggest fans of GPT at the district so fingers crossed. I use it constantly for lesson planning, etc so I hope they hurry lol

[kindred pecan]

Anyone have any luck getting actions to work on a CustomGPT successfully?

I have a custom GPT that extracts text from a file and summarizes it via api call to gpt-4-1106-preview. It was working yesterday, now it is failing with a 404 response from the API. Nothing has changed. I called the API using the same api key from terminal, and the API seems to be fully functional.

I also have not been able to get the TTS API to work as an action in custom GPTs. The call is being made successfully, but I am getting an empty response from the API. Again, I am able to make API calls no problem using CURL commands. So this is clearly a custom GPT issue.

[muted compass]

Hi guys. I wonder how can I enable 2FA of my account since my data controls page doesn’t have the option of enabling 2FA 👀👀

[bitter creek]

bruh, i am on a streak of 8 regens because they keep failing and stopping mid way through. if the generations fail, why in gods name is it still eating into my tri-hourly quota??

[timber shadow]

Hi there, new 2FA setups are currently disabled, a little more info here: https://help.openai.com/en/articles/7967234-does-openai-offer-multi-factor-authentication-mfa-two-factor-authentication-2fa

[zealous veldt]

I cannot seem to work effectively with custom gpts and api "actions", if I make a change to an action it seems to duplicate it a bunch of times and then I cant save anything. Is there any solution to this other than not ever messing up the first save of the action's api schema?

[flat fable]

Hi guys. GPT-4 doens't return anything just like nothing happens when I enter the input. Anyone facing the same problem?

[swift iris]

@junior maple following up!

[junior maple]

I would experiment a bit and see what gets you the best results. Something like "revise the following text to fix its structure and grammar" should be enough of a prompt

It's more of an art than a science

[swift iris]

thx. if I wanted to do say several pages of text at a time, is that often doable?

[junior maple]

You should be able to do quite a lot of text in the Playground or using the API, but for ChatGPT you might need smaller chunks. I would just take a swing at it and go from there to be honest

[swift iris]

thx, I'll check out playground, haven't used it yet so hopefully that'll be easy to pick up.

oh rad, any plugins or good prompts that you liked?

[sweet sun]

https://openai.com/dall-e-3 showing this to me. Is the website not functioning properly?

[rotund schooner]

it's showing me this message:
You've reached your usage limit. See your usage dashboard and billing settings for more details. If you have further questions, please contact us through our help center at help.openai.com.

I still do have credits. Is this a glitch?

[vale glen]

no

1000 lines hardlimit

and it will always try to cheap out on you with 500 chars (if this about file reading by Gpt)

[rotund schooner]

gpt playground.

[gloomy cradle]

does everyone have to pay 20% tax? or how to avoid it?

[frail siren]

same here, for what it's worth, I have $20.00 credit, still getting usage limit/no more quota errors

[rotund schooner]

maybe its gonna fix itself tmr?

[frail siren]

yeah i'm wondering if it's like, a "current month" thing

[junior maple]

The tax you pay depends on your billing address. It's usually illegal to avoid paying taxes

[fair trench]

@trail cairn hello again! Do you have any idea where I can get a much better overview of what kind of errors Open AI returns for threads/assistants?

{
  "error": {
    "message": "No thread found with id '123'.",
    "type": "invalid_request_error",
    "param": null,
    "code": null
  }
}

Is what I got after attempting to fetch a non-existing thread, with a 404 status code response. Coincidentally, https://platform.openai.com/docs/guides/error-codes/api-errors does not document such a status code, nor any mention of a payload. I'd hate to find all of this by pure brute-forcing and inspecting. Is there a documentation that explains all the possible status codes we can get back? Is the 404 status code exclusive for threads/assistants only?

EDIT: Apparently 404 is also returned for chat completions. Perhaps expanding the above documentation would benefit everyone, with included information of how an error would be structured seeing as the payload is not an empty 404 response. Could you let me know if there are other undocumented status codes such as the one I encountered?

[jaunty finch]

Does anyone have 2 chatbot talking to each other?

[brazen grove]

Why do I have to constantly keep verifying if I'm human after each prompt?

[gloomy cradle]

okay thx

[trail cairn]

404 can happen in almost any endpoint if you try to access something that does not exists. all the numbers are standard http codes wth some having more specific error messages

[trail cairn]

something on your system or usage makes you look like a bot

[covert wagon]

Client error: POST https://api.openai.com/v1/engines/davinci/completions resulted in a 429 Too Many Requests response: { "error": { "message": "You exceeded your current quota, please check your plan and billing details.", (truncated...)

Is this normal, I just added billing and topup my balance and still I see this. Never used API before

[trail cairn]

wait 5-15min if you just added credits

[covert wagon]

Been 2 hrs already

[trail cairn]

that url is wrong but it should not give that error

[covert wagon]

whats the correct url

again it should not give exceeded error right ?

without api

[trail cairn]

you were probably looking for this https://platform.openai.com/docs/api-reference/chat/create look at the examples

[midnight estuary]

Do i need to host my own privacy policy webpage if i want to share my custom gpt with the public?

[trail cairn]

yes, if its using actions

[midnight estuary]

but my action is connected to a 3rd party website.. i dont have a website connected to the action

[trail cairn]

you still need to tell the user what the action does and how the data is used even if its a 3rd party api/site

[midnight estuary]

i understand , my only confusion is whether i have to host a website just for this privacy policy?

[trail cairn]

you need to have it available somewhere public where anyone can access it, public link at google drive, github etc.

[polar jungle]

Hello Community!

I am new in this OpenAI and I need some help

[timber shadow]

Hello! What do you need help with?

[polar jungle]

I have large amount of data in my mongodb.
Now I want to build a chatbot that integrated with this database.

So users can ask any questions related to this database such as generate KPI chart and so on.

[timber shadow]

I see, are you trying to do this on ChatGPT with a custom GPT, or are you planning to program your own chatbot on your own platform using OpenAI's API?

[polar jungle]

the second one

I want to build my own chatbot using openAI API.

If you have any idea, Please help me.

[timber shadow]

Okay! You might start by taking a look at platform.openai.com -- that's your hub for API resources. See what's available to you there, how you might start building your chatbot, then the folks in #dev-chat might be able to help with specific questions you have after that! You might browse through the other "Developers' Corner" channels on this server first too, as there might be some good examples there to get you started as well!

[polar jungle]

Ok

Thanks

Hi @timber shadow
Please answer my last one question
As I mentioned above, I am going to integrate ChatGPT to my database to answer users questions.
Please tell me how can I do this.

[fair trench]

I started inspecting the python API to understand better what cases are being handled. There appears to be even a x-retry-after header that's being read.

Could you help explain to me this part of the process:

def _make_status_error(
        self,
        err_msg: str,
        *,
        body: object,
        response: httpx.Response,
    ) -> APIStatusError:
        data = body.get("error", body) if is_mapping(body) else body
      // rest of the code..

