community-help

[vapid warren]

Guys is Platform down?
When I try to signup for platform it redirects me to chat.openai.com

[charred parcel]

create an account and then use that account to sign into platform.openai.com

[vapid warren]

But I already had one

Of my company

API Keys are still working

[charred parcel]

then log in with that account

[vapid warren]

When I log in it kicks back

idk why

[brisk solstice]

now 40 for 3h

[tropic spade]

Custom gpt is way lower than 40 it's about 10 I think

[frail nymph]

I was on today just for 30 min and it kept giving me an error "There was an error generating a response & Regenerate". I tried simple questions and it worked. But complicated questions kept fail. Sometimes it would regenerate a response that I liked, but then disappear and give me the error again. I somehow reached my cap without getting anything today.

[brisk solstice]

It seems to be caused by overloading? Because there seems to be a bug https://discord.com/channels/974519864045756446/1174578478474088458 that causes free users to be able to use gpt4 as well

[neat tinsel]

Does anyone know if ChatGPT is working properly now?

My page says: "ChatGPT can make mistakes. Consider checking important information."

[brisk solstice]

I'm using it

[neat tinsel]

I can't, I can't even access the ChatGPT page

[frail nymph]

Does anyone know if there is a way to stay in the same chat but switch to 3.5 but later come back to 4.0? When I tried a couple months ago it locked that chat in 3.5. 90% of the time, 3.5 would be fine, but I want the chat to be the same

[uncut folio]

Hi! Apologies if this is the wrong spot. When I clidk on "My Plan" nothing happens. I don't have pop-ups blocked. What am I missing?

[frail nymph]

Seems to be a problem today. I could access it yesterday but now nothing happens

[uncut folio]

Thank you!

[brisk solstice]

me too

[carmine pumice]

I got problem when use chatgpt. The page is a whiteboard page, and no materials are loaded successfully, and the response status of all js and css file URLs (such as https://cdn.oaistatic.com/_next/static/chunks/4865-1af3a0b4fab98171.js)
that need to be loaded is 403. It seems my ip was banned? How to solve this problem, can someone help me?

[bright lily]

I'm surprised you went through several browsers 😅

[nocturne hatch]

I am a veteran in IT and lately, browsers have been so lousy that everything has gone through my mind. But otherwise, IT 101 teaches that first, look for the problem in the most obvious place, namely in the browser. Do you have any problem with that?

[bright lily]

It doesn't seem to be working for me either.

That's a pretty important page to not have access to 😐

They're probably like "Wow look! No one has unsubscribed in days!" 😭

[nocturne hatch]

I understand your sarcasm but I'm unable to appreciate it. Irrelevant. It's far from me to assume such things about the OpenAI. Enjoy your day..

[bright lily]

It's more of a joke c: I'm sure they're working on it. G'day good sir.

[lyric fox]

"My Plan" also not loading.

The admins don't seem to understand, either. The ‘MY PLAN’ page does NOT load AT ALL, and there are SEVERAL PEOPLE having this issue. We can’t even get to the Stripe page to update our billing info. I am demanding you refund me ALL Chat GPT purchases and give me an extra month FREE due to this GLITCH on YOUR end, or I will involve my legal counsel.

[charred parcel]

that's not going to happen

[rugged wing]

hey ayone knows why does the chatgpt site use INSANE ammount of cpu ?
i have i3 1115g4 so a website shouldnt take so much cpu power is this mining bitcoin or something to use so much processing power ?

[finite parrot]

Hi there, guys is anybody else experiencing an issue when opening an old chat? There's a message showing "Oops, an error occurred!" and a Try again button below, or just me?, I checked in 4 devices connected each one in totally different networks

[coarse wigeon]

I do, it might be plugin issues, reported in the bug-reports

[timber shadow]

Do you happen to have any browser extensions installed that might affect ChatGPT?

[rugged wing]

no

i have even disabled the annoying ai prompts in opera gx through the opera://flags/

[vague condor]

I have trouble accessing the "My Plan" section from the Menu. I need to retrieve my billing history and download receipts. Anyone else experiencing this?
EDIT: I see, others have the same problem.

[humble thorn]

I am experiencing exactly the same issue

[steel elm]

guys, i need some help.
how can i outpaint an image in the new api?

an image that i upload

[sage light]

hello guys, when i try to get into a chat I sometime get an error message and it says to try again, though the button is not working and i have to visit the page in another tab

[mild bloom]

same for me its not working

[sage light]

well

working on the phone app though

guess we'll have to wait

[finite parrot]

That's a relief for me... Thanks for sharing the info... I'll check

[compact gale]

Anyone understand how in playground we have access to a model with a large context window, but the max length is still the same???

[crude rover]

I have an active chatGPT+ subscription, but when I updated the app just now, it has me logged in as the free plan, with no access to my plus plan. Can you help?

[river dock]

How come I’m locked out of my plus subscription and how do I get it back? It tells me I have a plus subscription but it also tells me I have to upgrade to plus. How is that possible?

[crude rover]

I have the same thing going on right now.

[timber shadow]

@crude rover Is it the same behavior on both web and app?

[crude rover]

I'm in iOS, will check web now..... brb

[river dock]

Yes, same.

I can’t use 4. It only gives me 3.5. When I try to use 4 it tells me I have to upgrade but will not let me upgrade.

