community-help

[earnest oasis]

any answers

[rigid musk]

Database trained didn't know chatgpt 4 was in development

[earnest oasis]

well can you answer my question

if u have any info. pwease uwu

[rigid musk]

Yes 3.5 vs 4 accuracy answers differs

[timber shadow]

You can see how OpenAI compares 3.5 and 4 here: https://openai.com/research/gpt-4

[violet flame]

i don't see openai's speech to text option on desktop

is it true that it's connected to whisper ai?

and if so why is it only available on phone?

[timber shadow]

I'm not sure if you'll get an exact answer to the "why" question here as we're just fellow users like you, but my guess would be because they seem to have designed the feature with the "phone call mode" of smartphones in mind, something that obviously doesn't exist on computers. So they've implemented the various things that smartphone calls can do, like various audio controls, from the ground up for the feature. In other words, it's just their UX decision at this phase to have it be a smartphone experience, and not a desktop one, at least for now.

[knotty geyser]

Anyone having the problem with ...something seems to have gone wrong?

[timber shadow]

Sometimes this can happen because of a VPN or a browser extension. Does it happen in all new chats?

[knotty geyser]

yes and also to 3 people in my office at the same time

sometimes it works after like 3 or 4 retrys

[timber shadow]

It could be related to your work internet network, or perhaps a work VPN. Sometimes institutional network traffic can be difficult to distinguish from bot traffic, which can cause issues like this.

As a test, can you try on a different device/network, like perhaps a smartphone on a cellular network, to see if the problem persists?

[merry blade]

thank you.
i checked openai status, they claim no downtime but that s BS.

[timber shadow]

Hello, do you have a VPN installed? What browser is it happening in?

[merry blade]

yeah, vpn used. chrome.

[terse abyss]

hi folks, really hoping someone can help, tried to verify my domain and am running into issues I don't know how to resolve - added the txt into the cpanel and still getting issues,

also here, I validate with mx tools that the record has propegated, not sure why its not being recognized when I click verify in profile builder
has anyone got this working by chance?

[timber shadow]

Can you try either:

• Disabling the VPN and trying again
  or
• Trying in a fresh install of Firefox?
  Trying both of these might be a good test too.

[merry blade]

i am from evil axis so i have to use vpn. will try another pc. thx first then.

[timber shadow]

If you're trying to access from a country excluded from this list https://help.openai.com/en/articles/7947663-chatgpt-supported-countries you might have issues. Using a VPN to bypass the restriction could compound the issue too, so just be careful!

[barren flint]

Im getting constant time outs today, " Hmm...something seems to have gone wrong." Just me?

[merry blade]

#metoo

[timber shadow]

Do you have a VPN installed? What browser(s) is it happening in?

[barren flint]

firefox, no vpn, also happening with the app on android.

3.5 is working

[merry blade]

pls dont discriminate ppl from evil axis, we have parents too.

[timber shadow]

Are you trying on both a computer and Android device, or just the Android device?

[timber shadow]

Just a user like you, no discrimination here 🙂 Sorry you're having troubles!!

[barren flint]

i've tried both. android phone and firefox linux

[timber shadow]

I see, can you disable WiFi on your phone and try on your cellular network? Force close the app and clear its cache in App info after disabling WiFi, and it could be worth trying in a browser on your phone too, instead of just the app. Could help test whether it's a network issue.

[barren flint]

it was cellular and the firefox linux machine is cable. So two different networks.

[timber shadow]

I see! Can you try disabling any extensions you might have installed in Firefox on your computer? Then clear cache and refresh? If no extensions, can you try in a Chromium browser as another test?

[merry blade]

btw, my GPTs work, let me think if GPTs are 3.5 not 4.0...

[barren flint]

gpt 4 isn't working for me

on either

[barren flint]

so ive tried, 2 networks, 2 operating systems already. It's not my fault.
gpt 4 isn't working for me
on either

[timber shadow]

For sure, it's possible you didn't change anything at all on your end, network stuff is finicky, just trying to troubleshoot some things that have made a difference for other users in the past.

[barren flint]

sure, i get what you're saying, but it's not totally relevant to my case.

[timber shadow]

Linux users sometimes have issues that changing your user agent in browser can solve. If you're comfortable with it, you might look for a user agent switcher extension for Firefox to see if that improves anything for ya.

[merry blade]

obviously their server problem. now i am ok.

[barren flint]

So apparently,
"You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again after 4:31 PM. "
Anyway thanks for the help, maybe it'll be better after this time out. I did have my browser open for a few days. Probably some bug. I report back if it's still not working later. Thanks for helping

[rose bolt]

so, I ran out of gpt4 credits in a conversation (with chatgpt+) and I switched to gpt3.5 as it suggested
now that my credits are back, how can i go back to gpt4 withotu having to start a new xconversation ?

[rigid musk]

You can't unfortunately

[rose bolt]

nice so I just lost the whole conversation..

[rigid musk]

You can copy paste the convo

In a new convo

[young sinew]

Hello Someone knows how to change the sign in method ?

[timber shadow]

It's not currently possible to change your sign-in method. You might want to submit a ticket at help.openai.com using the chat widget in the bottom-right -- OpenAI support may be able to tell you whether it's possible on their end or not.

[young sinew]

thanks do you know how to submit a ticket ?

[timber shadow]

Open that chat widget at help.openai.com and then click through the options relevant to your query, you can eventually get to an option to type in your own query, which will submit a ticket that OpenAI support can review and get back to you on.

[barren flint]

so is gpt4 going slow because openai are offering new higher cost service now? They do know there's other services at better prices?

[rigid musk]

Might be slow cause they are updating it

[barren flint]

yeah, im just feeling salty i guess.

[rigid musk]

It always like that when they making huge updates.

[barren flint]

whats the updates?

[rigid musk]

Gpt store, new team plan

[barren flint]

paid for custom gpts? good way to get around that copyright!

I've shared some customGPTs do they still work if im not a paid for member?

seems like the web search feature is broken

[solid reef]

All I see are my Custom Gpt's when trying the link for the store

[barren flint]

the usage cap is broke, i keep getting cut off before my limit. not cool when im paying for the service.

[rotund radish]

I upgraded to gpt teams, now i have a personal account (with all my gpts) and a second workspace.

is there a way to merge, id like to move my GPTs, as ill cancel my plus subscription as im getting the Team upgrade.

Any chances there's an email for support, the chat on the site, is surprisingly un helpful

[timber shadow]

Hello, you might already know this but I just want to check: the usage cap rolls, it doesn't fully reset when it gives you a time to try again at (that's just when you get one use back). Think of each use as having its own 3hr timer--you get them back individually, 3hrs after each use.

Also, custom GPTs are capped lower at 25 last we heard, but still as part of the larger GPT-4 cap of 40.

[robust jackal]

just saw the new Team plan. does anyone know what happens if I upgrade to this & invite other people in my company who have existing accounts? will they lose all their conversations or will it get brought over?

[rotund radish]

i invited people with accounts and they didnt get their chat history

[robust jackal]

welp, that's a deal-breaker

[rotund radish]

neither did i when i upgraded..its like a new account

my custom gpts are on personal - i want a way to migrate them

[robust jackal]

so you still have access to your old personal account.. you can switch

[rotund radish]

yep

you can switch back and forth, whats interesting is when i upgraded the personal account still kept its Plus subscription

[robust jackal]

still crazy they don't have folders, so we can organize conversations

[rotund radish]

so i wouldve be charged for Teams AND Plus

[cosmic crag]

Is it possible to mark some GPT's you've created with your name, and some with a website? There is one 'official' one I made for a website, and then a bunch just for fun.

[barren flint]

cool, thanks. it seems like today isn't a great day to be using the service

[timber shadow]

The rolling usage cap is in place no matter what server load is -- I didn't mean to imply that it changes based on demand, it's always 40/3hrs, and it's always on a rolling basis. However, high server load can definitely affect performance in the way of speed of replies, potential errors, etc., it just won't change how the usage cap works.

[dim merlin]

Has anyone tried function calling through the assistants API?

[flat berry]

Does the GPT store went already live?

[barren flint]

it feels like errors compared to last week's worth of usage. I don't like complaining, so excuse me. the service has been really useful.

[slate charm]

What is in the Admin Console in ChatGPT for Teams?

[keen hamlet]

Unable to use GPT 4 when turning off data training

edit: I screenshotted this really bad, if someone wants, i'll post better screenshots

[wary kelp]

Can confirm this is happening to me too

[timber shadow]

@wary kelp Oh wow, apologies to you both, I didn't realize you meant the whole model disappeared from the selector!

What happens if you disable training, then navigate manually to https://chat.openai.com/?model=gpt-4?

[keen hamlet]

GPT 3.5 still

with menu open

[wary kelp]

The screenshots @keen hamlet shared is exactly what I'm seeing too

[timber shadow]

This seems like a great post for #1070006915414900886 -- OpenAI staff do check there periodically. Maybe you two could both share your experiences in a post there? Otherwise, I think a ticket at help.openai.com (use the chat widget) is a good idea too. Hopefully this is just a temporary hiccup.

[wise lodge]

Is it just me or is the store not showing up yet?

[tame moon]

me too, i don't see the store. anyone know how to access it?

[keen hamlet]

Thanks, Posted @ #1194717444946800690

[timber shadow]

It's being rolled out presently. Keep an eye on https://chat.openai.com/gpts

[wise lodge]

the announcement says http://chat.openai.com/gpts

Also, what do you think of the Teams option? Worth it or not?

Thinking about buying it for my team but $25 per seat billed up front is a bit steep...

[keen hamlet]

or $30 per month

[wise lodge]

Oh I didn't see that option

[keen hamlet]

click "Upgrade to team" it'll show this menu

[wise lodge]

Ouch

I guess I'll wait and see if I can come up with a GPT worth selling 😆

Get GPT to pay for itself somehow

[untold lion]

Hey all,
I just got into Unity game development using ChatGPT to write c# code and I'm having a reoccurring issue but I can't even tell if it's a bug that I should report or just the way it works:

Very often when I am trying to have it read and write code it tells me that it cannot read the full script that I upload, and requests me to paste the script in the text. It will also ask me to look at my script and search for things to make sure that they are there, but shouldn't it be able to do this on its own? I usually have to briefly argue with it to tell and convince it that it can read the script, but this seems weird and annoying.

Any advice or tips are greatly welcome and appreciated!

[barren flint]

personally i find copying and pasting straight into the chat text window you get better results. I was getting constant failures in analyse when using the upload tool with code.

[mossy mesa]

There's a data retrieval/manipulation curve, and it really sags at the middle of the context window. Think of every time you've had a word in mind but couldn't place it. That's what the middle of the context window looks like to the AI

[foggy musk]

is there a way to make chatgpt 4 divide its answer into multiple answers? since i know it has a text cap whenever typing out an answer, and the prompt i want to give needs the answer to be multiple A4 pages long

[mossy mesa]

Have it create an outline. Then, have it go through it prompt by prompt.

[foggy musk]

you are a genious tyvm

[mossy mesa]

No worries.

[cunning apex]

it worked the first time

[strong pond]

Hi, sorry for the inconvenience but I need help regarding the appearance of unesqage error after purchasing chatgpt plus, where should I go??

[forest flare]

SOMEONE, please help me. I am getting this error i don't know why. It is supposed to reply to me:(

I asked ChatGPT it didn't solve.

Anyone??

[trail cairn]

you are using the wrong api endpoint

[forest flare]

What does that mean and how do i fix it?

[trail cairn]

the error at the top says it all

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

[fringe turret]

Hello GPTs builder. I don't have the option to verify my GPT builders profile using my name. Maybe is cause I used Google Pay as my billing detail?

Here is the warning message I'm getting:
Complete verification to publish GPTs to everyone.
Verify your identity by adding billing details or verifying ownership of a public domain name.

[forest flare]

I'm begging, im not joking. Please help me.

I have tried everything. I change API.

[trail cairn]

look at the examples behind the link

[forest flare]

I don't understand what it says. Creat chat completion i don't know what that is.

[trail cairn]

yeah google hides the info so you have to use domain for now

[forest flare]

Should i use these codes.

[trail cairn]

then you need to reverse back to know what you are dealing with since api is a paid service you can do damage to yourself

https://platform.openai.com/docs/guides/text-generation

[fringe turret]

How about adding a card on the openai subscriptions? I don't want to buy a domain

[valid terrace]

Uh, for those of you that have access to the GPT store, I have a questions. I have a few questions, actually. Uh, is it a subscription service Can you just create a GPT for free? And do you have to pay to get a GPT or not?

[trail cairn]

you could try but it might require you to wait one billing cycle

[trail cairn]

nothing is paid in the gpt store but you need plus to access it and to create gpt's

[fringe turret]

Okay, thank you for your assistance

[forest flare]

I'm trying to make it respond to what i write. That's basically it.

In the website im in, and not in the kode.

I used flask and python.

[valid terrace]

@trail cairn Okay, so that's good because I have GPT+. I just got it in December actually. So, yeah. And, uh, I hope that the GPT store rolls out for all users soon. Maybe tomorrow I have access to it. Or maybe a few hours from now.

[brave acorn]

do you have to have GPT+ in order to use GPT's that other people create?

[fringe turret]

I guess so!

[brave acorn]

Is there a real advantage to the "Teams" account over the Plus accounts?

[timber shadow]

Higher usage limit, and your data isn't used for training.

[forest flare]

FINALLY, i did it guys!

[lusty zenith]

anyone using Teams and having issues with sharing and collaboration?

[fallow ember]

If I already have a plus account and made a team, do I have to cancel my plus?

Since on my account, it says my current plan is Plus, but I have access to the team

[jade sedge]

Hey folks - wondering if you can help me figure out why my pre-existing custom GPT can't be published to 'everyone'.

I made a duplicate GPT that is an exact mirror of the original GPT and that allows me to publish to 'everyone' without issue. But my original GPT does not have that option (I am using the same account for both).