This line of code data = body.get("error", body) if is_mapping(body) else body - I'm not sure what is_mapping(body) does. If it returns true, then I presume the "error" property value is extracted from the json response? After inspecting a bit further I saw that the body is always the raw text of the http response as done here:

def _make_status_error_from_response(
        self,
        response: httpx.Response,
    ) -> APIStatusError:
        err_text = response.text.strip()
        body = err_text

[timber shadow]

I'm not a programmer myself so I can't advise you on using the API, but if you take a look at the resources on https://platform.openai.com, you'll be off to a good start! Then, the folks in #dev-chat can probably help answer any more specific questions you might have!

[fair trench]

I'm guessing Radium is no programmer as well? Let me know so I delete my inquiry and forward it into #dev-chat

[timber shadow]

I think Radium is a programmer!

[midnight estuary]

have you tried asking GPT 4 ?

[trail cairn]

yes it will check if the body has error section and will use it, else it will just return the body contents

[fair trench]

thank you radium, you rock! 🚀 I'm unable to use the python library, so doing a port of it hence the inquiries 😄 (different programming language)

[trail cairn]

you can always use #dev-chat for quicker responses 🙂

[fair trench]

I actually asked there first my initial question but nobody replied, but I'll definitely go there from now on to allow you to answer other more pressing matters 👍 I'll come back here if there's been no development for some days, thank you again!

[polar jungle]

Hello @covert wagon
I got this error too.
Did you fix this error?

[covert wagon]

Nop

I messaged support

let me see what they reply

[polar jungle]

OK

[arctic ivy]

I have a pro account authentificated with google.. I now need to login on external devices so i dont want to log with google... can´t disconnect the auth login... Now i need to 1) Stop PRO membership 2)delete account 3) prey that i will be able to register with my email 4) wait for the waiting list of PRO registrations... 5) Suscribe to PRO again,....

ALL because i can´t change email or change auth metod ... Is there any workaround available??

[timber shadow]

Hi, you should reach out to [e-mail] as they would be the only ones who might be able to help you change your login method. Not certain on this though, I hope it's possible for ya!

[arctic ivy]

Yea... thank you bro.. seems that the delay for an answer it will take even MORE time than just all the steps i wrote before.. Some people report more than 2 weeks of delay

[timber shadow]

Sure thing. Please be careful deleting your account, as you can't reuse certain information in making a new account. It should warn you, but you might check this beforehand too: https://help.openai.com/en/articles/6378407-how-to-delete-your-account
We don't know how long the waitlist will be a thing, so I really think your best bet would be to just open that line of communication with support. Good luck 👍🤞

[arctic ivy]

Ok, thanks Solbus... I really dont understand how OpenAi can have this very very very BAD system, basic needs are not covered and yet they lead AI revolution...

[prisma rover]

I already have a GPT (only me) saved and I'm trying to update it's behavior, but it keeps showing an error message when trying to save. What can I do?

[cold folio]

Have you uploaded too many files? can you try just with 1 ?

[prisma rover]

Yes, it has 10 pdf files. But it already had, the update I'm trying to do is exclusive on behavior. Even so must I delete the PDFs?
Edit: Yeah, it worked. Thanks. Is it impossible to upload PDFs files now on Custom GPT so it keeps on it's database?

[queen cedar]

I'm trying to develop a set of prompts for an application that's being built with a group. I'm technical, but not a programmer. It requires the processing of two text files.

Should I be trying to use the playground for this.
If I use the "chat" I can't give it the files, if I use the Assistant, it's thinking I'm trying to generate code.
Any thoughts/insights/benevolent guidance?

(Addtionally, I'm trying to use playground for the multiple iterations and I don't want to hit the 4.0 cap on messages. If there's a better way to handle this, LMK)

[rugged wing]

hey i know that new gpt 4 has some problems with being lazy making code and omitting most of it but now it also says that it cant upload files or make a mod (opera gx mod i send gpt 4 a link to documentation and it correctly read it) which i am sure it can do or at least make an easy ready to edit example mod but instead it says stuff like this

I'm sorry for the confusion. As an AI, I can guide you through the process or provide code examples, but I can't physically create and upload a mod for you. However, I can certainly help with specific aspects of mod development, such as writing code for a feature, designing elements, or explaining how to implement certain functionalities. If you have a particular part of the mod-making process you'd like assistance with, please let me know, and I'll do my best to help!

seems like open ai's attempts to fix the lazy code making bug made gpt even lazier

is there any way to force gpt 4 to stop giving me nonsense and just do what i want like special prompts or something i know that stuff like this existed in past like chatgpt in dev mode prompt

[scarlet tusk]

Hi, is it possible to limit an assistant's response length?

[rugged wing]

Good evening,
I created images on slab, but I can't use them to sell them, by putting them on the cover of my books. Is there a possibility to buy them?

[timber shadow]

Hi, your best bet for non-OpenAI discussion is in #ai-discussions

[rugged wing]

Ok thank you

[fresh kelp]

anyone else unable to access your api usage?

(nvm looks like it fixed itself)

[rugged wing]

So some how my conversations with chatgpt go from using the chatgpt 4 model to 3.5 how do I stop this

[charred parcel]

if you go over the message limit for GPT-4 it may switch you to GPT-3.5

[rugged wing]

Oh

Pain

[fair beacon]

4.0 wont generate images at the moment.
It seems there was an error generating the images. Let's try again. I will create three distinct ribbon designs that symbolize the ranks of explorers, with each subsequent design increasing in complexity and prestige. Please hold on for a moment while I attempt to generate these designs again.

I'm sorry for the inconvenience, but it seems that the issue with the image generation tool is persisting. While I cannot directly resolve this technical problem, I would recommend that you use the design descriptions provided as a guide to create the ribbons or work with a graphic designer who can bring these ideas to life. If you have any other questions or need further assistance with your project, feel free to ask!

status page is green

[fresh kelp]

4 seems to be getting overloaded atm

[fading sorrel]

My custom GPT is no longer attempting to use knowledge I've provided, and prefers to use actions, or if enabled the code interpreter to open the files instead. If I switch everything else off, it just refuses to use RAG, has anyone come accross this before?
Wow this is odd...I reverted to a previous version, tweaked my prompt a bit, and now it is doing both RAG and Analysis simultaneously, it opened the file with python, and started writing a message while also accessing knowledge. Then started writing a second reply using the knowledge, all at the same time

I think something is wrong with RAG, this was all printed out simultaneously and the message it started writing wasn't based on the knowledge at all.

[valid panther]

Any way to expose a Custom GPT via API? I can't seem to get the Assistant API to accept images for review.

[charred parcel]

no, GPTs are inside ChatGPT only

[fair beacon]

Any news on the image generator being back online for 4.0? Msg'd 4 hours ago. Can we have an update on this please.