[crude rover]

Looks like I'm still okay on website. I have had the tab open forever and it's still working in ChatGPT 4. I'm in Chrome on Mac OS.

[river dock]

I can’t access plus on web or iOS app.

[crude rover]

What's your platform and browser on web?

[river dock]

Even though it shows that I have plus.

I use edge.

On windows 10.

But regardless, it’s the same on iOS. I saw there was an upgrade so I tried it but no change.

[toxic bear]

Is this a good place to report issues with the app?

[river dock]

I didn’t have this problem last night. just noticed this morning.

[river dock]

Where else? There is nowhere else.

[timber shadow]

@crude rover There have been reports in the past of new Plus subscribers not getting full access on all platforms right away, like it gets "stuck" or something. You'll find similar reports in #1070006915414900886, but it's been probably a week or two since I remember seeing such reports. It's not unheard of though. For those folks, it seemed to self-resolve within a matter of a day or so. I hope it resolves ASAP for you both! You might consider adding your experience to one of the bug reports posts, or starting a new once since we're in this "new UI phase" of ChatGPT

[timber shadow]

Here and we can maybe try to help troubleshoot. If it's just a bug, try #1070006915414900886

[river dock]

I’ve been a subscriber for several months.

[crude rover]

Thanks Solbus. I've had full access for a couple of weeks. Does that still fit your scenario? I reported the bug in the chat bot.

[timber shadow]

Oh I'm sorry I misunderstood you both 🙃 no, it does not fit that scenario, I'm really sorry!

[toxic bear]

This has persisted for a couple weeks now. Updated iOS, the app. Uninstalled/reinstalled. I can’t figure out any solution.

[river dock]

I also reported in the chat bot and it told me to come here for more help.

[crude rover]

Same here 🙂

[timber shadow]

You two should team up on a new #1070006915414900886 post! OpenAI staff do appear there pretty frequently

[crude rover]

I was using ChatGPT+ today for awhile and then I clicked the "update" button on the app to get the most current version of the app. That is when my problem started.

[crude rover]

Thanks!

[timber shadow]

So, simple question but: are your device's date/time settings correct/set automatically?

[river dock]

Yup

[toxic bear]

Yes, thanks for asking. They are correct. Although, I’m in a very remote place, the IP addresses may register to another location. But it worked here for a week before this issue began.

[timber shadow]

Do you have a VPN installed on this device?

[toxic bear]

Yes, I installed the vpn in hopes that would resolve the issue. Thought there was a chance that it had been blocked by the companies who own the infrastructure here.

I’ve tried with and without the VPN

[timber shadow]

Hmm, so it was happening before you started using the VPN? No behavior change at all with or without? What is the VPN?

[toxic bear]

That’s correct. ProtonVPN.

[timber shadow]

Is your location (actual location, not VPN server location) a supported country for OpenAI services?

[toxic bear]

Yes, I’m in Alaska

[timber shadow]

Okay good deal. Can you try both disabling the VPN, and also disabling ProtonVPN's specific "NetShield" feature? This will be separate from just disabling the VPN itself. I know you said it was happening completely without the VPN, I'm just trying to do some of the basic troubleshooting things first

[toxic bear]

How do I disable the NetShield feature? That’s on iOS?

[river dock]

I reported it in the bug channel. https://discord.com/channels/974519864045756446/1174773086709633095 let me know if you have one also.

[timber shadow]

It appears to be a feature per the ProtonVPN website. Try rooting around in Settings? It might go by another name too, Threat Protection, Web Protection?

[toxic bear]

NetShield is not enabled

[timber shadow]

And neither is the VPN now? Can you try logging in directly on chat.openai.com?

[toxic bear]

I can log on through the website

I cannot through the app

That has been the case all along

[crude rover]

I have one also. https://discord.com/channels/974519864045756446/1174773659437633696

[lyric fox]

Oh yes it will.

[timber shadow]

I see, sorry for misunderstanding that. I'm wondering about the "device check error" then. Do you happen to have a jailbroken iOS device? I'm on Android so I'm speaking a little out of my experience here, but I did have a little issue early on my Android phone because it was rooted

[toxic bear]

No it is not jail broken

And I wasn’t clear initially, not your misunderstanding. Thanks for the help.

[timber shadow]

Is it the same behavior on both a WiFi and cellular network, if you can try both with neither the VPN nor NetShield enabled?

[toxic bear]

NetShield is not an option without the paid subscription. I have tried both cellular and WiFi and both with and without VPN.

[timber shadow]

Shoot, I'm totally out of ideas. I think this is worth a #1070006915414900886 post. Share all the details you've shared here, and hopefully an OpenAI employee or a more informed community member can take another look for you. OpenAI staff are present there periodically

It would be odd but perhaps not impossible that the IT folks have somehow blocked the iOS app telemetry stuff, but not the web? Total guess on my part

[toxic bear]

Thank you

[crude rover]

Thanks DSMagic. Turns out my problem is only in the app. I can still access via browser. So I didn't comment on your bug report, lest it confuse them. Seems like maybe we have two different issues.

[river dock]

It wouldn’t surprise me if you closed your browser it wouldn’t let you back in.

[crude rover]

Yikes I will keep it open then. I did do a cache-clearing reload of the page and it stayed logged in. I hope they see your post and can fix whatever is wrong. You shouldn't have to pay for inaccessible service.