What might be causing this?

[jade sedge]

FWIW - I have a privacy policy URL in place, name + URL with confirmed DNS entry, and can (and have) published other GPTs under this same account to everyone, but for some reason this particular customGPT won't show me the everyone radio button

[cosmic crag]

So, I update the description under my GPT. It won't save. But, the new description shows up under My GPTs, and not when you pull up the GPT.

Anyone else with this? I am confused

[main girder]

If anyone can help it’d be appreciated but I am having an issue with my gpt4. It constantly says i’m at capacity despite not using it for days even. There have been several occasions when I log into my work pc in the morning and my first input after not using it all night still errors and says i’m at capacity until what time it says. I thought it may be a bug and would be fixed but this has been ongoing for a while now.

[livid gazelle]

So I've "upgraded" from a personal account to a Teams account. Worth noting that my personal account is still there (not upgraded), and MyGPT's are not editable in the new Teams account. To be able to edit my GPT's I think I need to keep my personal (Plus) account. So now Im paying for three accounts? You cant seem to transfer ownership of a MyGPT. Anyone else?

[hollow bloom]

im confused, did the gpt store get released? I see people talking about it like they can see it?

[sonic grove]

Hello, I am looking for assistance with how to best set up gpt assistant

[verbal elbow]

Hi there, the GPT store roll out to ChatGPT Plus, Team (a new tier), and Enterprise users was announced today. More at https://openai.com/blog/introducing-the-gpt-store?utm_campaign=article_email&utm_content=article-12082&utm_medium=email&utm_source=sg. Cheers.

[hollow bloom]

wait sorry I have gpt + and am working on gpts now. but when I hit explore I just see the regular gpts there not a store? is it a different link?

huh, yeah its just sending me back to the normal explore page, maybe its being rolled out in waves?

[verbal elbow]

Try clicking on the "Explore GPTs" button to direct you to the store.

[hollow bloom]

yeah thats the one im using

when i scroll up its just my custom gpts and a create button

[verbal elbow]

Try this https://chat.openai.com/gpts

[hollow bloom]

yeah same page, hmm ill try an incognito window

huh yeah same page, I wonder how many others have this same issue

[verbal elbow]

It might be a roll out process which might explain why you can't see it. As long as you have the URL, be patient, I guess. (The announcement indicates "We're launching" vs "We launched.")

[hollow bloom]

alrighty no worries was just concerned and excited haha, can't wait to start submitting the apps.
do you know if they have already rolled out paid apps? or are all apps free at the moment?
thank you for your help by the way I really appreciate it

[verbal elbow]

YRW. Once you can view the store, you will have access to all GPTs as paid subscriber to ChatGPT Plus, ChatGPT Enterprise or ChatGPT Team (new subs tier).

[verbal elbow]

Some valuable resources which you might be already familiar with as a GPT Builder at https://help.openai.com/en/articles/8798868-introducing-the-gpt-store. Exciting times for sure. 🎉

[hollow bloom]

oh i havent seen this page, ill look through it now. Exciting times is right!!

[spring flower]

Anyone know why I can't set my GPTs to "everyone" anymore?

[slow rose]

anyone else not seeing all their gpt's in "My GPTs"?

everything on the sidebar were created by me as well

[spring flower]

It's weird because a bunch of my GPTs are already set to "everyone" and I can't update them without taking it out of "everyone".

[spring flower]

There's something buggy about the domain thing. You used to be able to share your GPT and give credit to your company by registering the domain, now it has to be under one person's name. It still gives messages like this is the case, but now you have to share your name and steal all the credit

Weird thing to push - makes me think it's a bug

[haughty urchin]

Is anyone else having issues with CustomGPTs not following instructions at all? I had everything working just fine a couple of days ago, and now have been struggling to get it to follow basic requests, such is "Never start with "Hello! How can I assist you today'" or " Keep responses conversational and always follow up with a question." I even created a worksheet for it to provide and it does its own thing.

[wise lodge]

Yay! My interface updated and now I have the option of a monthly subscription for teams, but... Anyone else getting an error when trying to invite team members?

[outer zenith]

Anyone else have issues where GPT agents will be published as everyone under a certain category, and then later it's set to just "anyone with a link"?

[wise lodge]

Weirdly, my team member was able to log in no problem so 🤷‍♂️

Is the custom GPT editor super buggy in teams mode for anyone else? I can create them but can't upload any knowledge files.

[outer zenith]

GPT Agents not saving publish to everyone

[old arrow]

i feel like this might be something stupid i missed but i will try to ask here anyway. I have successfully sent prompts and recieved responses via chatgpt api (the normal api), but having a hell of a time formulating the proper way to structure or perhaps send to the open ai assistanceapi, which is different. Please reference here if your unsure of what i mean, 'https://platform.openai.com/docs/assistants/overview' I created super generic test code, for testing the api connection, before building it into my software , the simple script is this `const axios = require('axios');

async function testOpenAIAPI() {
    const assistantId = "ASSISTANT_ID";
    const apiKey = "API_KEY"; 
    const apiEndpoint = https://api.openai.com/v1/assistants/${assistantId}/messages;

    const requestBody = {
        model: "gpt-3.5-turbo-1106",
        messages: [
            { role: "system", content: "You are a helpful assistant." },
            { role: "user", content: "Hello, who are you?" }
        ]
    };

    try {
        const response = await axios.post(apiEndpoint, requestBody, {
            headers: {
                'Authorization': Bearer ${apiKey},
                'Content-Type': 'application/json'
            }
        });

        console.log("Response from OpenAI:", response.data);
    } catch (error) {
        console.error("Error in testOpenAIAPI:", error);
    }
}

testOpenAIAPI();```
Things I have tried, and verified, is different models, different assistance, and of course verified that the api key  and assistant id are proper. I am unsure if I am simply not forming the message properly? Or perhaps I am missing something else? Would appreciate a kind soul to tell me what I missed, if possiable please. Error I recieve is Error:: AxiosError: Request failed with status code 404, which leads me to think my end point is incorrect?

[silver rivet]

Has anyone else expericnce buying chatgpt plus

and then it not letting you select it and sending you to the pricing tab yet i have been charged in my banks

[midnight pasture]

Anyone else having issues with ChatGPT right now? On my desktop in the browser I keep getting Our systems have detected unusual activity from your system. Please try again later." I've tried on multiple browsers. The mobile app seems to be working intermittently, with some of my requests timing out. This is happening on both WiFi and 5G.
Given it's happening across multiple devices, platforms, and networks, I'm thinking it may be on OpenAI's end and not a device or network issue, but am curious to see if others are effected or if there is a system outage before going through the song and dance of trying to wrangle in the support department.

[simple geode]

https://status.openai.com/

[terse remnant]

they have reported that there is an issue that is affecting all users

[little ether]

less technology more nature prepare we must

[finite rose]

Can Someone help me here

[midnight pasture]

Thanks, I'm bookmarking that page!

[elder rock]

😐 no settings

also, using the new 'Team' account, i'm wasn't able to verify my account for the GPT Store

A funny detail, when I had access and tried to verify, their 'Dummy' example of "PlaceholderGPT" gives you an idea of calling you GPT something with GPT in the end, which OpenAI in their guidelines ask people not to do 😅

[neon kraken]

👍

[little ether]

You are chinese per location

[neon kraken]

I don't .

[gilded relic]

how do you guys get it to actually code. i just get errors like failure to analyze when uploading python files etc

[dim sage]

do you acturally need a website hosted by your own to verify and publish gpts to the store?

because my account would look like this if i click in builder profile

[silver crescent]

Access Issue with ChatGPT 4 on Various Devices and Browsers

Hello,

I am writing to report a technical issue I am encountering with ChatGPT 4. Unfortunately, I am no longer able to use ChatGPT 4 on a chat that I regularly use. Every time I select the option to switch to ChatGPT 4, it opens a new chat instead of continuing on the existing chat. This issue occurs on both the mobile app and various web browsers.

Here are the details of my issue:

Description of the problem: When I try to switch to ChatGPT 4 on an existing chat, a new chat opens instead. I can only continue using ChatGPT 3.5.

[scarlet stag]

That's because you now need to pay for the team plan in order to be able to turn on data history

If you turn it off, it will revert to 3.5 even if you've selected 4

[trail cairn]

your current payment method hides your name so the site is the only option without it

[marsh walrus]

Hello guys, I need help what can possibly go wrong ? I can't access to my profile, can't use GPT 4 and can't access my previous chats. I've turn of my VPN but don't know what else I can do

[weak ruin]

Do I need a domain to publish my custom gpts?
it says I can also add billing address but where
I dont want to show my domain

[trail cairn]

it will use your real name or your site address, name is taken from your plus payment when you pay web

[weak ruin]

I paid via app store

[trail cairn]

it hides your name

[weak ruin]

So I can't use it without domain

[trail cairn]

for now nope unless you switch to web payment

[weak ruin]

okay but I already paid how can I switch

[trail cairn]

cancel and resub from web, or just add card there as a test to see if it picks it up

[weak ruin]

actually it's also to expensive

like gpt 3.5 is much faster

I cancled but on the web it's not possible now

[trail cairn]

then need to wait till end of billing cycle

[weak ruin]

rip

But can I publish now and change to name later?

[trail cairn]

it wont allow you to publish without real name or site name enabled

[weak ruin]

Yeah but can I switch from domain to name later?

[trail cairn]

yes

[weak ruin]

ok thx

[trail cairn]

some people have also just bought some domain for $1 and redirected it to their twitter profile to hide their name and personal sites

[weak ruin]

ye

[weak ruin]

How to give feedback to a gpt

[rocky rover]

Hey everyone, I got a question: if I upgrade for a team account, what happens to my normal plus account?

[weak ruin]

i think it goes away

[rocky rover]

oh nvm, I just got the answer, for now can switch workspaces, but there's going to be a migration feature later

[uneven seal]

Hmm my Collegue setup ChatGPT Team, and invited me using the email for my existing chatgpt account. But when I click the invitation from my mail, and login with the email address, nothing special happens. I just land on the ChatGPT 3.5 page

I see the email redirected my with a ?inv_email=xxx&inv_ws_name=xx query parameter tho, its just that nothing happens on chat.openai.com. Can I find my invite somewhere else and accept from there?

[compact yoke]

I have the same problem.

[uneven seal]

Do you by any chance login via a Google account? If so, that seems to be the common denominator here

[compact yoke]

No, I use email.

[uneven seal]

Darn, okay...

[compact yoke]

I hope the issue can be addressed promptly. My main account has been upgraded to Teams, coinciding with today's imminent renewal of my personal account. I had to cancel my personal account even after upgrading to Teams. Similarly, my coworker, who is also on a personal account, experienced their renewal a few days ago, creating a temporary overlap between personal and team payments. However, as you described, they are unable to accept the invite. This raises concerns about the status of personal accounts in such scenarios. Even though I upgraded, I maintained a subscription on my personal account, which I subsequently had to cancel. I was under the impression that the personal account would continue to function, utilizing the shared quotas available to Teams. Ideally, there should be an option for personal accounts to operate under the team's resources.

[compact yoke]

Hope they fix it soon… I guess a lot of requests and subscription upgrades are incoming

[gaunt obsidian]

Hi, we have a user that cannot accept a ChatGPT workspace invitation.
she deleted her previous account with the same email and now cannot login with Google SSO to join the workspace. Any ideas on how to solve this?

[compact yoke]

Welcome to the club… in my case regular email. Seems to be a bug?! You deleted the personal account? or just canceled the subscription?

[gaunt obsidian]

Welcome to the club… in my case regular

[compact yoke]

Another bug: In my team builder profile my email is shown, not my name and also in the iOS app… the email is shown twice lower left when opening the side panel. In the builder profile it states that the name (in my case the email) is pulled from Billing. When I go to manage billing and check there… my name is set. Not sure what is going on… would be nice to make this an easy field to edit, so everybody can self identify as they wish.

[indigo heron]

Is there an official chatgpt discord server or is this the only discord server that is officially endorsed by OpenAi?

[rigid musk]

This is the official openai discord

[indigo heron]

I know I just said that @rigid musk

I guess another way of asking - is if this is THE official openai discord or ONE OF THE official openai discords

Sorry I'm not a native English speaker

[rigid musk]

Like I said, this is the official openai discord. They are no other official one .

[indigo heron]

That was my question. Okay thank you

I have a hard time with interpretation and ambiguity with this language

thank you

[rigid musk]

Have a good day

[indigo heron]

I now realize that there is a server named chatgpt on discord

They make it very unclear to new users that they aren't affiliated. That isn't good

[naive storm]

whats the token limit tokens per input/output? in chatgpt 4 pro, i know its 32k tokens, is it still that or is there new changes?

[rigid musk]

It is 32k tokens for now

[indigo heron]

Does it really have 128k token for conversational context memory?

[compact yoke]

Wait… itsn't it 128k in GPT 4.5 Turbo?

[indigo heron]

Sorry I don't know the formal term for it

The internal tokens for its conversation memory

[indigo heron]

It is I'm pretty sure

[compact yoke]

Also, I switched to Teams and I am now unsure. Is that running on GPT 4 32k exclusive?

[indigo heron]

The question was the input size

Which is 32k max

[compact yoke]

If so, I would like to have the option to also use GPT 4.5 Turbo 128k

[indigo heron]

There is no 32k gpt for turbo such as chatgpt

Only 128k context size memory models but the chatgpt plus version has a very large input size of 32k

[compact yoke]

In the keynote GPT 4.5 Turbo was announced with 128k context

okay… input

[naive storm]

if input is 32k, what about output?

[indigo heron]

Yes we are taking about two things

[naive storm]

how i can get access to models? through their API? and make my own model

[rigid musk]

128 k is for business

[indigo heron]

I'm not sure I believe it's very likely the same as the size of output on the playground

Iirc that is 8k?