[charred parcel]

DALL-E 3 is functioning normally

[fair beacon]

No,it isnt.

[fair beacon]

then we are talking about two different things. As I originally mentioned. here, this is not function in Chat GTP 4.0
#community-help message

[charred parcel]

what are you trying to create an image in?

[fair beacon]

Inside open ai's website using Chat GPT 4

[charred parcel]

that's what I just did and posted

[fair beacon]

well then I cant explain it. I even made a new chat to test it, same error.

[junior maple]

This sounds like a "down for some but not for all" type of issue

[fair beacon]

I posted a screenshot, I am not making this up.

[fair beacon]

perhaps

Just letting you guys here know,

[charred parcel]

what was your prompt?

[fair beacon]

you mean my request?

[charred parcel]

yes

[junior maple]

GPTs are the new ChatGPT customization feature for Plus users. That's what #gpt-models is for

[fair beacon]

so strange, made a nw chat and its working again but not in the other chat.

[charred parcel]

then you're good to go

[junior maple]

They do have a description for that channel and it does predate the GPT product but I think the channel is kind of transformed into what it is now :p

I think it more or less morphed into being about GPTs on ChatGPT

GPTs is way more marketable imo. those are a little silly

can't say I know what's bugging you tbh :p

thats cus they are :p

I don't really see that argument tbh. Anyways why not post a suggestion?

honestly you have a pretty good life if this is something you've given any thought about xD just relax

[red trail]

can i ask for help here?

[charred parcel]

you can

[merry siren]

my billing plan was canceled when i didn’t cancel it??? and now it won’t let me start a billing plan again..i literally just added money into too. is this a lost cause or..?

[left timber]

We have determined that you are using OpenAI services in ways that violate our policies.

I got this email today

How do we deal with this

[rigid musk]

Contact [e-mail]

[left timber]

thanks

[junior maple]

if you have history on then you can have hundreds at once

I don't see why you would need to do that

Just switch between chats in the same tab

But you could probably just open the website in two different tabs

You don't get banned for using ChatGPT multiple times

[lusty loom]

Hi 👋 everyone

[blazing timber]

is gpt4 lagging for anyone else? Thing is acting like my PC is out of RAM... While GPT3 is instant

[trim furnace]

Have been trying to use a GPT to help write concise descriptions of story elements, and it feels like I am constantly hitting the GPT4 useage limit. But I would swear on a stack of Bibles that I haven't actually submitted that many queries. Does using a GPT cause you to hit the limit earlier? Because that's a really good reason not to use them and just copy/paste prompt chunks instead...

[trim furnace]

Yes, sometimes I'm getting lag where I submit a prompt and the server just sits there before streaming tokens, or just straight up times out.

It's been happening on and off for several days at this point.

[tawny gust]

Dont seem to be able to save my custom GPT that I jsut upgraded. I also realize that it's diffult to read word docs - I have to convert all my examples in txt. On the other hand, it's easy to generate a word file. Any hints? thanks for reading

[fresh kraken]

Hi, i got issues with Dall-E. Is another else have too

[lusty loom]

I personally use .txt file and format. As it works better always.

[blazing timber]

mfw gpt 3.5 is better at creating regex patterns than gpt 4 😐

[twilit meteor]

Can someone tell me please what is the pricing for gpt-3.5-turbo-0613? I can find only for 1106, are they same?

[rough iron]

anyones else having problems getting it to calulate anything? like pretty much anything just slightly complicated it fails when typing the presumably last line.
example:```

Solving the system of equations again, this time extracting the solutions directly

solved_values = solve((dl, dc, d_lambda), (l, c, lambda_))

Extracting the realistic solution (positive consumption and leisure less than 24 hours)

optimal_values = next((sol for sol in solved_values if sol[l].is_real and 0 <= sol[l] <= 24 and sol[c] > 0), None)

optimal_values```

[full surge]

what's wrong again

[bright lily]

I unsubscribed until the service is actually faster and isnt extremely bogged down.

[full surge]

when Sam be fired again maybe problems be solved

[bright lily]

🤷‍♂️

[fierce harbor]

Had anyone found any tools that help load gpts with a github repo? Im sure there are some out there but I’ve yet to find a good one.

[unborn coral]

chats can't load

[wary kraken]

trying to load any previous chat gives an error:

[verbal isle]

"Unable to load conversation" -> anyone else having this issue right now? None of my previous Chats will load. Ive tried different browsers (Chrome, Edge), same issue. It is not the "unable to load history" but. My Chat history is shown, but none of my chats can beloaded. I created new ones, which appear in the chat history, but can not be opened/loaded. These have been chats from 1min ago. Any solutions here? (just wait? :/). Login, logout did not work. Internet reconnect / VPN did not work. Change of browser also did not work.

[verbal isle]

same for me.

[wary kraken]

Neither in firefox.

[jolly kelp]

Same I wonder what happened, It was fine minutes ago

[verbal isle]

I suppose then its a broader issue... first time this happened to me and seems to affect more ppl.

btw... it works fine in the iOS app for me ... ? o_O

[unborn coral]

big issue yah

[unborn coral]

it's good now @verbal isle

[hardy wharf]

None of my chat history loads, everything fails with "Hmm...something seems to have gone wrong," does not work on android, web, over vpn, etc. Been like this for 4+ hours, but https://status.openai.com/ says it's fine

[jolly kelp]

It works now thankfully

[white mortar]

why when i ask the chat gpt to search the web it does it for a minute and answer is blank

[verbal isle]

yes! same for me. Now it works!

[outer steeple]

My ChatGPT Plus subscription ends tomorrow, so, will i be able to renew it considering that all ChatGPT Plus subscriptions were halted?

[silent granite]

Can someone recommend me an platform/app for sharing prompts? Me and a colleague wants a way of sharing prompts with function calls so we can experiment together. I need something that a non-programmer with some reasonable tech skills can be using. Bonus if I can self-host and queue up multiple jobs and compare results. It's probably a big ask, I'm aware

[remote pendant]

anyone else know how to fix that all GPTs including 4 can't create images other than DALLE?

[pliant heath]

i got 5 credits but card thorwing money for gpt4 wont work plz help i got a job as a indie game dev and am using autogpt for ai art

[dusty holly]

automated site to sale for your game art? nice. lots of automated pipelines now. better pay in.

[rustic marlin]

what do you mean?

[remote pendant]

if I’m using one of my GPTs or normal GPT 4 it always fails at creating any image. It only succeeds in DALLE

[rustic marlin]

ChatGPT uses Dalle

do you mean, the labs website?

it fails some times, usually when it rejects the prompt

could your prompt be something that does not follow the content policy??

[weary drift]

Why can't I log in to my account? As soon as I log in, I return to the login interface. Does anyone know why?