[green crow]

does anyone know how i can make gpt "dumbify" these outputs to just text?

it started happening in the new updatre

[silent plover]

guyss, whatsap?! i'm having trouble with acess to chatgpt on my pc. I am using the same login e browser at androind and there works well.

In pc I open the site, write a messenger and appears this message: "Something went wrong. If this issue persists please contact us through our help center at help.openai.com"

What could be happening?

[timber shadow]

That looks like LaTeX maybe? But I don't know what that is for sure, I've just seen other people talk about it re: math formatting. So maybe ask for plaintext and not LaTeX formatting?

[timber shadow]

Do you have a VPN installed on your PC?

[green crow]

Nice, thsanks. what does Latex stand for?

[stray owl]

That is LaTeX, yes. @green crow
What you should do in your original prompt is say: "In your response, do not output LaTeX. Please only output your response in plaintext."

[green crow]

Yeah, thanks, it worked

[stray owl]

I'd like some feedback on my GPT if anyone is down lol

[silent plover]

Yes, I have. But sometimes worked, even with VPN on. Now, for example, the VPN is disconect but the wrong persists

[timber shadow]

What VPN do you use?

[silent plover]

OpenVpn GUI

[timber shadow]

I see. Does that program have a separate feature by chance that has to do with any DNS filtering? It's sometimes called Web Protection, Net Shield, Threat Protection, or something similar. Sometimes this feature needs to be disabled separately from the VPN itself being disabled. Can you check for that?

[green crow]

I use OpenVPN through Mullvad, and it works fine for me

try a different server/tunnel

[silent plover]

New browser?

[green crow]

Just chrome on the newest version

i use anyway. maybe try firefox/safari

[silent plover]

I can't. I am not the guy for that. But I have a co-worker tha can help me in this. If it has some filter you think can I disable it just for Chatgpt or not? Because this VPN is for work

[green crow]

Not all of my OpenVPN servers work, just try another location

It's not uncommon for CGPT to not work with VPNs

[spice wave]

Is anyone experiencing issues with ChatGPT? This is what I get when I ask a question

[green crow]

Kill your cookies and sign in again?

[silent plover]

I tried in Microsoft Edge. Same

[green crow]

sounds like a cache/session problem

[spice wave]

will try that

[green crow]

it happens to me more common if im signed in with a third party

[spice wave]

I cant event see the chat now 😢

[green crow]

are you using ublock origin by chance?

cuz if so, try refreshing the filters, that worked for me last time

[timber shadow]

What browsers have you tried so far, other than Edge?

[spice wave]

I am so I'll try that

[green crow]

if it works incognito, then its mostly likely it's because of a session issue / extension

[peak thorn]

Having issues with my gpt, its code just stops writing, could it be that I gave it too much data?

[cosmic crag]

Mine stops writing the code sometimes, too. I ended up just uploading a python script

Problem is that it has stopped obeying instructions, and won't follow the OpenAPI schema in actions.

[bronze pagoda]

just why?

[cosmic crag]

hahaha

[bronze pagoda]

im trying to not be racist

[earnest rain]

What’s the best way to get a hold of the actual company? I’m trading an afterschool program with ChatGPT as the cornerstone. It’s a pretty niche question.

[cosmic crag]

regardless, thats funny

Three ways I know of: tell chatgpt to carry a message, message modmail here and it may eventually get to someone in the ocmpany, or talk to sales.

[earnest rain]

Thanks

Any way to talk to a person live?

[cosmic crag]

No dm's on X, no messages on linkedin

[earnest rain]

Sales I guess?

[cosmic crag]

Actuallly, you could talk to someone marked as working at openAI on linkedin

Or, if you are brave enough, @ sam altman on X

[trail cairn]

it can be read that you are still asking for slurs, given the context be careful as you can get a warning email or a ban

[finite parrot]

Any luck about the Oops meme we've been facing since the early morning?

[peak thorn]

literally cant get past this

wrote like 20 lines of code

[earnest rain]

Honestly I’ll give it a shot

[cosmic crag]

Honest, I have given up interacting with gpt-4 or editing my gpt for the time being. I'll help people here, but when I stop getting 4 openAI status errors a day in texts, thats when I'll try again

[cosmic crag]

I wish you luck, sir

[cosmic crag]

@peak thorn what is the python code it is using?

or to what point does it get

[peak thorn]

thats it

[cosmic crag]

Ah, I had problems like this before, too

Just to double check, those are for sure the file path names?

[peak thorn]

yep

[finite parrot]

Flask app?

[cedar mountain]

why it's showing like this?

[cosmic crag]

In my experience, when gpt-4 analyzes to 'understand' the structure of a csv or a pdf or something, it fails. either stops or fails

So, you have a few choices. One: write the python script for it. Two: give it another file that is easily read like a .txt that details what it wants

[cedar mountain]

it's showing me since 24 hours

[cosmic crag]

AJ, this has been a common problem, especially this morning

[cedar mountain]

thank you for reply.. do you have any idea maybe when it will be solved?

[cosmic crag]

Honest, no idea. But hopefully soon. It seems like it is a minority of users, but still be fixed at least somewhat quickly

[peak thorn]

Well thats terrible

[cosmic crag]

Yeah. Another thing you could do, is tell it to use a different manner of reading the csv

The gpt-4 python environment is...quirky. Some python libraries act completely usless.