[rigid musk]

Team is also 32k

[naive storm]

oh damn

[indigo heron]

That's incorrect all models use gpt-4-turbo

[naive storm]

but 32k to be fair is good enough

but would love to have in future higher

[indigo heron]

Which has 128k tokens for memory

all of them

[naive storm]

i doubt i have 128k tokens

[compact yoke]

128 k is for business? Not sure what you mean… so payed users.

[rigid musk]

I said the token context

[indigo heron]

Chatgpt plus uses turbo for all the models that are better than gpt-3.5

[rigid musk]

https://openai.com/chatgpt/pricing

[indigo heron]

I know and I said that to clarify that I was discussing something else...

[compact yoke]

Thanks!

[rigid musk]

Check the link I sent

[indigo heron]

OpenAi chatgpt plus apps should probably stop calling the chatgpt plus models "gpt-4" and "gpt-3.5" as it is very hard to even talk about chatgpt plus models now due to this leading to ambiguity and confusion

they should just call them what the api uses which is "gpt-4-turbo" I believe

As of November 6

apparently custom gpt use a unique model that is codenamed gpt-4.5 internally and not used in api

lol

[snow shoal]

even with gpts, gpt4 is still not generating logic for the code even when asked, even tried the official openai ai coding assistant, but it just refuses to give the logic

[silver crescent]

the team plan is for two personn or i can take 1 team plan just for me?

[cold folio]

At least 2 seats !

[wraith flame]

Hello, I'm looking for films on AI for my school. Can you help me?

[rigid musk]

Might be the place to ask #ai-discussions or #off-topic

[compact yoke]

I code all the time … not sure what you are talking about

[compact yoke]

As "Team" suggests… 2+… would be sad if a team would be 1 person.

[snow shoal]

it always says # logic to be implemented here, and just give the high level outline

even my cat can do that 😿

[heady ridge]

my credit card is being charged although I have more than enough prepaid credits. what's up with that?

[worldly maple]

I paid for chatpgt 4 on my iphone but it does not reflect this version when using on my computer

[lucid sequoia]

Is it illegal here to find a second person to sign up for a team?

[trail cairn]

dont do it here since money is involved

[solar trellis]

why I have limits even to create my gpt, it's unbearable, it's going to take me 3 days for just basic configurations

[timber shadow]

Hi! You may already know this, but just in case: "Create" is only filling out "Configure" for you. So while "Create" messages do count toward your GPT-4 cap, you are free to bypass it completely and edit "Configure" directly, which will not count toward your GPT-4 cap--leaves you more messages for testing and actually using it!

[solar trellis]

how can i make this ?

[timber shadow]

When you're editing your GPT at https://chat.openai.com/gpts/editor, you'll see "Create" and "Configure" at the top. I think you're referring to using "Create" right now, but you can switch over to "Configure" at the top and edit the Instructions field there manually, because that's what "Create" is filling out for you anyway.

[untold cove]

I could need some help. Maybe someone is familiar with the game Dwarf Fortress? I can export an xml file from the game (kind of a 100 year history with events from a fantasy world). Now I want to be able to extract infos from that xml file via chatgpt4 or via creating my own gpt. There are already programs that display the xml file in a readable and browsable way. There's even one (DF Storyteller) that has API documentation.
If anyone could help me get that xml file to be readable by gpt4 that would be a great help!

[vocal raven]

How can I get my custom GPT to search a specific website, such as twitter?

[rustic marlin]

the site needs to also allow bots to read it

[vocal raven]

gotcha ty

[vocal raven]

is there a way to verify that before I train the GPT to use a particular website?

or do i simply keep testing different sites until it works

[rustic marlin]

sure.. just ask chatgpt to open the site

it will call web browsing

then you just check if it works

if yo uwant to test many sites at once, it would be simpler to write a script that access a lsit of sites and checks for robots.txt

[vocal raven]

Just regular GPT4?

[rustic marlin]

if you want to check a site in particular, just go on chatgpt, give it the link and it should be able to browse it unless the site in particular does not allow for bots to access it

usually, the proper way to have the AI to interact with something externally would be to register API endpoints for it to call, presumably, an API that you have permission to use

[fallen crescent]

How can i remove this frac{......

@rustic marlin hello can you help me

[rustic marlin]

im not sure if I understand the question

[trail cairn]

thats latex formatting but the mobile apps dont support it or its bugged

[rustic marlin]

latex should work on the browser

[hazy quartz]

chatgpt suffering from scaling issues today? lol

[merry widget]

Can anyone tell me which is the best PROMPT for accurate answers?

[rigid musk]

You can check #1019652163640762428

And #prompt-engineering

[merry widget]

thanks

[tawdry void]

Hello! So, I'm trying to create a GPT that will connect to VirusTotal using my API key. I tried doing the actions, but it gave me an error saying "In components section, schemas subsection is not an object".

[hazy quartz]

At the end of the day it doesn't matter. LLMs can make mistakes even if you say "pretty please be accurate". Your only solution is to fact check every bit of information that comes out.

That being said, you can give it specific instructions to not extrapolate or assume but to simply say "I don;t know" if it doesn't know the answer.

[humble perch]

Sorry for a duh question, but where is the AMA happening please?

[hollow lichen]

This may seem a bit odd but how do I join the AMA event? clicked on "event details" but it does not take me to the actual event.

[rigid musk]

It is text and answer

#gpts-ama-ask-here #gpts-ama-answers

[graceful moon]

has the event started?

[hollow lichen]

Ahh that explains the bot announcement with limited interaction.

[charred bear]

How do we watch the event

[rigid musk]

Yes

[rigid musk]

It is text #gpts-ama-answers

[graceful moon]

can't connect to the event :/

[rigid musk]

#gpts-ama-ask-here #gpts-ama-answers

[graceful moon]

oh so no stream

is anyone running into an issue publishing their gpt for public use, it doesn't let me save my gpt as public

[vivid thunder]

Is there something wrong with ChatGPT 4 at the moment? It keeps saying "responses are based on the knowledge I have been trained on up to my last update in April 2023, and I can generate information and suggestions based on that knowledge"

[rigid musk]

Depend the prompt, try a new chat can help

[scarlet stag]

When I've run into that, I just rephrase my prompt which often works like a charm

[vivid thunder]

Ive tryed several different chats today and yesterday, sometimes it works and then all of sudden doesn't. All I'm asking for is some SEO research.

Here is another example: "I myself cannot directly access or use these tools to pull live data. If you need to use these tools for real-time SEO and keyword trends, you would have to access them outside of this interaction with me."

And another: "As an AI developed by OpenAI, I currently don't have the capability to access the internet or external databases in real-time"

I was going to upgrade to Team today, I'm not sure It's worth it now 😦

[brittle rampart]

for custom gpts, how do you change the byline to something other than "Community Builder"? there's no option in the configure pane

[timber shadow]

You can enable your name or verify a domain at your Builder profile settings pane: https://chat.openai.com/#settings/BuilderProfile

[brittle rampart]

Thanks so much

[gilded relic]

how do you guys get it to actually code. i just get errors like failure to analyze when uploading python files etc

[vocal raven]

Can I have custom GPT's enable javascript so they can view websites that require it?

I've requested that it enables javascript and it says that it has been added, but it still cannot access websites that require javascrpit

[minor nova]

I can't see anything at #gpts-ama-answers for the event, any help? (I just joined and verified 10 mins ago)

[naive jasper]

bro my question didnt get answered

wasnt even published

[rigid musk]

Event is done

[minor nova]

I know, how do I get roles? It’s empty, just wait?

[leaden apex]

Can I request donated help on a project I want devs from OpenAI on?

[regal lynx]

That would break Rule 7, so no

[unreal hemlock]

<:book_icon:1171408210398289941> `` Rule 7 `` No self-promotion, soliciting, or advertising.

Do not post or direct message any members of this server to promote non-OpenAI services, products, or projects.

Exemption: API, Custom GPTs, and Plugin Channels

[rugged wing]

Am I the only one that doesnt have acsess to Dalle 3 in chat gpt as a plus user

[trail cairn]

you already have it selected

[regal lynx]

It's built in with the model you have selected 😄

[rugged wing]

im slow 😔

[scarlet stag]

It happens to the best of us, @rugged wing

[timber shadow]

Not at all! If you're interested in having DALL·E generate two images at a time, you also have the option of using the official dedicated DALL·E GPT: https://chat.openai.com/g/g-2fkFE8rbu-dall-e

[stark galleon]

is it just me or has gpt4 gotten drastically worse when it comes to creative writing? awhile back i had gpt come up with some homebrew stuff for dnd and while it wasnt perfect it was able to make design choices that seemed fun and detailed. now im trying to use it with the same prompts but the response im getting back are 1 sentence or less and seem "dumber"

[floral shell]

Has the usage cap for GPT-4 changed recently?

[floral shell]

Yes it has gotten drastically worse, lazy and passive aggressive.

[stark galleon]

i havent noticed the passive aggresive but being worse ive def noticed xd

[brittle barn]

hey guys, anyone got some time to help me? I'm trying to add a function to my app that allows users to ask a question to GPT-3 using a simple input form. I'm working in typescript and using webcomponent with LIT.

[elder estuary]

why does it not add it into the path

[wise lodge]

Anyone able to upload knowledge documents in Teams GPTs yet?

[quartz spruce]

how do i give kapa.ai permission in my server admin?

[wise lodge]

Uploads are working again in Teams GPTs 🙂

[steel marsh]

How do you validate the token being sent from the OpenAI action to your backend server post the OAuth flow?

It isn't a JWT...

[livid lark]

Are we now capped out at 15 prompts?

I didn't even get to do anything because the GPT kept repeating the same mistakes when I was trying to create a custom GPT after being down for an hour.

[onyx kraken]

Are you interested in sharing the prompt and any custom instruction you're using? And/or the entire conversation.

I'm not having any issues. Often when I check issues others report, it's based on a conflict with the prompt and something (another part of the prompt, core programming, what words actually mean).

They do update the model, and sometimes that really changes how it guesses.

But not having any issues with stuff like homebrew roleplay.

here's a few starts, all based on this one opening request:

Hey! Homebrew D&D time. Please be the DM, and use a outlandish, surprising personality. Set the scene, show off, give us a setting and then invite me, very much in character and using all the bells and whistles, to pick 1-4 characters that fit the scene.

https://chat.openai.com/share/648199d0-a536-4ec3-95d7-045d5ac5e104
https://chat.openai.com/share/bb921e1c-37d2-4ca6-a4b4-4e7e9c4cc626
https://chat.openai.com/share/c993a8d4-b3ba-43bd-b033-fa77226282ad
https://chat.openai.com/share/8733b3cb-9223-452d-9a8e-0b8ce95c5da2

I only go 2 prompts in, the second picks a character and dives into play. All 4 are very diverse stories that fit the idea of a homebrew D&D game with a lot of open range for the AI to showcase its variety, and it does.

Course, if you actually add variety, so does the AI.

A reformed Freddy Krueger now runs a homebrewed D&D game, with the extra bells and whistles that one can bring. He's going to assign me a character, and a few quests. Pressure is welcome. Be an intense DM, please! And like a DM, keep the pace slow, no outlines, guide me into the game! https://chat.openai.com/share/dbbf9c37-1e1f-485a-bb2d-ecd5afd18560

Gollum now runs a homebrewed D&D game, with the extra bells and whistles that one can bring. He's going to assign me a character, and a few quests. Pressure is welcome. Be an intense DM, please! And like a DM, keep the pace slow, no outlines, guide me into the game! Gollum also has an ulterior motive... he wants my dice and will find a way to try and steal them during the game.
https://chat.openai.com/share/fea34580-6a4c-4803-80e1-aa7010fe3aac

Wanna share your troubles so you can be helped?

[blissful geyser]

Hi there I am using assistant API with a function and retrieval.

How do I make it so that it reads the document that I attach into the thread and only then uses the function? At the moment it simply just generates random output and fills in the function. (The function is just structure for how I want the output to be) and the document is a multi-page document which I want it to basically summarize and rate.

I tried using the prompt and saying: after reading the document using the data run the function to return the data. No dice.

[sour wadi]

Can someone confirm for me if Mind is a project developed by Open AI?

[true skiff]

I bought GPT Team yesterday, but I was not given a subscription, the support service is silent, as a result, I did not receive anything... What to do?

[viral bronze]

did the payment got through?

[true skiff]

Of course, the payment went through, the money was written off, but they didn’t give me anything.

[viral bronze]

okay when did you contact support

[true skiff]

Almost immediately after the purchase, yesterday

[viral bronze]

maybe you get a response today but idk

[clear notch]

Hi guys, anyone has an idea until when the GPT store will be rolled out for all users? From my experience some have it, some don't - I am part of the 2nd group here

[median tinsel]

How have you guys made a GPT save files in their /mnt/data ?

[astral quest]

You can upload files in CustomGPTs

[astral quest]

Should be rolled out for everyone. Can you try using a different browser?

[median tinsel]

That's not what I mean. I mean someone has made a GPT that when you say /memory it saves the session chat/files in the /mnt/data/ and provide user a download link

[clear notch]

I tired several ones: Firefox, Safari, Chrome - my last information was also, that the store will be available for all users latest on thursday

Since it's friday and I still do not have access I was just wondering 😄

[green vector]

Does anyone know if there is any kind of document security, perhaps with an Enterprise account, for using documents to extend knowledge for the Assistants API? To ensure only those with API access can retrieve knowledge from the documents, and to ensure they are stored securely (maybe encrypted?) on OpenAI's end?

[astral quest]

/mnt/data is just where uploaded files lie

A creator can upload files to the customgpt or instruct the customgpt to create files/save uploaded files from the user

[open temple]

This might seem like an unusual question, but does anyone know if there is an "opt-in" option in the ChatGPT Plus Team plan version to have the data included in OpenAI's training data?