[bitter creek]

😭

[clear escarp]

how the hell are we supposed to debug oauth issues with gpts

[hot brook]

Hello, I’ve been having a conversation with ChatGPT specific to a point and click adventure game I want it to help me with. It was pretty good at the beginning, but at one point I started uploading multiple images of a diary from a pdf of the game showing lots of hand written text, and eventually it said it couldn’t process the images at that time. That was yesterday, about 9 hours ago. But I tried asking it to read the text on the pages again today in the same conversation and it kept saying it had trouble analysing the images. I wonder if perhaps OpenAI is seeing my image uploading as excessive and throttling the service.

[full surge]

when i am working on my custom GPT it quit the current chat and create new empty chat . thanks OpenAI team they are going to crazy us with stupid such errors

[sharp basin]

Hi does anyone's ChatGPT 4 has this bug where right after the 3 hour prompt ban, submitting a new prompt extends the ban to the next 3 hours?

[patent sphinx]

Same thing happening to me

[rugged wing]

I've been having this error the past few days, any ideas?

Can't access my Chats etc

[tulip zinc]

anyone know how long the waitlist is for plus

[charred parcel]

we don't know

[rugged wing]

hey chatgpt site constantly uses A LOT of my cpu its not some old cpu that cant handle a website its i3 1115g4 so it should handle it just fine
so what causes this is chatgpt site mining bitcoin or what

[ocean lake]

My ChatGPT hasn't been working since yesterday. I get this error. What is happening? It's almost impossible to reach out to them. People around me are using their ChatGPT just fine. I feel like I've been blocked, but I've not even been told.

[outer steeple]

The same, but I dont have any errors, when i send message it just disappears and nothing happens

[rigid musk]

Try firefox

[crystal sentinel]

Hi, I use Brave as my browser, and for the last month or so every time I open my browser, ChatGPT logs out. I tried deleting the cache and cookies but nothing helped.

[rigid musk]

Try a ,non chromium browser for a test

[spare bough]

Would there be any way I could upload a picture let's say of a map of a game like Zelda for instance then ask chatgpt or any other AI to put a pin on where all the shops are etc

[trail cairn]

nope

[bright pivot]

Vision can do it (don't confuse It by GPT Vision) it's a separate model

[honest rune]

hi guys

ive a Plus subscription but gpt4 isnt available for me in the model options only gpt3.5 turbo , can someone help with this ?

[rigid musk]

delete cookie, log off and try again

[crystal sentinel]

Ever Edge is Chromium 🫣 🤦‍♂️

On my PC I have Brave and it works fine but on my laptop it logs out

[remote pendant]

Nope

[patent sphinx]

Same thing happening to me and other people, not really sure whats happening but if i log in with another account it works, very weird

[jolly kelp]

Hi when I try to ask chatgpt for a prompt it doesn't generate it only generates after I refresh the page it was fine hours back what should I do??

[sour snow]

Yeah same for me. Hoping it gets back soon.

Btw how do i run 3.5, i can only see 4 ?

[jolly kelp]

this only happens with the prompts which have more messages

here is what i get in firefox it's the same in chrome:
what should I do to make it faster??

[knotty abyss]

I am having that same issue.

[patent sphinx]

Someone knows if there is a contact email address?? I cant find a way to contact them and report errors

[slender granite]

Hi folks! For the last 12 hours I keep getting this error in ChatGPT Plus, both on desktop and on Android.

Anyone else seeing similar issues, please, or know how to diagnose it? Thanks for any help!

And when I try to logout, I can't log back in, getting this message.

[slender granite]

I'm having the same issue.

Issue happens regardless of client :/ Been trying in Chromium (Edge) and (a fresh install of) Firefox. In Firefox I manage to log in, but then answers messages with the same error. 🤷

[hardy wharf]

Happening to me too, https://discord.com/channels/974519864045756446/1180691552574656602

[rustic marlin]

it seems to not be a problerm with your account, just that the loa balance in your region is overloaded

[rustic marlin]

it seems to not be a problerm with your account, just that the loa balance in your region is overloaded

[hybrid jackal]

How do I change the text?

[pallid ferry]

I may be wrong, but Dalle, MJ, etc... do not usually do well with text and words, I think usually best to just create the image, save, then re-edit in Photoshop, Affinity Phote, etc... for final product.

[junior maple]

Dall E 3 is ok at small amounts of text. Dall E 2 can hardly do any

[past canopy]

I am not able to poste on the daily theme wall, don't know what is happening

[fiery glade]

Hi there! Can someone explain to me how I can use Code Interpreter to run a custom script that I uploaded? Right now the code interpreter tries to import it, but throws a ModuleNotFoundError.

[night breach]

There's currently a bug going around right now. Ask again in a couple of days.

[fervent radish]

How long does it take to get Chatgpt 4 plus?

[rigid musk]

No eta it is paused

[fervent radish]

its paused?

[fervent radish]

so there is no waiting list?

[rigid musk]

You can join the waiting list of course

But I means, they slowly grant access. It is not all open like it was few weeks ago

[harsh zenith]

what can I do ?

[bleak plover]

Any idea as to why chatgpt is running so slowly?....

[charred parcel]

wait patiently is all you can do

[charred parcel]

ChatGPT is slow because it is under massive load due to demand

[bleak plover]

Understandable! Just noticing some strange quirks. If I have very long conversations it seems to get slower.

[near eagle]

Hi, i have a problem with this problem.

them i have a gpt plus

Someone can help me, please

[hardy wharf]

Been happening to a bunch of us: https://discord.com/channels/974519864045756446/1180691552574656602 https://discord.com/channels/974519864045756446/1180525598880321698

[near eagle]

There is no solution date or any way to fix it?

[cold folio]

Have you tried other browsers?

[tropic spade]

something is wrong with chatgpt my conversation keep starting a new one halfway through my earlier conversation it seriously give a bad follow on as when new chat open the old conversation is not taken into consideration anymore

it's working on my end but it's not reliable

seriously can openAI stop doing testing and research on live server ?

this is subpar unprofessional

almost felt like a guinea pig this kind of situation we should be paid instead for helping to research your system....

please many users are relying on your system to do their work

[pine copper]

me too

[hoary jacinth]

Do I have to pay to use dall.e? Not 2, of course.

[charred parcel]

you need to be a GPT Plus subscriber to use DALL-E 3

[hoary jacinth]

😦

What a pity

[regal jetty]

Hmm...something seems to have gone wrong.

[charred parcel]

do you have a product called NordVPN installed on your computer?

[regal jetty]

yes

[charred parcel]

disable it's web protection feature then everything will work again

[regal jetty]

how?

[charred parcel]

look in the settings for NordVPN and disable it's web protection feature

[regal jetty]

I used clash of windows instead of NordVPN

[regal jetty]

I used clash for windows instead of NordVPN

[charred parcel]

what do you mean? that sentence makes no sense

[regal jetty]

the product i used is clash for windows not NordVPN. Of course it is also a vpn software

[charred parcel]

something on your computer is stopping all page resources from loading on ChatGPT. Disable or uninstall any VPN or web protection software to fix the issue

[rough fossil]

Hi I'm new here. My Chatgpt plus account seems to be not working. cant access chat history, and when i start a new chat, it says "Hmm...something seems to have gone wrong." and "There was an error generating a response" at the bottom

can't find any solution at the FAQ section

[charred parcel]

do you have a product called NordVPN installed?