[peak thorn]

Its a massive data set, what do you reccomend?

[cosmic crag]

hmmm. Try Modin. It is built on top of pandas, but better for larger datasets, in my experience

[peak thorn]

Great, thanks so much will let you know how that goes

[cosmic crag]

please do

I'll say that, last resort, convert the csv's to json. gpt does so much better with json than csv

[tardy pumice]

does anyone have custom GPT for understanding Actions? thanks

[cosmic crag]

I am not custom gpt, but I can help with actions.

At least I don't think I am a gpt

[tribal talon]

Thats what a GPT would say

[cosmic crag]

*cough

[broken viper]

hi guys

[cosmic crag]

howdy doody

[broken viper]

Can I ask are you a OpenAI moderator?

[tribal talon]

I try not to moderate but I guess technically

[broken viper]

Oh okay.

[cosmic crag]

Hey, can you be my in at openAI? Looking for a job.

[tribal talon]

Probably not but you can mention you're discord active 😄

[cosmic crag]

Darn. Hehehe

[broken viper]

Are you looking for a Job at OpenAI?

[cosmic crag]

I just saw an AI eithics engineer listed at openAI a little bit ago. Kind of a dream job of mine

[peak thorn]

I told it to use Modin instead of pandas and then it proceeded to use pandas again

[broken viper]

Is it a Custom GPT?

[peak thorn]

Yes

[cosmic crag]

Nice. Hmmm

Are you knowledgable in coding at all, nack?

[peak thorn]

Somewhat

[cosmic crag]

You wanna try your hand at making a small python script?

Or I guess I can

If you trust me enough that is

[peak thorn]

I trust you man if you really want to! Very kind of you to offer

[cosmic crag]

Let's take it to DM's then

[manic cobalt]

My GPT - load more is not working. Can't adjust some of my GPT's

[green crow]

Is it just me or is there no option on iOS to clear all chats?

[trail mulch]

Has anyone else been getting this incredibly helpful error message when trying to update training data for a custom GPT? It happens whenever I try to upload a new file for training, regardless of what file it is. I've tried adding several different files through "Configure" and also doing it interactively through the "Create" tab. Any suggestions?

[cosmic crag]

Yeah, another common error lately. I had this yesterday, the whole day, but it seems to be workign today

@green crow I don't think there is a way on android either

[azure hare]

yall experiencing this?

[trail mulch]

Did some more testing. Created a brand new GPT. Gave it the first 10 files to add to the training data, it worked great. Adding any additional files at all, be it 1 or 10, causes it to error and not save. :/

Created another one and only gave it 1 file to start with, it loaded that file fine. When I go to add a second, it errors out. It's just not letting me give any new training data after initial upload, period.

[green crow]

sounds like a bug the devs need to fix

[cosmic crag]

I think what you are looking for is "variations", and it is only DALLE 2, not 3

[chilly ice]

Any way we can delete chatrooms without really clicking on them? If I click them they receive an "Oops, an error occurred" prompt. These are the ones with plugins. Not sure which plugins are causing the issue.

Could be all of them for all we know since there's a problem.

[balmy drum]

Hey I got a quick question, so longstory short I have been using a chat for a while and now it is getting really laggy whenever I try to open that chat.

I heard somewhere that this is an issue with the DOM generating all of the past text history at once, making longer chats result in having a lot of lag.

I want to find a way to either A. Reduce this lag without losing my chat history or B. Find a prompt that would let me tell the current chat I am using to generate instructions for the other chat to follow.

I have been trying to figure out a way to go about doing this, but I keep falling short of what I want to achieve. Any ideas?

[junior maple]

If your thread goes on for too long then it starts truncating it and only sending the last x messages to the AI. So there's no real point in dragging them out for so long and doing that

[cosmic crag]

Beat me to it. How dare you

[junior maple]

f

[bronze charm]

ChatGPT's GPT-4 currently uses 32k token context window which would take some time getting to but this is nevertheless true
8k for GPT-3.5

[cosmic crag]

To add, all that data that is chat history is sent with every request to the server. I think its more server issues than the bot itself

[junior maple]

I've never needed to bring a thread longer than 10 messages. Once it gets that long the AI usually starts to make obvious logic errors and not do so great

[cosmic crag]

I've had programming chats go on for much longer than 10 messages, but it is true that quality goes down the longer it goes

[balmy drum]

Oh I have a really long thread for this joke/experiment I am doing, pretty much in a discord server I have been copying and pasting what people say, and my instance has been generating responses in this really weird way that's just super funny. I can give an example if you want LOL

The lack of memory doesn't seem to be a problem for me as much, since there are central themes with the messages generated that tend to persist.

The lack of quality adds to it

[cosmic crag]

Thats great. I can't believe the real mr beast is with us in our humble server

[junior maple]

The chats are pretty much supposed to be ephemeral. Anything more than that and that's pretty much untested and largely unsupported

[bronze charm]

What if... you made a GPT that has some messages as examples in the instructions and it can base its responses off that

And you can just make a new one when you need it

[junior maple]

Unrelated but kinda related but I'd like to see GPT creators being able to finetune off of selected chats as an alternative to extremely long prompts / many-shots training

[balmy drum]

I actually tried that, thing is there is like weird lore? Like at some point it started calling itself the luminary of banter, and I am worried those details may not be captured.