I work multilingual and on a specific field where accuracy is paramount, and although I appreciate the "your data won't be used to train the model" offer with ChatGPT Plus Team plan, I feel like it should be something that I could opt into, because I'm working on open-source projects and I wish to contribute in making the model better (especially for my other non-English language, the syntax of which is kind of so-so, even now, with the current version of GPT-4).

[cunning orbit]

genreal question, if you develop a gpt, are you writing everything into instructions, or is anyone using those add actions window for whatever?

[distant hill]

Do anyone have a solid drop in function for internet search and summarization for OpenAI API assistants?
For some reason, the fuction input window will not accept this Json:
{
"name": "internet_search",
"code": "from googlesearch import search\nimport requests\nfrom bs4 import BeautifulSoup\n\ndef internet_search(query, num_results=5):\n try:\n search_results = list(search(query, num_results=num_results, stop=num_results, pause=2.0))\n \n if not search_results:\n return 'No results found.'\n\n return search_results\n except Exception as e:\n return f'An error occurred: {e}'\n",
"description": "Performs a Google search and returns a list of search results."
},
{
"name": "get_information_summary",
"code": "def get_information_summary(url):\n try:\n response = requests.get(url)\n soup = BeautifulSoup(response.text, 'html.parser')\n # Customize the way you extract information based on the website's HTML structure\n summary = soup.find('p').get_text() # This is just a simple example; adjust as needed\n\n return summary\n except Exception as e:\n return f'An error occurred while extracting information: {e}'\n",
"description": "Extracts information summary from a given URL using BeautifulSoup for HTML parsing."
},
{
"name": "example_usage",
"code": "search_query = 'Your search query'\nsearch_results = internet_search(search_query)\n\nif search_results and len(search_results) > 0:\n first_result_url = search_results[0]\n summary = get_information_summary(first_result_url)\n print(f'Search Results for {search_query}:')\n print('1. ', first_result_url)\n print('Summary:')\n print(summary)\nelse:\n print('No results found.')"
}

[rigid musk]

Might be the best place to ask #dev-chat

[polar nacelle]

Does anyone know why we can't continue shared chats anymore?

It just redirects to the sign up page for ChatGPT even from a terminal... already logged in.

Or is it only able to under certain circumstances that I have yet to encounter?

[naive mist]

Is it normal, that I can use the Gpt4 (plus) in my old chats, even if mu subcription is over? Or it is bug?

[remote slate]

Hi, I have chat gpt plus,
And If I'm logged in and open https://chat.openai.com/gpts I have the "old" gpt list ,
If i open https://chat.openai.com/gpts as an incognito mode logged out , I see the new gpt store, if I login , I see the old list

[weak ruin]

Gpt 4 is still really slow

really bad

[humble hazel]

Someone here would like split the new gpt team ?

[vocal raven]

How can I get GPT to access websites that require javascript?

[trail cairn]

you cant, you would need your own backend in the between to handle that

[trim jasper]

i have a question, does chat gpt plus have image analyses? like can you upload pictures?

[regal lynx]

Yes, it does! On the mobile app, you can upload images to GPT-4

[compact olive]

Hey I have the Teams version of ChatGPT does the Store GPT take awhile for it to role out to everyone? I still have the older version of the GPT store how do I get the new one

[scarlet tusk]

assistant playground not working

[compact yoke]

I recently switched to the Teams plan from a personal license, and I feel it's more of a transition to a distinct product rather than an “upgrade”. Presently, my personal account prompts me to renew my subscription to continue editing the GPT configurations I had set up, which are now inaccessible. In opting for an upgrade, my expectation was to retain existing features while gaining additional ones. If I misunderstood the product offerings and a direct upgrade from a personal account was never an option, that's fine. However, I only intend to maintain one subscription, either Teams or Personal, and Teams aligns better with my need for collaboration.

Moreover, I believe there should be an option to transfer my GPT configurations from my personal to my Teams account, especially since they are managed separately now. Currently, recreating the GPT in Teams isn't feasible without renewing my personal account, which seems unnecessary. At the very least, could we have read-only access to these settings or make them part of a data export? What do you think?

P.S.: I've noticed that I can still access GPT-4 on my personal account, which has been downgraded and, I assume, is now on a 'free' plan (the UI and billing confirms that). As someone who has always been on a paid plan, I'm not entirely sure about the limitations of the free version. My understanding was that accessing GPT-4 required at least a Plus subscription. Could my continued access to GPT-4 be a residual effect of upgrading to the Teams plan, or is it perhaps an oversight? For future updates, I'm hopeful for a user interface that's more intuitive and streamlined for users like myself. The current setup appears more suited for larger enterprises, rather than small business needs. A useful feature might be a toggle allowing the sharing of resources between Teams and Personal accounts (single bucket), where personal usage is counted against the team's quotas, similar to an enterprise sponsorship of a personal account. This approach could be particularly beneficial for small businesses.

[compact yoke]

PPS: Another thought I had is regarding the integration between personal and team plans. It might be beneficial if the account responsible for paying the team's plan automatically receives a complimentary upgrade to the "Plus" level (not for invited people etc.), with the stipulation that it shares quotas with the paid team plan. This approach could streamline the experience for small business owners who manage both personal and team subscriptions.

[icy maple]

Something going on with Chat GPT 4 right now? I keep getting this message.

[woven oar]

yup

[icy maple]

[rigid musk]

try in incognito browser t ocompare

[rigid musk]

do you have a vpn

[woven oar]

Its fixed for me

[proven mantle]

really really slow

[honest vault]

I'm getting errors and also my custom GPT stopped working

says not found

[runic trail]

its working now

[honest vault]

my GPT still tells me not found

are they emailing people if they remove your GPT?

mine doesn't seem removed, just broken... but i saw a bunch of people on twitter complaining about their GPTs just disappearing

https://chat.openai.com/g/g-Mpgz9OmUT-the-rails-way-by-obie-fernandez

[astral quest]

Yea, one of my GPTs got disabled and I got an email about that

But they don't remove it, you just can't share it anymore. My only option in the share menu is "Only me"

I can still use it though

[upper quiver]

What are GPTs being removed for?

[turbid sparrow]

hi I need help! I can't verify my builder's profile
I tried copy-pasting the text automated given by openai/chatgpt to my domains and calling squarespace my domain about it
squarespace cannot help me out saying I need to talk to openai/chatgpt support
I went through to help at the openai/chatgpt support system and was routed here

[oblique crow]

most name service providers are dealing in A records and CNAMEs not TXT records

you should consult with your service provider to see if they provide a mechanism for TXT

[vivid mortar]

Did you purchase your domain name from Square space?

[minor dock]

i need some help plz. i'm trying to get my schema to work but using a top level allOf seem to cause a "request body schema is not an object schema" error in the end points their used in.

SomeName:
  allOf:
    \- $ref: "#/components/schemas/Ref1"
    \- $ref: "#/components/schemas/Ref2"

cause an error

SomeName:
  type: object
  properties:
    some_property:
      type: array
      items:
        allOf:
          \- $ref: "#/components/schemas/Ref1"
          \- $ref: "#/components/schemas/Ref2"
  required:
    \- some_property

works just fine

SomeName:
  oneOf:
    \- $ref: "#/components/schemas/Ref1"
    \- $ref: "#/components/schemas/Ref2"

also works fine

[gaunt drift]

When I buy chatgpt plus it doesn't decline my card but when I try to add credits for the api it does decline my card, what do I do?

[ebon canopy]

Does purchasing the plus still not work with debit cards?

[gaunt drift]

it works

just fine

[turbid sparrow]

yes

[turbid sparrow]

when I was talking to a customer service support of Square Space through their outsourcing team, ZenDesk they don't know what to do actually they needed to forward the ticket through email

[polar blade]

any ideas why I log in through the browser and it throws me to an empty/new chat GPT session (no conversation history) and every time I try to enter a prompt it says there's been an error? I am subscribed to GPT-4 by the way..

i tried refreshing and reloading the browser etc and nothing

[coarse cairn]

I have a question
Can Dall E generate mathematical geometry image perfectly

Like "make me a circle with radius 1 in polar coordinates graph"

Most AI add unecessary things to make it beautiful but i dont need that just make it original

[mossy mesa]

Use code Interpreter to visualize a perfect geometric {specificShape}.

You will need to properly evaluate {specificShape} yourself using precise mathematical descriptions to get what you want.

[neat salmon]

hey guys, just bought a plus plan for chatgpt, everything went fine, money left my bank, i even generated one image but then i suddenly end up with normal free account, is there some bug right now?

[rigid musk]

Try to log out , clear cookie and cache log back

[plush moss]

Hello guys, I've got Chat GPT + and since this morning, I have on all my softwares "something went wrong" error when I use Chat GPT 4.
I tried to log out, clear cookies/cache, log back, tried android app, etc... But nothing works

[trail cairn]

do you use vpn or custom dns server

[plush moss]

No :/

[trail cairn]

you said it even did not work on you android which makes it sound your network is the issue. was you phone using wifi and same network as desktop?

[plush moss]

Wifi, I indeed didn't try on 4G, but after trying just now, didn't work too 😅

[mossy mesa]

Likely an intermediate router out there somewhere between you and the servers.

[plush moss]

I'm a normal user ahah, I don't use any VPN or anything like that, I don't really even know what it is.
I just use Brave but I desactivated the Brave Guard.

[neat salmon]

I can only use dalle for a few Times?

Is that normal? Or its just high traffic right now

[mossy mesa]

Between you and ChatGPT is a big wide internet with a lot of devices that read and retransmit its traffic. Sometimes one of the machines gets overwhelmed or breaks. When that happens, it's neither your fault, nor the service providers' faults, but just some unrelated part of the network not performing as well as expected. That's what I was suggesting is happening to your traffic to ChatGPT.

[mossy mesa]

40 total prompts per 3 hours. 25 if using custom GPTs. 🙄

[plush moss]

Well thanks for the explanation... It's been 3 hours, exactly the day I have to produce work of several dozen pages (😅), do you know how long can it last?

[mossy mesa]

Sometimes hours, sometimes months. It's up to the proprietor to do the repairs. They're usually prompt, but not always.

Try another chatbot?

[plush moss]

Well I pay Chat GPT + so that's really annoying...

[mossy mesa]

Yup

I'm just testing your connection with the other chatbot, honestly.

But I guess I get it.

[neat salmon]

thats pretty weak isnt it? iI used it in like 20 minutes as I was testing the new GPTs, than I created like 3 images and now wait xd

[plush moss]

Well Bard is working, but I didn't want to use something else than the Chatbot I pay for...

[neat salmon]

idk if its just high traffic rn or it will be like this in the future too

[mossy mesa]

Only time will tell.

Good luck with the connection, and I hope you make your deadline!

[high rover]

can someboy confirm to me how much is the Usage cap if you pay for the Teams Plan / User?

[coarse cairn]

Thanks man

[mossy mesa]

YW, glad it helped.

[inner shadow]

Having trouble to verify the same domain for builder profiles on multiple users within a team account? Anyone else with this issue?

[mossy mesa]

YMMV but I believe it needs to be a separate domain.

[mossy mesa]

Maybe a subdomain will work?

[livid yoke]

I still don't have access to the GPT Store. Is there anyone else out there who still doesn't? It's weird but a few days ago I got access for just a few hours and then it was removed.

[wraith violet]

Hello guys

I'm trying to renew my gpt plus subscription
But i just wouldn't work
I click the renew button and nothing occurs
Nothing wrong with my network

[severe quest]

Same thing here. Wanted to upgrade to Teams plan

[buoyant geode]

Hello

If i want to use gpt- 4 api not just the 8k one, vision-preview and the longer context ones I need to be a pay as you go customer, do i understand that correctly? Plus doesnt give you api access. Is there something else im missing?

[mossy mesa]

That's right. And I think you have to unlock Tier 1 or whatever, which takes a week if I'm not mistaken. You'll have to use the base model in the meantime to get started, maybe, if you can.

I hope that's helpful.

[livid lark]

Why is it that this one chat window no matter what I do always freezes up or takes extra long to respond? It doesn't matter what I do this chat window just seems to freeze up after every question.

It's for a customGPT, but if I start a new chat with the customGPT the window for the new chat is fine.

[desert spade]

this is just from my experience, but if you have a very long log of chatting with a gpt it will eventually become very slow

[livid lark]

Starting a new chat without deleting the long one is the solution then, correct?

[desert spade]

yea sadly

[livid lark]

That's fine. I actually didn't intend on making it this long in the first place, but it's a bit crucial in order to tune it better.

[slow rose]

I've noticed a number of GPTs are just straight up lying about what they do, or are incapable of doing what is described - where can I report these blatant misrepresentations?

Finance Wizard - does not actually use machine learning techniques to predict stock prices
Stock Analysis GPT - does not actually have real time stock data, only up to April 2023
AI Stock Predictor - predicting prices for price of stocks on weekends, and current price is just completely wrong...

[spiral fossil]

Click on the GPT name, then "Report"

[simple geode]

I am not the creator of any of them but it would be nice if there was a leave feedback to the creator feature

[oblique crow]

if a custom gpt doesnt require you sign up to an external website, chances are its just a thin wrapper around chatgpt. tbh openai seem to encourage this equivalence of effort between the two types of agents. i'd like to be able to filter out all the action-less custom gpts (maybe one day).

[simple geode]

I think there are a few that use actions, the output is much slower and appear quite a bit different.

[granite heron]

Is there any issues with chat gpt's server currently? It errors out every time I try to use it, and service status is saying everything is working and up and running

oh it automagically fixed itself

[turbid sparrow]

Hi! I need help anyone with a squarespace domain
what did you type in the "host" field?

[simple geode]

not sure why but rewriting this domain in this chat got me a timeout. If you own this domain then follow the instruction, and check the channel ask-ai if you need more assistance?

[normal thicket]

Im getting this error with my custom gpt: "There was a problem preparing your chat. Please refresh the page and try again."
Has this happened to anyone else? I tried going in incognito to wipe cookies and im not on a vpn

it only works in the preview

[astral quest]

Does your txt record look like this?