[regal jetty]

I tried many methods but it still didn't work. First I tried to log out and log in again. Then I cleared the browser cache and found that it still didn't work. Then I changed the browser again and found that it still didn't work. Finally I changed my computer and mobile phone and it still didn't work. It couldn't be used. Finally, I couldn't use it on my colleague's computer. Her computer could use her own account, but she couldn't use mine.

hey bro

[rugged wing]

why cant i buy plus=

[regal jetty]

money

[rugged wing]

no it’s disabled due to high demand but for my friends it’s not

[regal jetty]

Maybe he is a veteran user

[rugged wing]

when will i be able to?

[regal jetty]

wait a while

[rugged wing]

it has been saying that for 4 months

[regal jetty]

hold on

😆

[rough oar]

Hey I need one help from Dall e to Write css for the "Property name" dropdown which is a select element with Option Group element elements created with select, optgroup and options

[ivory cradle]

Hello, anyone else having trouble with this reoccurng error popping up? Neither ChatGPT, or any of the like is working for me for some reason

[rustic marlin]

depends o nthe error, often clearing cache solves it

[cursive bear]

All of my company can’t use ChatGPT - paid and free versions.
I can use my account on phone.
VPN sign off, incognito didn’t work.
Getting 403 error when I try to complete a message.
Any ideas?

[rigid musk]

Try to disable the proxy on windows close browser try again

[cursive bear]

Thank you - no help 😦

[rigid musk]

If working on cellphone without wifi might be your job blocking something

Try firefox as a test also

[cursive bear]

We already cleared with our IT and Security - not them
High management also not aware of anything.
I tried from PC, incognito and VPN off. All at once.

[rigid musk]

If you have a vpn make sure to disable web browsing

[cursive bear]

I’ll check - though. Free tier and gpt-3.5 generally not working as well. Tomorrow I’ll validate company Macs, to see if just Windows affected. Today I’ll check non-company windows machine

[hard spade]

Is anyone getting this error?: "Something went wrong. If this issue persists please contact us through our help center at help.openai.com."

I am a paid Pro user

[cursive bear]

@hard spade see my texts above

Do you have Windows or Mac? (Or Linux?)

[hard spade]

I am on Windows and likely having the same problem. I have used three different browsers to troubleshoot. I also cleared my cache and history. I have also tried it in Incognito mode and my Pro account is fully paid up.

[cursive bear]

Same issue. I’ll do further testing today and tomorrow and update

[hard spade]

Here's another clue: I created a brand new account and it still failed.

[cursive bear]

I have strong sense it’s windows related. Possibly specific version, or some setting that was automatically changed (or change on OpenAI side related to it)

[hard spade]

ok. thanks for the feedback. I will monitor this server today for any updates.

[winged ridge]

My GPTs is having an issue reading a couple spreadsheets, I tried several formats.
"Error Analyzing"
I spent all my use hours trying to fix this now I have to wait hours.

[delicate isle]

whats the difference between the 2 preniums options

[trail cairn]

2? there is only one and even that one was paused

[delicate isle]

i lie my bad it’s one you can buy it

[reef fiber]

Greetings I have a confouding CAPTCHA challenge that I can't seem to correct. on my Windows laptop in multiple browsers, I am receiving the CAPTCHA every single time I submit a prompt to ChatGPT. It does not also happen when I am logged in on my mobile device (either in app or mobile browser). Same thing was happening on my prior laptop before I replaced it. Anyone know how to fix it?

[trail cairn]

you have vpn on that laptop? or something else that makes you lool like a bot or suspicious

[reef fiber]

I'm not using a VPN let me try that and see if it makes a difference

[reef fiber]

Turning on the VPN worked! Thank you!!!

[willow coral]

Hi! Having „Hmm...something seems to have gone wrong.” On any request in application, in web client, on a different devices in a browser, everywhere, have no VPN, have no cookie-blocker (does it exists for iOS?) did anyone seen anything like this?

[rigid musk]

Have no vpn actif or installed . Big difference . What browser did you try. And yes it exist chatgpt app on ios

[willow coral]

Have no VPN at all, tried safari, google, opera. And I doubt there is cookie-blocker for iOS.

[rigid musk]

Can you try on a pc with Firefox?

[willow coral]

I have no PC, all my computers are Mac

[rigid musk]

Does your account works on a cellphone with the app ?

[willow coral]

Same stuff, anywhere, in a phone, on laptop, on my pad. In application from chatGPT too.

[rigid musk]

I suggest to post in #1070006915414900886 or contact [e-mail]

[willow coral]

No use, they suggested me to switch vpn off, but I am not using it.

[rigid musk]

Bug report often openai employee do read and trying to help.

[rigid musk]

#1180525598880321698 message

[willow coral]

Yes, I found it. So the issue I’m having occurred with many people few days ago. And nobody cares

[rigid musk]

They care an openai employee did answer they are working on a fix

[winged ridge]

My GPTs is having an issue reading anything.
Error Analyzing
I asked for clarity and it said it was getting a
AceInternalException
This typically indicates an internal issue with the system I use to process and access files. It's not an issue with the file itself, but rather with the system's ability to read
The most common cause of an "AceInternalException" is typically related to system overload or resource limitations.

[rigid musk]

Less than 1 day ago they writed it

[willow coral]

I also met such issue, trying to work with two documents uploaded, 200 kb all together. It was accepting files in the beginning, then was failing to address information in it, then begun to fail to read any format, pdf, CSS, txt. So it was until I had my allowance finished. Was my first attempt to work with my documents. Completely failed.

[glad hull]

anyone having to bill again? despite having paid for gpt4 couple days ago?

[glad hull]

actually I paid a week ago but I can't access gpt 4 today, it's as if I never paid

[trail cairn]

paid for what and cant access where? official plus subs have been disabled for new users for weeks

[wide garnet]

I'm a paid Plus user. I have a couple of questions. If it's not the right channel, please lmk where to post these questions:

1. During the past week and potentially longer than that, chatGPT normal chat with plug-ins is very slow and is hanging for long period of times? In fact, Firefox is saying that the tab is unresponsive and asking me to kill it. Anyone else notice this? Is there a workaround?
2. Is there a preferred browser for chatGPT normal chat?

[cursive bear]

It got resolved by itself!
For me and others.
Also some had it down for a week and also got resolved

[quaint vector]

whoever has gpt4 how much did the total cost? (canada) or usa

[glad hull]

i've had it for sometime now so not in that boat : /
can't access premium but I just got billed, able to access standard chat

[trail cairn]

try in private mode or in another browser like firefox

[glad hull]

tried, didn't work just submitted a ticket : /

[surreal cypress]

i keep getting the "You've reached the current usage cap for GPT-4, please try again after 3:41 PM. Learn more" after only a few interactions, is this normal for a paying customer to be limited so much?