[cosmic crag]

My new band name

[junior maple]

Some GPTs I'll just come up with some lore in my first message and they'll understand the assignment and start making up stuff too. I love doing that trick with my Kato by referring to it as a "powerful sorcerer"

[bronze charm]

To build off of that, you can update the instructions whenever the GPT does come up with new lore

[junior maple]

If we could see user chats and finetune off of them that would be incredibly OP tho. I remember farming outputs off of GPT 4 and finetuning 3.5 with them and actually getting better results than what I put in

Hooks for text adventures in case you're curious

[cosmic crag]

I actually have a question. Why would my GPT not be available on mobile, yet?

[hexed fern]

Hello

Can someone tell me how do you know how many people are using your GPT bot?

[timber shadow]

Hello. If you go to chat.openai.com/gpts/discovery, you can see how many chats have been started with your GPT. It includes your own chats, so you'll have to subtract that

[normal bloom]

Newbie question: in Windows, what keyboard combination or keyboard shortcut clicks "Add and run" button in https://platform.openai.com/playground?

[cosmic crag]

So, you want the same base, but you want bigger differences than what the variations API gives

[modern light]

why, no matter how hard i try, can dall e 3 only generate girls with one or two afro puffs

if i say 3, 4, 5, etc. or even describe in details how the hair looks

it always ends up with one or two puffs

[timber shadow]

DALLE, much like GPT, can't really count. You might try using different kinds of qualitative language like "a few" or "many" to see if something like that gets you a more consistent output around the number you're looking for.

[modern light]

mk

[modern light]

still only two, i've said "a lot of" and "a very large amount of"

[lament mango]

My chats with "Web pilot" plug in or any plug in which allows link reading is not working.
The previous chats are displaying an error messahe.

This is the error. Can anyone help?

[timber shadow]

A professional headshot of a woman featuring a unique hairstyle with many small Afro puffs. Each puff is distinctly rising separately from her scalp, showcasing perfect alignment. The hairstyle is meticulously crafted, highlighting the individuality of each puff. The woman has a confident and professional appearance, dressed in sophisticated business attire. Her facial expression is composed and inviting, with a gentle smile. The background is subtle and professional, focusing attention on her hairstyle and facial features. The lighting is soft yet highlighting, emphasizing the texture and form of each Afro puff.

Maybe something to start from? Had to try a few times, so I see what you're saying!

[tame wigeon]

i have acess to GPTs, but I cant see GPT's that other people make. When do I get access to that?

[modern light]

i'm using bing image creator for it
mirror selfie of curvy darkskin black girl with a large amount of afro buns, vitiligo, bushy fake lashes, freckles, gold circle frame glasses, gold square small hoop earrings, gold square nose piercing, hourglass baggy very pastel yellow shirt with a NYC Basketball graphic, high rise blue ripped jean very short short shorts, white air force 1s, sitting in room on tire swing, white walls, white ceiling, very fluffy white carpet, a lot of afro buns

[timber shadow]

Store launches later this month! For now we have #1171489862164168774

[lament mango]

My chats with "Web pilot" plug in or any

[modern light]

it's also not generating the square piercings but that's not as bad of a problem

[manic cobalt]

How come Feastables are not available in The Netherlands?

[tame wigeon]

thanks! so only limited people currently have access right?

[timber shadow]

I don't think Store is live for anyone yet. They just said on DevDay "Later this month"

[hard meadow]

My chatgpt plus plugin chat shows errors both the old and the new chat s

wht should I do

[timber shadow]

I've read a few users say enabling more than one causes this bug right now. Does it happen with just one plugin enabled?

[hard meadow]

no I have more than one

[timber shadow]

Can you try it out with just one to see if it works?

[hard meadow]

but it worked before I need to get my old chat info

it works with only one

[timber shadow]

Seems to be a bug, you might add your experience to similar post(s) in #1070006915414900886 -- hopefully fixed for you all soon!

[tame wigeon]

Ohh but you can share via URL currently right?

[timber shadow]

Correct! You can publish GPTs just for yourself, for anyone with the link, or publically, but publically is pretty much the same as anyone with link right now as there's not the central hub of the Store yet. You have to enable either your name or a website if you want to publish Public right now.

[tame wigeon]

btw @timber shadow you’re killing the discord support man 🔥I’m out of questions

[timber shadow]

😄 appreciate that! ask away any time, tons of super helpful folks in this server

[rugged wing]

I can't access Bing's AI Generator whenever I sign-in, can someone help here, please?

Nevermind, I got in

[cosmic crag]

I leave for a couple hours and solbus takes my place

Goshdarnit

[timber shadow]

There is no one wiser in Athens than you, may I only strive to be as wise 😎

[cosmic crag]

And yet, I know that I know nothing

But what really is knowing?

[astral quest]

I kinda doubt that with that amount of traffic right now. But I also hope so!

[timber shadow]

That's that secret sauce!!

[timber shadow]

Good point, I wouldn't mind a little more dust settling first too 😁

[cosmic crag]

The GPT creation should be less painful by the time the store opens, I would hope

[stone hazel]

wonder if someone here can help.... I created 12 custom GPTs, but a few of them have randomly disappeared. Is there a limit to how many you can create?