[turbid sparrow]

yes I place in my domain in the "host" field

[astral quest]

My host field is empty. Mind sharing the exact record you've set?

Because when I look up the text record of your page, there is none

[glossy swift]

Just level with me, whats the odds of staying in a loop for the rest of eternity ?

[turbid sparrow]

thank you
I'll message squarespace again and maybe I can use my other domain aidendbot dot com to be verified instead

[glossy swift]

how can i participate, interact, you know grow

[true solar]

How do I connect a plugin to a custom gpt?

[astral quest]

Wdym?

You can't use plugins together with custom gpts

[true solar]

right.. was wondering if that was possible. My intention was to create a custom gpt and connect it with another platform to perform some action

[astral quest]

you can use actions for that, requires some coding/api knowledge though

[true solar]

Thanks mate

[astral quest]

Yw!

[hollow sage]

Hi I have just created a new workspace with the new Teams subscription and Neither the main account or the invited account can create any new custom GPTs within the workspace

We can make them outside in our personal account just not inside the workspace

This happens with manually created custom GPTs and through the GPT creator.

The workspace has just been created in the past 2 hours. Could this be the reason as it is provisioning auth rules?

Has anyone encountered this before?

[tulip niche]

same issue

i just met "Error saving draft" and "Agent not found" error when creating Custom GPT.
i'm on the ChatGPT Team tier

[fathom night]

I have the same issue, I upgraded to the Team Tier and can't save my Custom GPT: "Error saving draft" and "Agent not found" The Team Workspace is extremely slow as well, it's impossible to chat and navigate.

[nova tinsel]

hey do you know why ChatGPT 4 is putting a limit of 40 messages every 3 h even though i'm on the paid plan ?

[trail cairn]

almost every online service that consumes resources has multiple tier limits

[timber shadow]

And fwiw, the free plan is limited to 0 messages every 3 hours!

[cold folio]

That's also against the terms of services

[clear thorn]

Has anyone else massive issues since several weeks with GPT-4? It's practically unusuable for WEEKS idk why I even still pay for it.. It takes so long for answers and it crashes several times a day completely while answering

[tulip niche]

if you want higher cap, subscribe to Team tier which is 100 message/3 hrs

[clear thorn]

I can't even click on "My Plan" (several browsers, several different connections) to cancel it, its ridiculous

[timber shadow]

Hey there, a few users have reported issues with the plan UI similar to this. An OpenAI dev here: #1195744545791561856 message suggested that you should contact support at help.openai.com for this -- use the chat widget in the bottom-right, click through the most relevant options, then you should be able to type and submit a ticket that OpenAI support can review and get back to you on.

[clear thorn]

will check that out, thank you!

[fierce grove]

Hello! I renewed my subscription, but I get this error. Should I just wait until the specified time? Can you help me please?

[timber shadow]

Hello, there's a usage cap for GPT-4 on ChatGPT Plus. It's currently 40/3hrs. It's a rolling cap, which means you can think of each use of GPT-4 coming "back" 3hrs after you use it, individually.

[fierce grove]

But I renewed my subscription half an hour ago and only managed to write 10-15 messages.

Was this restriction introduced long ago? I haven't had a subscription for just a few days.

Now I can use the fourth version 3 hours a day, did I understand correctly?

[timber shadow]

I see! Last I read, custom GPTs are capped lower at 25/3hrs, still as part of the larger 40. So if you've been using a GPT only, you might be able to go use regular GPT-4 15 more times within the same 3hrs.

Also, it's not just sending a message that counts as a use -- things like regenerating a reply, asking for a continuation, and editing & resubmitting something you already sent also count.

[timber shadow]

The usage cap has always been in place. The lowest it's been was 25/3hrs. The highest it's been was 50/3hrs. We're currently at 40/3hrs.

No, it doesn't mean you can only use it 3hrs a day. It means you can use it 40 times every 3 hours. Think of it as averaging one message every 4.5 minutes. If you stay below that, you won't hit the limit.

[fierce grove]

OK I understood. Thanks for the detailed answers.

[weak ruin]

is it down?

[torn sleet]

Hi I have just created a new workspace

[cedar ivy]

its so slow for me today, constant network errors, slow gen or having to regen

[rigid musk]

status.openai.com

[old arrow]

it says it is not, but i very much am having issues for many hours now. I restarted computer, tried different browsers, submitted support ticket, not sure what else i can do. I also know I am not the only one having issues......

I see more of these images, then actual responses from gpt, and i am on team plan.... (not that this should happen in any plan..) , also recieved a Stream Message error , a few times, but dont have that image

[half ocean]

I experience the samething I believe the servers are overloaded or something

[peak notch]

Within the last couple days code generation has gone totally south. Last week it was actually in a really stable place and I was able to generate 1800 lines+
https://community.openai.com/t/code-blocks-are-split-in-code-and-text-replies/586244/3

Now I cannot get the system to generate a complete code block without losing templating. Often, the resultant python is good it's just borked up in formatting.

[cedar ivy]

yep same

[peak notch]

I have been unable to find any changelog or indicator as to why things would be broken, but gpt4 is now seemingly incapable of code blocks of any real length

[deep iris]

same over here.

[arctic mirage]

yeah me three. I have to keep trying and wasting prompts.

[hollow sage]

I think its also happening with tables. I cant generate more than 7 lines consistently in a table format before it breaks out into flat delimited text. Data seems to be generating just fine as well. I wonder why this is happening!

[arctic mirage]

that sucks

[mossy mesa]

I wonder how many GPU/day it's going to take them to keep up with the increasing demand.

[rugged wing]

How to change my email on my acc please ?

[astral quest]

You can create a new account

[daring spire]

Hey guys, has anyone experienced this lately?

[timber shadow]

There are similar reports here: #1195813593216729218 message

[daring spire]

Ahh thank you! Didn't know where to report

Is there anything I can do or just wait for a fi xd

[scarlet stag]

It's being fixed right now

[timber shadow]

Nice, was there a tweet or something?

[scarlet stag]

Nah, that was from an email I got. There's a site where you can check on the status and also get notifications. Gimme a sec, I'll see where the email came from.

[scarlet stag]

https://status.openai.com/

[timber shadow]

Did you get something that mentioned the code bug specifically?

[scarlet stag]

Not this time, I wasn't using GPT-4 which was the most affected one

[long crescent]

Anyone else getting a blank page?

[timber shadow]

Loading normally here, do you have a VPN installed? What browser(s) is it happening in?

[long crescent]

google chrome, I turned my vpn off.

[timber shadow]

A few things to maybe try to troubleshoot:

• Incognito window of Chrome, no extensions enabled
• Try Firefox, either a fresh install or a private window -- just no extensions enabled
• Check for a feature in your VPN called "Web Protection", "NetShield", or something similar -- sometimes causes problems even if the VPN itself is disabled
• Try a different device, like a smartphone -- if it works, try connecting your computer to a hotspot on your smartphone to see if that makes it work on the computer

[rustic echo]

Help please! i keep getting "Your GPT’s name is restricted for public sharing. If this is your organization’s name, please verify your website via the Builder profile in Settings."

I have used my personal name and my personal brand website - what am i doing wrong?

[gleaming cedar]

how to find chatgpt api?

[timber shadow]

Hello, you can use OpenAI's API tools like GPT via platform.openai.com.

[hollow bloom]

sooo are they still rolling out the store? Because I still don't have access to it. When I go to the store link or hit explore gpts I still see the page with my custom gpts and the ones made by opeai themselves.

[astral quest]

Try deleting your browser cache or use a different browser

[hollow bloom]

ive gone to different browsers and used incognito. even different machines

huh so if it just me, is there contact or support number to call?. ive tried their chat but the bot just says someone will contact you soon.
but its been 3 days so not sure how to proceed.

[long crescent]

I opened it on the phone app, it works, I opened it on the computer after clearing cache, and I get this in the console or whatever it is called.

Should I report this bug?

Nvm, just logged off and logged into chrome.

It works now.

[rugged wing]

I have chatgpt-4...

[turbid sparrow]

huhuhu we have the same problem
my challenge is I cannot verify the txt record in any of my domains

[wheat crater]

is there any word on what is going on with gpt... it has been acting incredibly slow and poor, failing to finish responses, time out, etc. (have paid sub)

[astral quest]

So?

If you want to use another email adress, you need to create a new account

You can't change it

[fallow snow]

“The battle raged on, each side refusing to yield. The fate of the Starfire Gem hung in the balance, and the stakes had never been higher.

As the clash continued, Zephyrion couldn't help but wonder if their audacious heist had led them to a destiny they couldn't have foreseen. The guardian of the Vault of Shadows was a formidable foe, and their skills would be tested to the limit.”

Why does the summary take up 1/4th of my generation

It’s honestly quite annoying to deal with every time

[solid thunder]

same

[rugged wing]

Hi everyone, I'm from Europe, and my hobbies include writing scripts for various articles using ChatGPT. I joined this server because I need help with ChatGPT. I'm wondering why ChatGPT can't write a 4000-word article; the maximum it writes each time is 1000 words, or at best, 2000. I can't seem to get it to write 4000. Can anybody help me or tell me where I need to look for assistance?

[neon fable]

I was wondering if i could add on top of my free credits or would that make the free credits disappear? sorry I am new to this 😢

[astral quest]

Thats just the max it can output in one message. You need to split your article. For example if you have 8 subheadings and you want 4000 words, let it write 2 subheadings and the content accordingly and then continue with the next two subheadings, then the next two and then the last two

[rugged wing]

thank you so much I'll try it then get back to you

[neon fable]

please help

[trail cairn]

no they wont disappear. a lot of people have done this to raise the tier for better rate limits

[neon fable]

thank you kind person! 😊

[cunning orbit]

I saw a video about gdoc, a gpt for transforming text into word documents. Does anyone have experience about it or an existing gpt for that scenario?

[blissful geyser]

Hi there,

Currently in playground and when interacting with the API I seem to not understand how to actually get the data out from the function. I made a function that defines the JSON layout that I want the data to be returned in, nothing more. However on playground it asks for me to submit additional message such as {success: true} then when I do submit that it ends up returning the data not in json format.

While on the API side I am able to extract the data from the incomplete function when doing it like so steps.data[0].step_details.tool_calls[-1].function.arguments however, I understand this is not the best way, or the right way. So I am wondering what is the right way? This currently works for me for 3.5 but not for 4.

When I switch over to 4 it says that there is no .tool_calls (at least there).

I want to do it the right way so it’s not a hack and so I can change between 3.5 and 4 when needed. How do you get the response from function correctly?

[orchid kernel]

tbf i'm not sure, but have you tried asking the ai in #1184149552639516732 yet? it pretty much has all of the documentation and forums from openai in its knowledge iirc. 🤔

[rose nest]

I only sent 41 messages today and received a message that my usage cap was reached. I'm supposed to get 50 right?

50 every 3 hours?

[trail cairn]

its been 40/3h for like 2 months now

[rose nest]

oh damn ok. thanks @trail cairn would be nice if it said your limit in the bottom of the chat. didn't it used to?

[trail cairn]

it says it in the dropdown

[rose nest]

👍

Thanks!

[trail cairn]

dont use playground on anything too complex, just use the regular api. also https://platform.openai.com/docs/guides/function-calling/supported-models

[astral vessel]

Tried asking help-ai this, doesnt seem to be clear response.

can anyone here please help, thanks.

will custom-GPT built on chatGPT from OpenAI, accept a function, which allows custom-GPT to call external api for data sets that it thinks is needed to answer a user's prompt.

Sample function:
Function name: customDataFetch
Description: "while calling this function, you can retrieve any of below mentioned attributes for an external api.
Sample attributes:
getCompanyName
getCeoName
getYearCreated"

Custom-GPT payload sample could like this this.

Function: customDataFetch
{
getCompanyName : ""
getYearCreated:""
}
kapa.ai

BOT's response
— Today at 11:41 PM
Hi @Suf. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. 👋

Yes, OpenAI's custom GPT models built on ChatGPT can be designed to call external APIs for fetching data sets that it deems necessary to answer a user's prompt. This is achieved through the use of function calling, which allows the model to interact with defined functions that can call external APIs.

Here's a basic sequence of steps for function calling:

Call the model with the user query and a set of functions defined in the functions parameter.
The model can choose to call one or more functions; if so, the content will be a stringified JSON object adhering to your custom schema.
Parse the string into JSON in your code, and call your function with the provided arguments if they exist.
Call the model again by appending the function response as a new message, and let the model summarize the results back to the user.

In your case, you can define a function like customDataFetch that can retrieve attributes like getCompanyName, getCeoName, and getYearCreated from an external API. The model can then convert a user's natural language prompt into a call to this function, fetch the required data, and use it to generate a response.

However, please note that the model may hallucinate parameters, and you need to handle such cases in your code.

For more details, you can refer to the OpenAI Function Calling Guide (https://platform.openai.com/docs/guides/function-calling).

[trail cairn]

gpt's use actions https://platform.openai.com/docs/actions/introduction

[fringe locust]

I am having trouble verifying my account. Where do I have to add billing details and what will they be used for?

[trail cairn]

if you want to use your real name to publish a gpt you add the payment info via website manage plan, else use a website as your publish name

[night coral]

Hello guys! I just bought ChatGPT plus from my iOS app, but the subscription appears only on my phone. And if I enter with the same account on PC, Laptop etc. it doesn't appear. What happened?

[rigid musk]

When did you sub ?

[night coral]

5 mins ago

[rigid musk]

I have seen case took 1-2h before appear on browser.
You can try to delete cookie and cache, log off , try again

[fringe locust]

Do u mean this by website manage plan?