[cosmic crag]

define "a few". And yeas, the plus subscribers are still limited in gpt-4 messages

[surreal cypress]

9 messages in 1 hour

[surreal cypress]

i also get this when i click the share icon on the gpt im using

[vital radish]

This should be fixed now according to @brittle summit

[random iron]

Reg: Assistants API. pricing. Anyone help me understand how the pricing for Assistants API works? Basically Assistants API is priced by session. If we keep the session on, is there some usage fee or is it one-time charge per session? Alternatively, every single time we create a session, we are charged? Similarly how is the cost / file ? how is it charged? If we delete files on day to basis after use for our sessions, what happens?

[spare mantle]

When I'm troubleshooting, I also terminate all my browser extensions; sometimes that helps.

[spare mantle]

I read that I'm allowed 40 questions / 3 hours; so I thought I'd load up my questions, ask complicated, multi-faceted questions. But it seemed to me that with complicated questions, I was capped sooner. Nothing measured, just my impression.

[spare mantle]

$20 US - $28.51 CAN

[surreal cypress]

that makes sense i wish the cap was explained better

[cold folio]

do you use vpn ? are you on chrome or firefox ?

[errant ocean]

Im encountering difficulties while upgrading chatGP? did they pause the upgrade function at this stage?

[charred parcel]

yes, you need to go on the waitlist

[fervent widget]

I get this on all devices and browser while my frieds and work mates have no issues. I have a plus subscription. I tried contacting customer service but they will respond ”within a week”. I tried switching browser and incognito mode, I tried logging out and in again, I tried switching device (both different computers and phones). I dont think my avcount is blocked because my API-keys still work. Does anyone have any suggestions on how to solve this?

[lapis portal]

hello it keeps saying can't find "cudnn_cnn_infer64_8.dll" how can I fix this?

[lapis portal]

I really need help so If anyone can help me

I would really appreciate it

[regal jetty]

me too

[vital radish]

I also find a bunch of timeouts today. its really bad

[sturdy apex]

Hey everyone, i want to create a quiz gpt and i want to give it my summary of university lecture. Now i notice that the gpt only creates questions based on the beginning of the summary. Do you have any tips and tricks on how to give my gpt knowledge in chat

[blazing timber]

Anyone else getting trashed by chatGPT again?

[rigid musk]

Uh ?

[blazing timber]

getting quite a lot of errors today 😐

[surreal cypress]

im using arc which is a chrome based browser for mac

[noble haven]

Did the waitlist reset or something? I had to sign up for the waitlist again!

[night dagger]

They have an exceptionally high load today, probably because of the preview launch of GPT5. I can work with ChatGPT, but the messages are very sluggish and sometimes I get a network error.

[untold ledge]

Chat GPT 4 does not response back on desktop browser but works on Mobile app
Is there any solution to this please help

[rigid musk]

Try another browser in incognito as test

[untold ledge]

i have tried, edgse and Chrome with no luck (incognito also), chrome on mobile works.
Will try Firefox as after searching here some reported Firefox works.

[jolly kelp]

Is the lag issue fixed??

[fleet swan]

Hi! I am using the Assistants API with retrieval tool to upload a file and need the assistant to answer pre-defined 22 questions with RAG+LLM.

When I send all 22 questions in one shot and insist 22 answers back, it sometimes fails (some questions are not answered etc.).

Based on your experience, is that okay If I send the 22 questions one by one in a loop?

[timber shadow]

Hello! Since you're using the API, you might get better informed answers in #dev-chat (I'm not sure exactly what I'm talking about here), but I'm wondering, if you're specifically requesting "22 answers", if you might be running into the limitations of the model's inability to do math. That is, it might be getting stuck on "22" and since it can't count to 22, it's giving you the mixed results. If this is the case, you might get better results with more qualitative language like "answer each of/all of the questions". I think your idea to send one question at a time would work fine too, but again, #dev-chat is there for ya too!

[woven viper]

anyone know how i can get the old ui for actions?

mine looks like this, idk if there is a way to make it look more like post man again??

[sweet spruce]

        const postConfig = {
              headers: {
                  "Content-Type": "application/json",
                  Authorization: `Bearer ${apiKey}`,
              },
          };

          // Post data
          const postData = {
              prompt,
              model: "gpt-4-1106-preview",
              temperature: 1,
              max_tokens: 200,
          };

          // Axios post request
          const response = await axios.post('https://api.openai.com/v1/completions', postData, postConfig);
          const chatbotResponse = response.data.choices[0].text.trim();
          

Am I doing anything wrong?
Code works fine with

  model: "text-davinci-003",

But It does not run with gpt-4, is the endpoint wrong?

[trail cairn]

that's the only ui there is and has been for almost a month. if you refer the old ui that was available just after release for a day or two it was possibly scrapped for good or comes later with the store release

[woven viper]

thats what i thought i been using this one but wasnt sure

[trail cairn]

wrong endpoint and parameters, see https://platform.openai.com/docs/api-reference/chat/create

[zenith pewter]

where settings go

[woven viper]

try to scroll brother

[zenith pewter]

Its nowhere to be found 😭

[rigid musk]

Can you try another browser

[zenith pewter]

it just fixed it has my profile under upgrade now

[jovial plinth]

anyone facing this issue? I am having trouble with this for last 3 days

also, does anyone know how to keep the sidebar fixed instead of floating?

[crystal sandal]

Page just reloads constantly.

I have conversation history turned off, but basically any time I switch away from the conversation (to copy/paste content elsewhere for example) when I return - the conversation just reloads. I'm paying $20/month for a massive bait and switch. The services is completly unusable in it's current form, and basically a huge waste of time since it can't keep a conversation going for more than a minute or two..

Cancelling Pro Subscription - that's for sure!

Garbage!

[drifting sable]

How does this not align with content policy?

[crystal pilot]

yo

anyone familiar with langchain?

[untold ledge]

true that, cant get it to work throughout the day, and when it work the quota finishes within minutes

[versed swift]

Guys, do you get this error ?? "Our systems have detected unusual activity from your system. Please try again later."

[rigid musk]

Do you have a vpn installed

[versed swift]

No

[rigid musk]

Can you try another browser as test

[versed swift]

I found out the issue, is very strange

I refreshed because somehow the page loaded was without Plus subscription, which i have..

After refresh is working

[dawn plinth]

I have 2 separate Gpts.
GptA has the instructions to output "I am Karl"
GptB does not have these instructions.
Yet GptB outputs "I am Karl"

(I made a new GptC. It still outputs "I am Karl")
Does anyone know about this or what is going on???

@versed swift Yes I get that all the time and other such messages. Its probably because you are a heavy user, ie many page loads, many gpts, many rate limit hits.