[astral quest]

#1174906536473149470

[timber shadow]

There are reports of having trouble loading some GPTs behind a certain number, see #1174906536473149470 message
For now it seems safest to manually save your GPT URLs. Hopefully you can find them in your history?

[astral quest]

you can open them via the link, but not the editor

[stone hazel]

Yeah that does seem to be the case as I could load them in the list but had an old one saved in another tab

[stone hazel]

Are we expecting this to be resolved any time soon?

[timber shadow]

Maybe throw a ⭐ at that bug post for visibility! OpenAI staff are present there periodically

[astral quest]

I hope so. You can upvote the bug report

[stone hazel]

I've thrown a star on it. Just the lonely 1x at the moment hopefully others do the same 😄

Appreciate the help guys thank you

[astral quest]

Nope you didn't :(

[stone hazel]

oh i see not just the reaction but the tag... done now!

[astral quest]

Nope, still no reaction on that post lol

[stone hazel]

for real?

[astral quest]

yea

[astral quest]

oh you made a new one

[stone hazel]

a new forum post? nope...still the original

are we looking at the same one? 😄

oh you made one i see it now

[astral quest]

I guess not. I am talking about the one solbus and I linked

[stone hazel]

yeah i see

[astral quest]

didn't know you had another bug report up

[stone hazel]

all good

the more the merrier. hopefully it gets noticed

would rather not have a bunch of tabs up for different GPTs...lord knows I already have enough tabs xD

[astral quest]

yea they get out of controll quickly 😂

[stray owl]

Hi! Is there anyone here willing to give user feedback for a GPT I made

[cosmic crag]

Send me that link. I won't go easy on ya. Warning.

[stray owl]

I appreciate you!

#1171911165861970001 message

[night breach]

It's kind of like a copy of mine #1172301858300235846

[stray owl]

What lol mine came out way before you

[night breach]

How do you figure that?

[peak thorn]

Anyone have expierience setting up azure open ai?

[wise lodge]

Hi all! What the heck is openai schema? I can't find documentation for it anywhere 😦

Nevermind, found it

[woven osprey]

I can't log into openai, can anyone help me?

[floral summit]

• Click on your name on the bottom left
• Click on "Settings & Beta" -> popup with tabs on its left side appears
• Click on "General" tab -> See "Clear all chats" with a button "Clear" next to it.

I haven't tried it, and not found a way to clear single chats instead except clicking on them first. I hope it still helps. Sorry if this has been answered and I overlooked it.

[wise lodge]

How do I fix this?

None of the provided servers is under the root origin https...
In components section, schemas subsection is not an object
Server URL http... is not under the root origin https...; ignoring it

My server URL is an IP address and the port is 8080. The same request works fine in postman and curl

[pine cargo]

I'm lost... How do I use the openAI api key?

This doesn't work, is it outdated?

openai.OpenAIError: The api_key client option must be set either by passing api_key to the client or by setting the OPENAI_API_KEY environment variable

But I do pass "api_key" to the client? (x is to hide)

[trail cairn]

https://platform.openai.com/docs/quickstart/step-2-setup-your-api-key

[ruby bough]

I have a question regarding the Privacy Policy.

If a user who is not an Enterprise subscriber creates GPTs that connect to the API, what is the privacy policy for the data obtained from the API?

[trail cairn]

you need to write one and list everything your api has access via the api to and what it does with it

[hazy sorrel]

which framework of python is best for gpt development:
django,flask or fastapi ?

[ruby bough]

The API accesses user activity stored on its own data servers, allowing GPTs to retrieve and analyze information from the API.
This is something we don't want used for training.

[spark mango]

Hey guys i can't run even the simplest of python code using the openai library

Anyone come across this before? ubuntu 22 lts with python 3.10

[trail cairn]

anything transferred back to user and their gpt session is a subject for training based on users settings. do not transfer PII

[ruby bough]

Thanks, I'll look into using Enterprise.

[trail cairn]

fyi enterprise requires +100-200 users and one year commitment, api offers assistants which are nearly the same as gpt's

[warped breach]

If you use gpt 4 turbo from a plus membership, compared to registering an enterprise account to a personal llc. Would you get a "smarter" or more powerful model? Or is it just higher token limits and some admin functions? Could you analyse a longer pdf for example, or write a longer book?

[astral quest]

Enterprise is only available for companies with more than 100 employees

But no, its the same model, with better availability and a higher usage cap

[warped breach]

So the same as buying tokens etc

[ruby bough]

I see, the Assistants API allows for a nearly implementation.
That was helpful.

[viscid abyss]

hey guys, where can I find the API implementation for DALLE 3? And I was trying DALLE 3 on playground and it is throwing me this error

[faint token]

Hi guys, i keep getting to verify myself as human on litteraly every prompt i send, anyone got a fix for this?

[rigid musk]

Vpn often causing this

[rough spade]

Why I can't reply on OpenAI's Forum?

[rugged wing]

Since yesterday new chats created an error. Now I also can not open existing chats.

[timber shadow]

Are they Plugins chats by chance?