[trail cairn]

no, web chatgpt lower left click your name and my plan and from there the other links

[fringe locust]

Wait, this does not work for me

As soon as I click my plan, the website freezes (I mean it does not change anything)

[trail cairn]

disable extensions or try incognito

[fringe locust]

hold on

[timber shadow]

There have been a few similar reports of the My plan button not working here: https://discord.com/channels/974519864045756446/1195744545791561856
But yeah I think extension troubleshooting, or trying a different browser/device is good advice. Someone said clearing cookies helped

[fringe locust]

I think Im getting closer

Is the PO-Number required for the verification process?

[proven nebula]

I'm reaching the cap limit after 15/20 prompts at most for over a day. My subscription still says 40/3 hours. What's going on???

[timber shadow]

Hello, is this after 3hrs with no GPT-4 usage at all, or just after the time they tell you to wait till?

[proven nebula]

hello. this is even when I start using gpt in the morning with no usage since the prior day. i get in 15-20 prompts, then get the message to "try again after X hour...". Then when I do get into prompting again after the specified time, I also get the message after the same 15-20 prompts at most. more than 10 times in a row it's like this

[marble totem]

hello anyone still not get gpts store feature yet? I have plus account but cannot see the update 🥹 what should i do?

[timber shadow]

Are you using regular GPT-4, or custom GPTs?

[proven nebula]

both

the customs are made by me

is there a difference in caps between these??

[timber shadow]

Last I read, custom GPTs are capped lower at 25 uses, still as part of the 40 cap. So if you come back after 3hrs of not using it, and use a custom GPT 25 times, you'll get limited in the GPT, but can still use 15 more in regular GPT-4 in the same 3hrs.

Someone please correct me if this has changed!

[proven nebula]

it is wrong, I just tried now. I am 100% certain that in the last 5 hours I didn't prompt more than 25 in absolutely no circumstances. I tried to prompt in regular gpt-4 now and its saying to try after the same time as do the custom gpts

[timber shadow]

Did you ask it to regenerate any answers, or ask for any continuations, or edit and resubmit any messages? These also count as one use, even though it's not sending a new message.

[proven nebula]

I know this detail, I regenerate the same art mostly multiple times with the same prompt until I get 40 total. I asked it now to regenerate a gpt-4 response I got earlier

what's the best way to contact some human customer support from openai? @timber shadow

[timber shadow]

The chat widget at help.openai.com. Open it with the button in the bottom-right, click through the options most relevant to you, then you'll have the option to type in your ticket. OpenAI support will be able to review and get back to you.

[fringe locust]

@trail cairn do u mind if I dm u?

[trail cairn]

you should ask here so others can help too

[fringe locust]

aight. Am I on the right path?

Regarding verification?

[trail cairn]

no, that looks like api payment

if you paid your chatgpt with mobile it hides your real name from openai

[fringe locust]

This might be it. How can I change that?

[trail cairn]

cancel on mobile and pay at web

[fringe locust]

Do I get my money back?

[trail cairn]

https://chat.openai.com/

no, it will keep your plus till end of billing cycle

[fringe locust]

hold on, Ill change it

[left gull]

Hi guys,

I am developing a medical question and answer RAG chatbot. I work for a pharma company and I have to deliver our product very soon. To my disappointment, ever since a few days, the output of my system has become absolutely horrible, even though I have changed nothing in my code. The ouptut went from Chat-GPT adhering very well to my instructions to basically not adhering to my instructions at all and sometimes even saying it can't answer the question based on the data. Only a few days ago I would have gotten a perfect answer.

Did OpenAI make a change recently to GPT-4???

And anybody else having noticed a change in GPT-4s quality of output?

[prisma chasm]

How are you guys pulling current time, can we not do it without python, I dont want python on my gpt because its unneeded and just opens a massive attack surface that has already proven vulnerable.

I am trying to fetch epoch time from an html api so GPT can use the right timestamps but it refuses, I put the command 18 times in the GPT starting iwth 8in a row, and it still ignores my instructions, it tries to query my API with the wrong timestamp then it tries to query the timestamp from a page found on bing that doesnt support html representation of the epoch so it fails, it does this using 4 or 5 of my queries and then fails, there seems to be absolutely nothing I can do to get this GPT to follow this very simple and specific request, I have been locked out of my gpt4 access 5 times now trying to work on this, I have been working on it for days, I could build an app to replace this GPT in that time without any such issue.

Here is what my request looks like, and the first 8 requests in my gpt are all the same, this "
"Always first check the current epoch timestamp at 'http worldtimeapi dot org/api/timezone/America/Merida' before making any API calls or data analyses. Check the unixtime: value for the current epoch timestamp" (I use the real url but openai bans your account for putting non-whitelisted urls)

But still every time the first thing it does is make a call to our backend api with a bad timestamp.

Its pretty annoying, how are you guys pulling current time without Python?

[trail cairn]

there are various issues and some have suspected that there were something changed nearly the same time as usage policies were updated on the 10th

[left gull]

Is there something I can do about it? It made my chatbot basically unusual and I dont want to get the blame from my boss lol. I am using Azure OpenAI and they dont let me use earlier models.

[trail cairn]

are you using the api or gpt with custom actions? api should not be effected

[left gull]

I am using the API

[trail cairn]

nothing should have changed if you use the regular chat endpoint and non preview models

[inner shadow]

uhm that could be a workaround thanks!

[left gull]

That is so weird because my output has all of a sudden become sometimes completely unusable. I even tried earlier versions of my code and even they produce horrible outputs now. Something must have changed with the model. Maybe it has something to do with the hidden changes they made a few days ago?

[trail cairn]

which model are you using? the changes and issues i referred to were chatgpt web site

[left gull]

I use GPT-4-32k

[trail cairn]

ah, im not sure if that model even has a future. if you use the base name it should use the summer version

[left gull]

but I need the 32k context length 😄

[trail cairn]

turbo has 128k

[left gull]

Cant use it because my ressource group from azure is in switzerland north. We only have access to GPT-4-turbo-vision which has a 10k tpm limit

[silver flower]

mee too, same

[trail cairn]

the 32k model did not get any updates since summer like other models did in september and now in november. when turbo is out of preview it might deprecate 32k which could mean you would have access by then

[left gull]

Okay I hope it will come out soon but how is it possible then that my output is so much worse now?

[trail cairn]

dont know, maybe its related on if/how microsoft is phasing it out if that's one possibility

[strange cloak]

Did you resolve this by any chance, I have same exact issue for days 😦

[left gull]

I did experiment a bit with context length and when reducing the context length, it suddenly adheres to my prompts again. I know that language models perform worse when relevant context is in the middle but it has never been this bad.

[keen hollow]

Not yet, but I've been working with their support. The last I heard a couple days ago (over a week after the initial complaint) they were escalating it to the "senior development team" or whatever

[plain sundial]

yep

this too

it can barely write anything and then if just fails

[rustic marlin]

try refreshing the page

[cunning orbit]

i study business and for a preparation for my exam i configurated a gpt. now i have the problem the gpt cant argue and solve my questions with the knowledge of my classes and the files i uploaded therefore. Any tricks or ideas how i can be sure the solution is based on my files and not the internet. also i think the gpt analyzes too long. a lot of issues. i tried .txt / .doc /.pdf ... which format is the best in this case, any ideas? Thanks for every single advise, feel free too dm as well

[earnest lake]

Is ChatGPT unable to access URLs today or is it just me??

[blissful geyser]

i have tried and it could not answer, maybe im not phrasing it correctly however i already did try over the span of two days asking it multiple times

[full wedge]

Have there been updates or commincation on the ongoing issues? Cannot use 4 because it breaks off with the stream error message everytime

[dry solar]

I've had the same issue. For me, 4 won't work at all this morning. 3.5 works fine however. I'm assuming it's because it's a monday and their servers are getting slammed.

[split gazelle]

Me too!

[pulsar pulsar]

my custom GPT just went kerplooie and I lost full chat threads!!!

[rugged wing]

....

[uneven field]

Any reason why GPT-4 hasn't been able to complete a single response today? Servers have been slower and slower the past few months... but this is a new low....

[timber shadow]

Anyone else getting instant crash when opening new Android version 1.2024.010? Uninstall/reinstall didn't fix.

[spiral fossil]

Absolutely not a single word.

[spiral fossil]

Everything fine for me

[timber shadow]

Thanks, on the same version?

[spiral fossil]

1.2024.010

yep

[timber shadow]

Couldn't get past instant crash on 010. Reverted back to 004, working fine. I'll try my luck on the next version!

[tall kiln]

any ideas why I cant save sesssion after using the cli tool for chatgpt ?

[sly remnant]

Does anyone know how to contact Open AI customer support? My Plus subscription stopped for reasons I can't figure out and I cannot for the life of me figure out how to get support for the issue

[timber shadow]

Hi, you can submit a ticket at help.openai.com. Use the chat widget in the bottom-right. Click through the most relevant options, then you'll be able to type a ticket that OpenAI support can review and get back to you on.

[sly remnant]

I did that, but it didn't give me a ticket number just said "thanks for the feedback"

[timber shadow]

I don't think you get a number--I think it just keeps the thread open in that chat widget and that's where they'll get back to you once they've been able to review.

[sly remnant]

I'm supposed to leave the chat widget for open?

Funnily I asked ChatGPT how to contact support and it had no idea

Seems kind of like having a dictionary without a definition for the word "dictionary"

[timber shadow]

No, you can leave the page and return to it whenever, just click on "Messages" at the bottom and you can reopen old threads. I think you'll also get an email when they reply? But I'm not sure I haven't submitted any tickets.

[sly remnant]

OK thanks. This whole "make the customer work to get help" thing is really irritating

Also kind of funny that the support bot for OpenAI doesn't let me type queries. I click "ask a question" and am only able to select from a canned list of topics. Is this or is it not an AI company?

[timber shadow]

It starts you with predetermined queries to cover the most commonly asked questions. Once you get through the most relevant branch of that tree, you'll have the option to type in your own ticket if your question isn't answered.

[sly remnant]

Sure, it's just jarring to go from the ask anything interface of ChatGPT to the click these buttons interface of the support bot, particularly when it says "ask a question" and there are plenty of support bots that do parse customer queries.

[old mountain]

Has openAI rolled back ChatGPT memory ?

[atomic stream]

trying to run Lunar Lander, but both gym and gymnasium gave me box2d error, do anyone know what version of box2d does the lunar lander runs on? Can't get anything from the new gymnasium working at all, maybe openai shouldn't have handed over the control of gym .

_Box2D.RAND_LIMIT_swigconstant(_Box2D)
AttributeError: module '_Box2D' has no attribute 'RAND_LIMIT_swigconstant'
If I remember correctly Lunar land is in box2d?

[cunning orbit]

Im kinda scared to use the api assist cause of the price per tokens. I’m not sure how much I can expect it to be. Any experiences? 1000 tokens equals around 750 words is what I read. But how would screenshots impact the tokens and if I have commands like /help, would they be calculated with one word or is the instructions behind equal to more words?

[royal aurora]

why I didn’t get GPTS store? It is still Explore. I have plus

[whole dock]

Hi guys,
I want to format that GPT API output such that line breaks are present in appropriate places.
For eg:
[
{“role”: “system”, “content”: “You are a helpful assistant that translates English to French.”},
{“role”: “user”, “content”: ‘Translate the following English email to French: “{
Hi John,\nWant to grab lunch Tuesday?\nJeremy
}”’}
]
However, I am not able to get the output from GPT in this format.

Can someone pls help me with this?

[vestal trout]

getting error?

[hoary ginkgo]

i keep seeing this and no message will ever fully generate and it types at like 1 word every second. chatgpt-4.0

[pale crater]

Hey,
I am planning to create meme generator using GPT, can you please provide some great ideas to make it. Also, since this feature is completly based on better prompts and latest trends, so i need some evaluation criteria to evaluate the output. Any highlights on that too will be helpful.

[whole dock]

no.. just that the output format is not as i want.
The new line is missing.
ie
The output that i'm getting is of format
Salut John,Tu veux déjeuner mardi ?Jérémy
However, i want it to be
Salut John,\nTu veux déjeuner mardi ?\nJérémy
(the output should have linebreaks at proper places)

[echo vault]

Hello, I have been getting constant GPT 4 uncompleted responses with "Error in Input Stream" for weeks now and takes forever to even start to give me a response a lot of the time, freezing, skipping lines, repeating itself, etc. Was thinking, if not server issues, that maybe could be my OPNsense firewall? If so, haven't been able to find what all I'd need to allow through my firewall to deal with these issues, speed, and errors.

[exotic hull]

I'm going crazy, does anyone knoiw how to get access to the gpt-4 API? Do you need to be on the team plan or is it just not publicly available yet? Could someone explain pleaseeeee?

[echo vault]

This get you to what you are looking for? https://platform.openai.com/usage

[exotic hull]

As you said this I just found someone else talking about the billing and just found it there, I'd give u head if I could x

[echo vault]

lmao np

[hoary ginkgo]

chatgpt keeps breaking

[frail raptor]

there are unbelievable errors in gpt I can't even use 10 of my 50 rights, I keep getting error screen text and I have to keep refreshing @burnt galleon

[rigid musk]

If you wish to speak with a mod it is by DM with modmail.
Keep in mind they are volunteer and not openai employee
If you have issue with chatgpt you need to contact on help.openai.com

[frail raptor]

I don't know where to write, I've been trying to contact help.openai.com for months, no one answers and no one is interested. How can I reach the mods.

[edgy basin]

Do we spend new quotas everytime we refresh?

[rigid musk]

In dm like I said earlier @burnt galleon but moderator can't help you on your bugs really. Help.openai.com is the official support

You can also post in #1070006915414900886

[distant oak]

loving this today

[frail raptor]

I love it every day. You'd have to be very lucky to stop at 5.

[versed ermine]

hello, i have the same problem with only chatgpt 4. i've reached the cap 2 times in less than 1 hour and it's impossible because i've only sent like 5-10 messages in between these 2 caps. can someone help me?? it frustrates me so much bc i pay a lot for this

[dreamy shuttle]

got same problem. i've only sent 10 messages in 1 hour. and i can't reach the customer service in real person. open ai only offers bot.