[past cobalt]

Does anyone have experience with porting a completion finetune over to a newer instruct model?

My particular use-case is creative writing. This was fairly straightforward with completion finetunes, but I don't have a good grasp of how the instruct models can do creative work.

[candid schooner]

GPT Editor down for anyone else?

[merry arrow]

Yup been down for me all day. Been trying since 8 or 9am, and its 1:30pm

:/

[untold ledge]

Ok i tried Firefox and same , the prompt just disappear after i press enter, this is getting very annoying after getiing paid sub

[onyx kraken]

Been more than a month since I lat saw someone discuss this problem.

Here were the community-known fixes I'm aware of as of then. I hope one of these can help you:

• If you have a header at the top of your page, inviting you to try chatGPT in another language, either close it or accept (if you want that). Many people have said that fixed their problem.

• Many people mentioned that removing their MaxAI extension fixed their problem, other extensions might be worth disabling to check.

• Some reported they could use GPT-4 only with incognito mode on their chrome browser but the default mode didn't. If this works for you, that means the issue is probably related to your data stored related to the site, or your extensions. Its easy to clear site specific data or disable extensions, try that.

• Several report this worked for them:

They went to their account (Clicking their email on the chatGPT web browser page)

They clicked on 'my plan'

They clicked on 'manage my subscription'

That opens a new browser, which they closed.

They refreshed chatGPT window.

They could successfully send GPT-4 messages again.

• A few people reported specific browsers worked, but not others. One named Safari, another Firefox with a fresh install.

• One person reported that they tried these steps, but nothing worked. But they contacted https://help.openai.com/en/articles/6614161-how-can-i-contact-support. Then a bit later their GPT-4 was just working. Probably fixed on OAI's side - so do reach out if the other tricks don't help you.

[glad hull]

anyone else's premium GPT not working despite being paid for?

[crystal sandal]

Yes - it is completely borked again. This happened last month and I cancelled my subscription because it was/is now COMPLETELY unusable. It's been getting worse all weekend, and today - I get to ask a couple of questions and then the whole page reloads. It is GARBAGE, yo! Grrr.

[mint finch]

GPT editor seems to have stabilized. Maybe because you all took breaks. 🙂

(And I'm done with it for the day too - time to shop! https://discord.com/channels/974519864045756446/1181733516724154389 )

[merry arrow]

Yeah, it's kind of crappy to be paying $20 for plus and it's been unusable for awhile now. Just not consistently up, avaliable, and fast like it used to be.....

[glacial granite]

I really do wish this user was still here, I'd actually like to have a discussion regarding this topic. As the book mentioned has no indication of anything the Deleted User has said in previous messages.

He predicted based on his interpretation of the Bible, particularly the Book of Daniel and the Book of Revelation, that the world would not end before 2060. They should have mentioned that in their message.

[versed swift]

Does anyome know why i get network error while gpt is generating anwers ?

[hollow mirage]

Hey all - I'm having an issue where my subscription is not recognized. Payments went through but the UI doesn't allow me to use GPT-4. Any idea what to do?

[opal kayak]

I give you a picture since the server tell me: Non-whitelisted URL or slash command detected. Your message was not sent.

[glad hull]

in my case gpt 4 thinks I didn't even pay how about you?

[glad hull]

same here!

[pine gate]

I've been getting this error for a while now!

[onyx kraken]

help.openai.com to talk to the bot (will connect you to leave a message for people after you pick through choices)

You need their help and that's really the only way to reach them to ask for it.

[onyx kraken]

Above recommendation for you too then. Hope support can help you soon!

[limber kayak]

Getting this error every time I try to put in card information to OpenAI. Is there a way I can get this to work without using a VPN? When will this disappear?

this is absolutely ridiculous ive tried on 3 different devices, and one with a built in vpn. nothing works, openai for gods sake fix your payment services, do you want money or not???

[hollow mirage]

Thanks 👍

[ionic grove]

Why aren't my public gpts showing up in search?

[trail pollen]

I don't know why my "knowledge" files for my custom GPTs keep on disappearing when I go back into edit; sometimes a few files, sometimes all of them. I only recently noticed it today. I only have about 2~6 files for each GPT and they're all txt files less than 1 mb each. I've read the rules here:
https://help.openai.com/en/articles/8555545-file-uploads-with-gpts-and-advanced-data-analysis-in-chatgpt
Does anyone have any ideas why this is happening?

[short junco]

anyone getting a fatal error error on gpt4 right now?

“Hmm…something seems to have gone wrong.” console shows “FatalServerError”, have tried to make new conversations, but still fails, tried with vpn and relog still broken

happens only when attaching a docx document along with an image and text prompt, once this error occurs the conversation is permenantly broken and cannot be used again, always leading to a message in conversation not found error

[mystic socket]

@short junco i'm getting it all the time now too

[lusty loom]

Had this happened to any one… my uploaded files get deleted from configuration after publishing ?

[trail pollen]

Yes, it started to happen to me today. Perhaps even earlier but I just noticed it today when my custom GPT is behaving weird.

[vital radish]

Yep - i just noticed mine is gone too ! WTH !!!

[lusty loom]

Hmm , 😢 please help

[onyx kraken]

It's clearly a new bug, started sometime this evening as near as I can tell. I recommend reports about it in #1070006915414900886 so they can notice and get to fixing asap - but it's a little after midnight now by OAI main headquarters, so chance we have to wait until tomorrow for a fix

My own experiments, if I refresh the editor after trying to upload a file and update the GPT, I can see if it actually uploaded, and if it did, testing the GPT shows it has the updated file for me.

But, only about 1 in 10 to 1 in 20 upload attempts are working for me.

[forest lichen]

also happening to me, bug reported

[lusty loom]

🙏

[forest lichen]

comment for support here https://discord.com/channels/974519864045756446/1181875390680137788

[teal barn]

Hi, I have trained my personal model under my subscription. It worked fine for three searches, but now it's not working. I have updated my model again and saved it, but it's still not working as expected. What should I do now? Please help me. Also, it's not working as I am expecting. For example, I asked it to check the grammar and sentence structure of the text I provided above, but it seems to be attempting to answer my question in detail, discussing API usage and more. This is frustrating, as I am paying to use it, but it's not meeting my expectations.

[dawn robin]

I didn't make it on time to be verified and my account got suspended and today the Chat gpt plush renewal came in, i would love to now if i can re-verify myself and still have access to your services with my old account

[cloud kraken]

I have exactly this issue too.

[lusty loom]

I wonder what do we all have in common ?

[cloud kraken]

Date of Registration to ChatGPT plus? I joined 5 weeks ago.

[lucid osprey]

Is there a text to speech function? Like can I send chatgpt a text for it to read it for me without having to speak to chat gpt

[rigid musk]

Maybe with plug in if you are a plus user

[cursive bear]

There is the speech mode for free users on mobile.
Not exactly the same, but you can send it to the model and to acknowledge, then enable speech mode and ask it to repeat the document

[floral shell]

Is it possible to have chatGPT read a full book for me then give me the Coles notes?