[rugged wing]

I switched of plugins, used other browser, restarted my Mac. Nothing helps

In previous chats I used plugins

[timber shadow]

Sorry, I'm talking specifically about the GPT-4 Plugins model, where you can enable up to three plugins? Many users have been reporting errors when trying to use a chat or start a chat with more than one plugin enabled. Does it work if you try to start a new chat with just one plugin enabled?

Or are we not talking about the same plugins here? 😁 not talking about browser extensions, to clarify!

[rugged wing]

Today, withoutout plugins or 1,2 or 3 plugins activated in chat gpt that chat works. The chats of yesterday and most existing chats load for a few seconds and give an error

[timber shadow]

You might want to post in #1070006915414900886 or add your experience to a similar post there! May be related to the other stuff occurring lately

[rugged wing]

hanks, will do.

[zenith solstice]

It start showing "The conversation is too long, please start a new one" now

[neat gazelle]

hi guys, do you know if openAI offers any kind of plan for education purposes ? my company is some sort of AI club, and we want to purchase licenses that our members can access, but with some sort of control of who accesses it and when.

[spark cliff]

how do i fix the error when i upload pdfs into gpt 4? a few days ago it worked pretty good, now all i see is error or retry

[silent granite]

Hello. I'm reading the docs but there are some things I don't yet understand.

• Custom GPTs, is this basically the same as Assistants API?
• I currently have a LangChain setup using RAG. Is this obsolete now with retrieval?
• When I create an Assistant using the API, does it exist forever or do I have to recreate it? If it exists forever, can I update the retrieval files without creating a new "assistant"?

I would love to read some more guides about the Assistants API, can someone recommend me anything? Preferably using Typescript. Thanks!

[timber shadow]

I can only answer one small bit of your question: custom GPTs are not the same as Assistants API. They're basically like more complex custom instructions, with the additional ability to upload knowledge files for reference, and include some API actions as well.

You might try asking in #dev-chat too if no one gets back to you here!

[silent granite]

I can see instructions, code interpretation, upload files and retrieval in the assistants playground though (https://platform.openai.com/playground?assistant=new&mode=assistant). That's why I assumed they are the same, but it's something else?

[timber shadow]

They may have some functions in common, but like with everything OpenAI: ChatGPT is a curated, and in many ways simplified experience of their tools, and API tools can be used more flexibly in your own programs, if you know what you're doing! Which it sounds like you do! Unfortunately, I'm about at the end of what I know, no API usage myself.

[silent granite]

I understand, thank you!

[timber shadow]

For sure! Another shout-out for #dev-chat too, lots of knowledgeable people in there

[stark jackal]

I have been developing with openai in python for a couple of weeks but all of a sudden i get errors when trying to install the openai module in all of my coding spaces

[timber shadow]

Maybe try in #dev-chat if you don't hear back here!

[zenith solstice]

Anyone has the problem that I cannot load into previous pages

[tranquil jackal]

Hey all, doesn anyone know how i check the allowance for my GPT4 Plus account? Is this just random or do i need to keep track of how much i'm burning through it?

It just shows up after 10-20 minutes of usage it's quite dissappointing to always be getting this as a paying customer.

[rigid musk]

It is 40 prompt per 3h

So you have to wait till 1 prompt is available

[tranquil jackal]

pretty sure i haven't done 40 prompts.

[rigid musk]

So about 1 prompt every 4.5 min

Regen, error count as prompt

[tranquil jackal]

How do i increase it so i can use it all day?

Is there a next level tier i can pay for?

[rigid musk]

Unfortunately you cant

Other solution will be using api instead of chatgpt web

[tranquil jackal]

I'm in the API section now trying to figure out where i can see my allowance but it's not showing any allowance stats for GPT4 Plus account. It just shows my free allowance from when i first signed up back in the day.

[rigid musk]

Different payment , different products

[tranquil jackal]

Oh dear... really thought $24pm would get me more than 20 minutes worth of work

[rigid musk]

Do 1 prompt every 4.5 min and you never reach the limit

They will increase it in future

You can always create a new chat and use 3.5 as temp solution , but don't convert existing gpt 4 to 3.5 you can't revert

[slender sinew]

it's kinda tough tho, esp when you're testing out your GPT

[rigid musk]

It is a temp solution they even had to pause new subs

[slender sinew]

by default, it's using GPT 4 - and we can't really swap to GPT3/3.5 haha

[rigid musk]

If you testing gpts I understand. If this is just to have fun randomly you can create new chat with 3.5 as alternative

[marsh gate]

Why if I have version "chatgpt 4" I get this error when using GPTs: You have reached the current usage limit for GPT-4.

[timber shadow]

There is a usage cap for GPT-4 on Plus. Custom GPTs use GPT-4, including testing them and using the "Create" part of builder.

[marsh gate]

Ok, thanks @timber shadow

[night breach]

We should write a petition to increase the usage cap, so we can give OpenAI people a higher incentive to upgrade their servers faster.

[rigid musk]

You can always make #1070006151938314300 yo have higher cap

[marsh gate]

I agree, it would be a great help

[night breach]

I honestly don't think suggestions would help for that case.

[vital radish]

shouldnt really be a suggestion - they surely know that the current cap is ridiculous

[rigid musk]

They had to pause sub. Server are overwhelmed

[timber shadow]

Yeah at least we're not back to 25 😁

[marsh gate]

how the limit works. Does it go by monthly, daily, weekly limit?