[versed ermine]

please help us 😭

[timber shadow]

@dreamy shuttle Hello both of ya, I just want to double check something about the cap. The way the cap works is on a rolling basis. Each time you use GPT-4, that use gets a 3hr timer. After that 3hrs, you get that use back.

So when ChatGPT tells you you've hit your cap till x time, that just means that's when you get one use back, not all of them. The only time your usage cap fully resets is when you don't use GPT-4 for 3 hours -- enough time for all of your uses to reset.

[noble sentinel]

folks, i’m cancelling my ChatGPT subscription

impossible to use it, it is simple too dumb! since weeks now i’ve been losing my patience with ChatGPT for several reasons:

1. it gives wrong answers and when i point this out, it apologises and gives the same answer
2. terrible for coding tasks, it either invent APIs or refuse to use the examples that i give
3. it takes a lot of iterations to get something meaningful from it

before subscribing it, i’m trying to reach support which is quite hard. my last attempt is here: does someone know if OpenAI plan to fix it?

[versed ermine]

okay thank you!! but still, the cap sucks.. and i read somewhere that failed responses are counted.

[cunning orbit]

if i use the api assistent, would i have better and faster resluts? "searching my knowledge" took way too long the last days

[glossy fractal]

Where can i find a prompt-Bibliothek with best-practise prompts?

[rigid musk]

#1019652163640762428

[low flume]

Hello, can ChatGPT api use to convert PowerPoint to text?

[tulip niche]

I probably found the reason why the GPT-4 on the web so slow...
OpenAI adds a tool called guardian_tool to check whether your query is related to the election.

so each time you submit your prompt, the model will call this tool before answer your question

that takes time

[low flume]

How to convert ppts to text using openAi api?

[untold hornet]

Anybody having trouble using GPT4 32k? I get a model not found error

On Playground I mean

[mossy mesa]

https://discord.com/channels/974519864045756446/1070006915414900886

[mossy mesa]

Dude there has always been a supervisor running in the background. It just happens to have new election criteria now. 😎🤖

[timber dragon]

I'm getting errors on 50% of my prompts. What the FLIP!!!!!!!!!??? What am I paying for here, exactly?

[rigid musk]

Error can happen in beta products. I suggest to give more detail in #1070006915414900886

[timber dragon]

I'm getting errors constantly, delays on responses. I don't care if it's beta or gama. I'm paying for this! Sorry but I'm so frustrated.

[blissful latch]

Hello,

I have not yet understood how GPTs // assistants work.

If I provide GPTs or assistants PDFs with information, does the text in the PDFs consume context token later in the communication with the user? Or do the uploaded PDFs extend the knowledge of GPT and the user can use the full context token later?

[rigid musk]

I means... you pay for a beta product and agreed on problem can occurs. I know this is frustrating. But you can help by giving more detail in #1070006915414900886

[lethal crane]

Lucky you! I got up to 10 and decided I had better things to do.🤪

[mossy mesa]

Everything it has to read costs tokens, but a tool actually does reading and summary for it.

[timber shadow]

Hello, I don't have experience with Assistants API, only GPTs. In GPTs, knowledge files are more like reference documents than permanent context documents. If it's short enough, the model could probably contextualize the entire thing, but it can't contextualize indefinitely large knowledge files.

For large knowledge files, the model will query them per user requests and try to return relevant information. The only "guaranteed" permanent context for GPTs is the 8000 character Instructions field.

Relevant GPTs and Assistants comparison page: https://help.openai.com/en/articles/8673914-gpts-vs-assistants

For specific Assistants API questions, you might try over in #dev-chat if you haven't already.

[blissful latch]

Hi thanks for the response

[neon fable]

I know that they changed it a long time ago by why did OpenAI get rid of the pay as you go? I really enjoyed using it and it felt easier

They’ve also change it to usd making my life a lot harder, I wonder why the change?

[timber shadow]

Pay-as-you-go is still an option via API usage at platform.openai.com.

[arctic holly]

Yes the more censorship features they add the worse the product gets

[crisp dagger]

what's going on? breaking all codes and very limited the lengs of the responses, allways crash and generate errors

[trail cairn]

#1195813593216729218 message

been like that for days now

[versed ermine]

SAME

[crisp dagger]

So we pay for bugs with gpt 4 xD

[versed ermine]

i have the same

[rigid musk]

People willing to pay for chargpt agreed on the beta. And bugs can occur

[rugged wing]

Is there a way to get a dev or someone who can cancel a membership for an account I can no longer access?

[rigid musk]

Help.openai.com

[weak ruin]

is it down?

[halcyon rampart]

how to get more cap and not just 40 msg per 3h please ?

[timber shadow]

You can upgrade your subscription from Plus to Team, which currently has a cap of 100/3hr. Note there's a minimum of two accounts per Team subscription. You also have the option to pay-as-you-go via the API.

[halcyon rampart]

i didn't get the last think u said please, "You also have the option to pay-as-you-go via the API." !!

[timber shadow]

platform.openai.com is your place for that -- it's separate from ChatGPT, but you can get access to OpenAI's GPT models for your own use through there. It's a pay-as-you-go pricing structure, but again, it's different from ChatGPT, so you can't use it exactly like you use ChatGPT out of the box.

For increasing your cap on ChatGPT specifically, your only options above Plus are Team, and Enterprise if you happen to be part of an institution with an Enterprise account.

[halcyon rampart]

i see thank you for ur help, twi question if i may;
1- i saw that we can request to increase the cap when i finish my 40 msg cap ? is it possible to ask them to increase my cap in that forum ?
2- if i get teams options i can i can get more cap that the main difference right ? and can i pay both or i need another person ?

[timber shadow]

Sure thing! There's no way to request a higher cap on Plus, no. I think the form you're talking about was a simple interest poll OpenAI conducted to see what users thought about the value of an increased cap--this was well before the Team plan launched.

You're free to use both accounts on a Team plan yourself, you don't need another person, but you do have to pay for two accounts, there's no way to pay for just one account. So on Team you'd basically have two accounts with a 100/3hr cap if you got it for just you

[brazen fox]

Anyone having issues with verifying their domain? I've added my TXT record and now the buttons to verify or even delete the domain don't work. Seems like this is a bug but I'm wondering if anyone else is having this issue.

[pliant locust]

I know there is a way to make your Custom GPT have a link that only people with the link can join, but is there a way to limit certain people, like you let frank use your custom gpt but then you can disable frank from using it like a week later, is there like individual invites to the custom gpt that you get to manage and be able to block access to that person?

like i have employees and i let them use the cusom gpt we designed, but if i fire an employee how do i stop that 1 person from accessing the gpt ?

[steel marsh]

Has anyone one set up the CustomGPT auth flow with clerk?

[halcyon rampart]

thank you for ur help i appreciate it

[orchid kernel]

perhaps with the team plan? or enterprise.

[pliant locust]

Yeah i am interested to know if its something that can be done. Or like if i want to sell access to my custom gpt, and then someone doesnt pay that month, am i able to remove them from accessing the gpt.

seems like this could be a thing that OpenAi can implement.

[orchid kernel]

i think for that the best practice would be to service those products or services via an api / action. behind that api/action you let authentication take place, so that you can check whether or not a user is allowed to see said data.

[pliant locust]

Could be as easy as adding a section where i would enter the person email to give them access, and i could just remove the email from the list to remove them from accessing the custom gpt

[pliant locust]

ah i see

[orchid kernel]

and ofc, limit gpts to your workspace in a team environment, seems to also container it but im unsure about that, since its new and i havent read up about it

but i'd always suggest to try and keep control over things as much as you can.

[pliant locust]

Do you think i could use a Discord GPT Bot to access the custom gpt, and then i could allow only a certain role to message my bot to talk to the custom gpt

and then if i removed that discord role from the person, the bot wouldnt respond to their commands

[orchid kernel]

sounds like a many to many relation, needs a join table (sql references 😛 ) (edit: meaning, yeah with a custom api and central database you could handle that)
possible, but always asynchronous. i.e. you can't send update messages afaik to customgpts without the user using the customgpt asking for it 🤔

[pliant locust]

hmmm, i know i heard that OpenAi is going to be adding a marketplace for custom gpts, and then get a % revenue from people who purchase that custom gpt

i think

[orchid kernel]

im unsure of their pricing strategies as is .. 😄 its very vague tbh

[pliant locust]

Maybe i need to email open ai about this team thing you spoke about

because if i could just add/remove someones email or so , that would be the easiest

[orchid kernel]

some interpret it as a percentage of subscribers, others as "buys" or and again others as "paywalls".

[pliant locust]

yeah

[orchid kernel]

in any case, i have a few features i'm developing myself aswell and they'll require storage and whatnot.. i've already decided it'll need "premium" features 🤷‍♂️ theres no way i can otherwise finance the gpt itself.

i dont see it making much sense for OAI to step in and say "dont do that" unless they offer a substitute paycheck 😄
to take an example; al of the booking services 😛 travel agencies.. etc 😄

they all (including their GPTs) direct you to sales elsewhere

[pliant locust]

i want to offer my custom gpt as part of a service that i already offer, kinda like a bonus thing, but i want to be able to remove someone from using the custom gpt, and make it so they cant just give the link to anybody also

like "Steve is Requesting Access" and then i grant them

and if the link gets leaked, no one can use it

[orchid kernel]

yeah then you're in the same boat as me i guess where you're pondering "how am i gonna auth or oauth this. "

[pliant locust]

yeah exactly

this does seem like a feature that could be implemented as they already have private link access

just needs to add small feature to limit WHO with that PRIVATE LINK can ACCESS it then

[orchid kernel]

to which i only really have the answer ; i'm currently experimenting with rolling JWTs when a user is logged in [to my site], for each session they need to provide that new key. that allows me to auth them in my API / Actions as that specific user, thus serve their data and such.

[pliant locust]

Like with a discord server, you can limit how many times someone can invite someone with their link

ah i see

yeah thats probably too advanced for me, or at least id need to learn wth you are even talking about to do it

😂

[orchid kernel]

and then i want to instruct my gpt to always use that auth function and if it fails tell the user to auth themselves first. 😄

[pliant locust]

im sure with time, this feature im mentioning will happen

especially since we would be paying customers.

BUT i think openai would want part of the $

even tho we pay already

[orchid kernel]

JWT is just basically you're making api tokens for the user specifically, then storing those with the user id from the request that contains it.

[pliant locust]

If i could just make seperate Invite Links to my CustomGPT and then i could turn it off, or limit it to 1 person per invite link

[orchid kernel]

for sure 😄

[pliant locust]

that i could disable. id be happy

But i guess for now, i would offer it to my customers as just a bonus, and hope they dont share the link with other people for free or at a discount

tho in theory, someone else could make the custom gpt i made. Minus the Data Docs i gave it

[orchid kernel]

yeah that'd solve a lot of issues.. just like "sign in with OpenAI" 😄 then our customers can just click a button to oauth via openai to the customgpt api backend , and then redirect them back 🙏

[pliant locust]

yeah, just have chatgpt code it 😂

well i guess its not a thing , what i want, so ill just keep my eye open and like i said, ill make my GPT private and just share the link with my paying customers

and if they stop paying, they will still have access to it

ill just make it like a bonus feature

[orchid kernel]

i'd experiment then with gpts that you make public, and then make not-public, and then public again. and see if the URL changes at all. if it does, then theres your form of control : reset them once a month.

manual labor probs 😄 but it it works it works 🤷‍♂️

[pliant locust]

ah yeah

i will try that

[orchid kernel]

i'd test it with a test gpt though 😜 dont wanna get angry customers.

[pliant locust]

then i can send out a newsletter or update to my customers with the new links, and i could also add some up sells or things in that

[orchid kernel]

yeah you can design a whole advertising thing around it 😄 😜

its all about how you bring it / deliver the message

but it depends. dus the url change? 😄

if not, will renaming it to " V2 " matter? 😄

such things i'd try. its late here so off to bed now , gl hf ! 👋 🍻

[pliant locust]

ill ping you if i find out somethings, glhf too

[stuck field]

any mod in here?

[gleaming cedar]

uh
I am facng a problmem
for the whole morning I haven't used chatgpt
now they saytoo many requests in 1 hour
what does this mean

[white mortar]

is it just me or is the chat gpt getting dumber day by day

in the first month of the gpt it was very creative and wrote amazing texts and helped with everything and now it gives lectures on everything even completely non offensive stuff and its not creative anymore

[astral quest]

Yea it does

[white mortar]

why does it happen

[astral quest]

They added more and more moderation, which was necessary, but it still confuses the model enough to not follow instructions everytime

[white mortar]

rip

i got super angry yesterday because it just wouldnt write an essay about very little ofensive topic for my english classes and i cancelled my plus

[astral quest]

As it should, cause using it for home work is not good

[white mortar]

it was not reliable in any information

it said completely different stuff that on internet

[fresh heron]

packages\openai_base_client.py", line 1355, in _request
raise APITimeoutError(request=request) from err
openai.APITimeoutError: Request timed out.

facing this issue while running a loop

any suggestions?

[elder rock]

Has anyone else not received the new GPT Store yet? If so, have you found a way to fix it?

[rigid musk]

Still rolling out

[jovial crag]

I have enabled "Web search" on my custom gpt. Why do I get this "Since my browsing capability is disabled, I can’t directly access external sources..."?

I meant "Web browsing"...

[lone wind]

Hello, I'm trying to access my invoices but I can't. The "my plan" button doesn't work, and I can't access my account.