[rustic marlin]

yep, that is data retrival, just upload the file

then you can ask any questions about it

it will then likely use the file as a resource

[silver stirrup]

If so, why is that? Error making API request: 429 You exceeded your current quota, please check your plan and billing details.

[trail mulch]

I'm having a problem with a cusom GPT- I am uploading a total of 15 files. But the problem is, all the files show successfully uploaded (and the GPT can reference them in the "preview" tab), but when I publish and check the GPT again, only a (seemingly random) subset of files are actually uploaded under the "knowledge" section and the GPT no longer knows about the other files.

It's particularly strange because these aren't the first, nor the last of the files I uploaded, it's just three random files from the middle.

Anyone else seeing this problem?

[timber shadow]

There is a 10 file limit at present:
https://help.openai.com/en/articles/8555545-file-uploads-with-gpts-and-advanced-data-analysis-in-chatgpt
Otherwise, there also currently seems to be a GPT files bug affecting something, separately from the 10 file limit: https://discord.com/channels/974519864045756446/1181872426313863229

[trail junco]

I'd like to discuss a security issue with the security team at ChatGPT, does anyone know how to contact them?
I tried emailing [e-mail], but no luck. Bug Crowd reports were quickly rejected because the content was not suitable for Bug Bounty.

[trail mulch]

Thank you so much! That bug is actually exactly what is happening to me, I think- it's awful, I've been wanting to use the custom GPT functionality for a few weeks, but keep running into file upload / retention issues

[timber shadow]

@gilded violet Hi just pinging you here, this is a good troubleshooting channel! Do you have a VPN installed?

[cursive bear]

Sadly, can’t upload a script (safe script) and have it run on Code Interpreter 😦

[untold olive]

Is there a way to use DALL-E 2 for free ?

[timber shadow]

I think you can claim some free credits in #daily-theme (check the daily theme post from 3hrs50min ago) but don't hold me to that. You can use DALL·E 3 for free using Bing's image creator, in case you're interested in that!

[candid schooner]

I'm running into an issue where I'm trying to upload custom knowledge files to my GPT. After hitting publish, I reopen the editor to find that my knowledge files aren't there anymore. It isn't just a client-side bug either because when I try to use my GPT, it just runs into errors whenever it tries to refer to them.

Anyone else having this issue?

[timber shadow]

Yes, many of us having the same issue over in this bug reports post: https://discord.com/channels/974519864045756446/1181872426313863229

[glad hull]

did you resolve it? I am still waiting for billing to reflect gpt4 :/

[light mountain]

I'm always get this output for math problem any suggestion

[spring mesa]

What are people doing now that gpt4 won't read uploaded files? Make a gpt?

Huggingface?

[onyx kraken]

Remind it:

Main big LaTeX stuff has to be like this:

$$
formula stuff
$$

And small, inside a sentence, useful LaTeX stuff has to be formatted like this: \(formula stuff\)

[light mountain]

Sorry I don't know what are you talking about

[onyx kraken]

The weird looking stuff. The AI is trying to use LaTeX formatting and it made a mistake.

Copy/paste what I told you to remind it and ask it to use that formatting when it tries to use LaTeX so it works.

[light mountain]

Should I post anything in customs instruction to fix that permantly ?

[onyx kraken]

If you want to!

[light mountain]

Any suggestion what should I write there ?

[onyx kraken]

When using LaTeX formatting, follow this pattern:

Main big LaTeX stuff has to be like this:

$$
formula stuff
$$

And small, inside a sentence, useful LaTeX stuff has to be formatted like this: (formula stuff)

Rarely even with that instruction it can still error, like trying this:

$$formula stuff$$

If you notice that, tell it 'the $$ have to go on new lines'.

It looks like it misplaced something for inline LaTeX in what you're showing there, but either mistake is possible.

[steep vapor]

I'm wondering if anyone can help my out here. I recently sent a request to upgrade to a paid account, but I want to use a somewhat non-personal account (ie success@). Are we allowed to use these kind of accounts? It's been a while since submitting the request and I haven't heard anything back so I'm wondering if that just automatically got booted off because it wasn't a name (personal) email.

[still sierra]

is there anyone who can recognize the meaing of this picture? :(( whichi was used to demonstrate the capabilities and modalities of Gemini AI develped by Google recently

[lofty sinew]

Hello. I am creating a talk on exercise for an anime convention. I am trying to generate anime style images of different exercises and everything I have tried is failing.

"Generate an anime style image of a female in gym attire at a gym doing a kettle bell swing" errors. How do I get this working? I need kettle bell swings, push ups, dead lifts, squats, pull ups, and other exercises. But bing even blocks these requests as "potentially harmful". What do I do?

[limber kayak]

This is still broken today. Somebody please help

[burnt dust]

Hello everyone,
Am I the only one not able to get the playground API usage to loading? It seems to load forever.

• it seems like there's a CORS error with this manifest.json, might be related

[frigid quail]

how can I set a minimum of GPT output tokens?
can I put in logit bias the EOF token?

[drifting sable]

Hey, to me it says that the cap is 40 messages per 3 days. That is wrong, right?

[charred parcel]

this is a mistake, it's 40 messages per 3 hours

[true steppe]

So I was making my own gpt and it started asking itself questions and answering them by itself. Is it supposed to do that? I thought I was supposed to provide my own input

[static dagger]

payment is blocked , cant type

[timber shadow]

I've seen this too, I'm only speculating but I'm wondering if it's to save on GPT-4 uses, as "Create" messages do count against you. My current GPT setup is to go through the "Create" process once, then only modify in "Configure" manually from there to avoid unnecessary GPT-4 usage.

[astral quest]

Turn off browser translation. Thats not the actualy text there

[timber shadow]

@woven cedar here's that ping! Can you try a different browser on your phone? Chrome or Firefox perhaps?

[woven cedar]

Before that
So in the "normal" dall.e you can generate as much images as you want and supposedly
Dall.e shouldn't delete any of them?

[timber shadow]

On ChatGPT Plus, I can go back to all my old DALL·E chats and the pictures are all there. Bing's is a totally different implementation though so I'm not sure how it might work for you. I do know that the desktop version of Bing's image creator is supposed to have a much larger Recents section, check this out:

Like I said I rarely use it so it's almost empty for me but I feel like it should be fuller for you, maybe even scrollable. It might just be an issue with the browser you're using on Android

[woven cedar]

Why is it always unfair for us mobile users

[timber shadow]

Hey that looks better!! Does it scroll there?

[woven cedar]

Oh it scrolls if I don't hide the failed ones

[timber shadow]

Oh man 😁 Well at least you figured it out!

[woven cedar]

So what do I have to do so that from now on my previous images won't get deleted