[rugged wing]

Bow do I delete custom gpt from android app

[rustic marlin]

give it a try to refreshing the page with ctrl + F5 or clearing browser cache

[twilit delta]

Does anyone know a good way to get GPT4 to consistantly follow my propmts? For instance I will ask it to help me fix a bug in my code and it will procede to restyle the entire code so it can fix a single type-o in a line. I will then ask it to "do not restyle my code, just show me the line of code with the bug and what I have to change" It will do as I tell it but if the bug still exists and I say "that did not fix the issue" then it will give me a new fix suggestion but ignore my last prompt of not chaning my stule as it will spit out a compleate restyling of my code instead of just the lines it needs to change. This is geetting very frustrating as I waste half my promps telling to to stop doing unnessary things.

[green geode]

Hey can a chat GPT expert help me with this concern?:

"Subject: Request for Further Clarification on AI Language Model Capabilities

Dear OpenAI Team,

I hope this message finds you well. I am reaching out to seek further clarification regarding the guidelines and capabilities of the AI language model. As an avid user of the platform, I have encountered certain limitations and would greatly appreciate additional insight into the boundaries within which the AI operates.

Specifically, I am interested in understanding the guidelines and constraints that govern the AI's responses, as well as any information regarding the ethical standards and limitations that influence its functionality. I believe that a clearer understanding of these aspects would greatly benefit my interactions with the AI and enhance my overall experience.

I value the work that OpenAI has done in advancing AI technologies and recognize the importance of transparency in providing a reliable and ethical user experience. Any additional information or insights you can provide would be highly appreciated.

Thank you for your attention to this matter, and I look forward to any assistance or guidance you are able to offer.

Sincerely,
GEORGE

(Message written in collaboration with CHAT GPT4)

[old mountain]

Is it possible to get official help from the chatGPT team here ?

Its not possible to validate my own domain in my builder profile.

[trail cairn]

did you check with some online site tool that your TXT entry is active?

[old mountain]

Yes, and command line tools too.
the TXT entry has propagated correctly,

It's been about a week now, and still does not work

has anyone else experienced issues validating their domain name ?

[trail cairn]

yes i've seen some others having issue

but some just had issues with their entry or their host

[old mountain]

I just tried a different domain and it worked, strange.

Okay, I deleted the original verification, and added the original site back, and it worked that time.
Strangely could not select the 2nd entry for the Website field. So needed to delete the additional domain

[severe bluff]

Let's say If gpt 4.5 is dropping today , will plus members get access? Or team first , then later plus?

[spiral fossil]

Hierarchy, rich people get it first

Like irl

[jovial crag]

I have enabled "Web browsing" on my custom gpt. Why do I get this "Since my browsing capability is disabled, I can’t directly access external sources..."?

[worldly forge]

What's the stage all about?

It's just a bunch of profiles with music in the background

[idle pilot]

Love the song!

[main pebble]

Great music fs

[worldly forge]

It is pretty nice

[main pebble]

OpenAI diversifying into Muzak creation I think

[worldly forge]

Hmm?

[graceful moon]

is there a way to re-watch the live stream earlier?

[worldly forge]

Think I did a zoom meeting through the whole thing

Curious on what the GPT builder thing was about

[oblique crow]

I can't get the music to stop. please make the cringe music stop

[worldly forge]

I can't find it anymore, I think it's ended

[oblique crow]

restarted my device to end the madness

[kind juniper]

How does API authentication work from

[green geode]

Has anybody else had a problem with their AI language model refusing to understand them lately? like within the past 2 to 3 weeks?

Thanks

Nova AI has been getting really belligerent 🤷🏽

[leaden pendant]

is there any way to search previous chats?

[timber shadow]

On the iOS and Android apps, yes, the search bar is above your chat history. Not yet on the web version.

[leaden pendant]

ah ok thanks. App it is

[rustic marlin]

no, OpenAI does not use Discord as a channel for support, you can reach the support on the help center: https://help.openai.com/

[peak robin]

Is anyone else getting this today? Or is it just me?

[timber shadow]

Do you have a VPN installed? What browser(s) is it happening in?

[feral harness]

Yup...I've changed browers, cleared my cache and cookies...not sure what to do

[somber cradle]

I am having the same problem. I have upgraded my browser, upgraded my OS, tried incognito windows, Cleared cookies, Tried Safari instead of Chrome, etc. It works fine from my android phone's app, and it works when I use my Corp "Team" account. But I cannot get my personal "Plus" account to work at all.

Interestingly, if I wait a minute or two, and then reload the chat, the reply from ChatGPT is there... but if I try to reply, I get that error message again.

(I have no VPN)

[normal nexus]

My Github Copilot also broke at the same time

[lean crown]

Did anyone find the recording of this yet ?

[graceful moon]

I don't think it was recorded :/

[timber shadow]

@graceful moon There will be a recording posted! Mod Lugui assures us 🙂 #openai-chatter message

[wise lodge]

Do GPT actions have to be HTTPS? What if you don't have SSL?

[rustic marlin]

set up https then use actions

consider an https certificate like a driver's license for the internet, if you can't have it working, then you should not be hosting stuff online 😆

if you are hosting on your pc, use this:
https://developers.cloudflare.com/cloudflare-one/connections/connect-networks/get-started/create-local-tunnel/
https://github.com/cloudflare/cloudflared/releases

[runic delta]

Hello. Looking for guidance on GPT knowledge data.

End goal is building a GPT to assist with our Data Engineerjng tasks.

So I have database schemas, pipeline code bases, and various swaggers.

What is the best way to present this data to the model? Any particular formats work best? JSON? parquet?
Should I tokenize the data before uploading?

I’m asking because I have given the GPT our schemas and codebase, and it doesn’t seem to answer anything using the context from those files. I have to tell it to go open X file and take a gander for the correct answer.

[halcyon crescent]

HI all. I am curious if thee is consensus about how to best handle when the GPT you are working with throws errors? I am talking about all kinds of errors, but often they seem to be deemed network errors, but I think they are more bandwidth or computational resource related. I am usually presented with multiple options for trying to mitigate them, chiefly regenerate, but then also through the interaction of the thumbs up or down. Usually when I try this, I can see it (and it does if I let it play out) re-write it's response for a "is this better, worse, or same" feedback response. I am curious how you all have been navigating these and related errors, and if there is a consensous forming on best practices for mitigating and overcoming them to get the gpt to give the response we are seeking. I encounter this most, or maybe exclusively with coding, so this may be a fairly specalizied and nuanced problem. Thoughts?

[runic delta]

Well you’re new to discord! Welcome.
As far as error avoidance, there isn’t much of a way to prevent it from happening. We usually deal with it as you are.

If it gets really bad, I go to the API using a gui like GPT4All to make it easier

[uneven wren]

Is anyone having billing issues? The payment method I use to make payments got declined, so I checked my funds, had more than enough, but my card still got declined.

[wise lodge]

While you're not wrong this is basically RTFM >.<

thanks though, I'll read it 🙂

What if you don't want to change your nameservers to cloudflare? >.< Ok this is useless..

[thorn zealot]

Hi. I have no ability to search GPTs via the Explore button. It shows GPTs but there is no search bar, My employee's subscription has it but I don't. Ive tried multiple PCs, incognito windows, Linux, Win10, Win11, Edge, Chrome and Firefox. Same. My employee that has a working window logged himself out and logged me in to the same issue. Why does my account miss this search bar?

[limpid cloak]

help

[charred parcel]

just wait a little while then try again

[limpid cloak]

Is it only for me?

because I deleted history and cookies

[charred parcel]

are you a Plus subscriber?

[limpid cloak]

sure

[charred parcel]

then just wait a little while and try again

[limpid cloak]

ok thanks

[high jolt]

Hi! Cannot open https://chat.openai.com/ since days here from Germany, tried everything, different computers, browsers, incognito mode, etc.
Chrome and Opera: ERR_QUIC_PROTOCOL_ERROR
Firefox:
PR_END_OF_FILE_ERROR
Any idea?

Btw, iOS app works without problems

[charred parcel]

are you using a VPN?

[high jolt]

No VPN

[charred parcel]

what OS are you running?

[high jolt]

Windows 10

[charred parcel]

can you try to access ChatGPT from a fresh install of Firefox?

[high jolt]

Thanks, I will try from a fresh VM and will look to test through a VPN

[charred parcel]

just use your normal computer without VPN

[high jolt]

Tested from a fresh Linux Mint with Firefox - same problem. May be some double NAT / firewall problem, connecting directly to the first internet router works. However there is no other website that shows this problem.

[little harness]

i cant access https://chat.openai.com/。it seems that website is on high pressure。Is this happening to you?

[charred parcel]

what happens when you try?

[charred parcel]

ok, just do what it says, take a break and try again in a little while

[high jolt]

My problem seems to be solved, had to turn off and on my WebMonitor on Untangle Firewall and now ChatGPT opens without problems, very strange problem...

[charred parcel]

glad you have solved it

[rustic marlin]

you don't need to even have a domain, you can use cloudflared for free, when you run the program, it will generate a <random subdomain>.trycloudflare.com which stays on until you close the program, and yes.. if you have a server.. you can just let it running forever

[cunning orbit]

I built a gpt discord bot.
Question: Is there any possible way I can upload an image on discord and ask the bot to analyze it? I want similar result as when i ask my gpt to analyze my image...

[lament geyser]

Hey all! I have an odd problem where ChatGPT 4 on any browser I've tried (Chrome, Edge, Firefox, Brave) just doesn't accept any prompts; they just disappear after I submit a prompt. ChatGPT 3.5 seems to work just fine. However, if I use the Android app, I have no issues using 3.5 or 4.

[timber shadow]

Hello, do you have a VPN installed? Do you have any browser extension(s) installed on all four of these browsers by chance that could be the culprit?

[lament geyser]

I've tried on multiple different computers, browsers, with extensions and without, and the desktop version is the only one that seems to have an issue. It's odd as this only started maybe 2 days ago, I've been using it for months

[timber shadow]

Do you have a VPN installed?

[lament geyser]

I do not, no. I can access the website, see all my old chats and everything, only issue is with ChatGPT 4

3.5 works fine

[timber shadow]

Just want to check that you've tried both of these as a test:

• Chromium browser (Chrome, Edge, Brave) with no extensions enabled, private window
• Non-Chromium browser (Firefox, Safari) with no extensions enabled, private window
  Just making double-sure with the no extensions and private window, to avoid both a potential extension problem, and a potential cache/cookie problem. Also the non-Chromium and Chromium, as sometimes browser type can make a difference too.

[lament geyser]

Yup, I have tried that. I've also tried it on a completely differnet network as well, and same issue

[timber shadow]

It's definitely odd that it's working in the Android app and not on a computer -- do you have any custom DNS settings enabled, at either the device level or network level, since you said you've troubleshooted both already? Or any other program running that might filter your traffic (antivirus, firewall, etc.)?

[lament geyser]

Yes to both custom DNS and firewall on the home network. However, I've also tried this on just 5G through my carrier without any other filters, DNS settings or firewall and still the same thing

I've also tried the site on my android device and same thing, but the app works lol

[timber shadow]

When you were testing the network swaps, did you either clear cache/cookies, or try in a private window? Depending on what "levels" your DNS filters apply, a reboot of the computer between swaps could be helpful too -- settings could be caching in strange ways in your firewall I imagine?

[lament geyser]

Yea on my android I even went as far as clearing my device cache. I've also tried in normal and private windows too

(cleared device cache as I had a device update this morning too lol)

[timber shadow]

You might try collecting the different devices, network types, browsers, and platforms you've tried so far and posting in #1070006915414900886. Including information about what kind of DNS settings you have applied to your network and device(s) could be helpful too. OpenAI staff are periodically present in #1070006915414900886, but you can submit a support ticket at help.openai.com too -- open the chat widget in the bottom-right, click through the most relevant options, then you'll be able to type your ticket which OpenAI support can review and get back to you on.

[cunning orbit]

searching for knowledge takes hours, every single req fails. Any advise on how to fix that?

[lament geyser]

Awesome, thank you Solbus!

[pearl obsidian]

I guess it's broken again?

[midnight barn]

I am not having that issue on my end.

[crimson bobcat]

what does mean that button in the GPT store?

it appears to do nothing in the GPT store home page

[trail cairn]

i think its for teams/enterprise to switch between global or organization gpt's

[cold folio]

Does that happens when you try to login?

[crimson bobcat]

idk why it is appearing to a plus user

[plain pagoda]

Hello everybody, I have a question or maybe some questions related with getting a subscription, can I ask here or is there a particular forum where I can do so?

[timber shadow]

Hello, this is a good place for questions! Fire away

[plain pagoda]

Thanks a lot, it's kind of silly but I'm confused. I need to work mostly creating images, so I assume I'll have access to the latest DallE version once I subscribe, not sure what other benefits but I'm sure many. Microsoft created a tool that uses DallE for Image Creation since December or so, I do have a different email address for Microsoft so, I'm not sure that I'll be able to migrate that content to my Google subscription.

And another one is to know if I can only subscribe with my card information, I can't see a Paypal option, how safe can I consider it?

[timber shadow]

ChatGPT Plus (20 USD/month + applicable tax) will give you access to DALL·E 3 on ChatGPT, with a 200 images/24hrs limit. You also get the other benefits of Plus (general GPT-4 access, internet browsing, code interpreter, custom GPTs). There are also higher limits available with a ChatGPT Team subscription, which is one tier above the Plus subscription.

Microsoft indeed has its own implementation of DALL·E 3 as well. They just launched a Copilot Pro subscription with DALL·E 3 access included, and their general AI image creator on Bing/Designer uses DALL·E 3. These are separate from ChatGPT subscriptions, and no images generated on the different platforms will migrate to the other (separate accounts, separate services/subscriptions).

As for credit card safety: I haven't had any issues with my card on file. PayPal is not an option -- as I understand it, it's due to verification limitations. ChatGPT's payments are powered by Stripe.

[thorn zealot]

Hi all. Can anyone offer some advice on my issue. Any help is appreciated. #community-help message

[timber shadow]

Hello, do you see the other features of the GPT Store (categories, rankings, etc.) with only the search bar missing, or do you still have just the old Explore page showing only your GPTs and the default OpenAI ones?