[night breach]

Every 3 hours

[rigid musk]

Indeed that's reassuring

[vital radish]

yes i know - but everyone's $20 that previously paid for 50 is still $20 - its up to them to make sure the service they provide is adequate no ?

[timber shadow]

Rolling. Think of each message as having its own 3hr timer, then you get it back. Counting edits/resubmits/regenerates

[rigid musk]

They also state products is in beta and they can adjust by demand

[marsh gate]

ok

[tranquil jackal]

Thanks for the info Neko. Appreciate it.

[vital radish]

yeah - look i know its a very complex operation, but still - people are frustrated that they cannot use the product to its full potential.

[rigid musk]

Sorry I couldn't help more :/ have a good uwu day

[tranquil jackal]

I submitted a form the other day asking for increased usage allowance but i'm sure everyone else has probably done the same,

[hard basin]

Good morning. I am new to this channel and may not be posting this request in the proper place.

Yesterday (November 16th) I started getting an error message "Oops, an error occurred".
After doing some research, I found a support article saying I should clear my browser's history, which I did.
After that, it worked throughout the day.

Today, November 17th, the same error is happening.
I do not want to keep clearing cache and cookies all the time.
Can you please look into this issue and let me know how to fix this for good?

Thank you and best wishes.

[rigid musk]

Can you try firefox as test

[night breach]

I agree. They should get their stuff together and get more investors and get this thing upgraded. The cap limit is absolutely unacceptable.

[timber shadow]

Is this happening with all chats, or just Plugins chats by chance? Maybe just Plugins chats with more than one plugin enabled?

[tranquil jackal]

Sign me up man

[night breach]

I'll have chat GPT make a petition form and get back to you guys shortly

[plain pagoda]

I have questions... I signed up for the upgrade waiting list since I want to be upgraded in order to use DALL-E, but apparently I have two different profiles, one is the old free version (which is the profile I want to upgrade) labs.openai.com, and recently Edge created an extra profile with my Microsoft email address, the images there in the Microsoft are really close to what I want to create once I educate myself in the platform, the designs in labs.openai are yikes. They look like child drawings vs professional illustrator, any advice?

[tranquil jackal]

This echo's what @night breach has suggested. We're the early birds, and we deserve at least enough allowance to test our gpt's. I completely understand if you're going crazy with a gpt that does a lot of image processing and what not, but when it comes to mere chat, optimizing and chatting with the gpt through text... we really shouldn't be getting kicked out after such a short period.

[timber shadow]

A bit of humble feedback: I'm sure OpenAI wants to offer higher limits to Plus users, so I don't think they're "holding out on us" per se

[vital radish]

At the very least - they should place a notification like status.openai.com on their website informing their customers as to what they are doing to improve the current situation, and to acknowledge that there is an issue - the silence is deafening......

[rigid musk]

Unfortunately people that subs agreed on early products, bug and error can be there. You can always discuss with [e-mail]

[night breach]

It's not about holding out. It's their business model. And as I stated in the past, it's not a very good one. We want upgraded servers yesterday and we deserve to have it because we're paying customers.

[tranquil jackal]

And we also need to know when we're nearing the end of our usage cap. At least if we're notified of that then we can prioritize the next few prompts with extra care to finish off what work we can without the sudden interruption.

[night breach]

Yes, that's for sure.

Long overdue feature.

[vital radish]

On the other hand - its kind of sad that there is not an alternative to OAI. Im sure if there was some competition - things would have looked a lot better at this stage. When people vote with their wallets ( and purses ) - things start happening. OAI know they have the market - so no rush.

[timber shadow]

And yet, I do get the sense myself that they're rushing. I don't get any vibes that suggest they're resting on their laurels

[vital radish]

how will we know though ? They are not communicating anything ? Or have i missed it?

[night breach]

Yeah they don't want to communicate anything with us. Only when they release something new.

[tranquil jackal]

Big company like OpenAI i'm sure they're got eyes on the conversation.

Having an active line of communication with the discord users can also be a recipe for disaster

[timber shadow]

On the contrary, an OpenAI dev has commented explicitly in this Discord that they're a small company relative to the number of products they have. I'll see if I can find. I hope you all have your job applications submitted!! They need the help!

[timber shadow]

#images-discussions message

Valuation doesn't necessarily equate to company staff/resource size

[tranquil jackal]

I know but it's enough to make a man grimace haha

[night breach]

That's honestly just a company excuse. They're right now one of the biggest companies in the world in terms of reputation. They will need to up their game today. Because all these usage caps is just not cutting it.

[timber shadow]

Hey you've said in the past you have your application in right? Glad you're committed to the cause!

[night breach]

Glad you're always so committed to having gratitude.

[tranquil jackal]

I do feel like there are random changes going on in the background each day. 2 days ago i got hit with the usage cap after about an hour. Yesterday i was all day long testing gpts with only a one hour lunch break from it. Today i only got 20 minutes of mere text chatting in/out.

[timber shadow]

Yes I try to imagine like a live dashboard of what the current limitations are but it'd probably be so chaotic and maybe even more frustrating. Like maybe it's regional? Maybe limits only affect some accounts, not all? Who knows how many variables are at play. Moxi has talked about how there are separate 15min, 30min, 1hr, etc limits applied to DALL·E for example, if I'm remembering that right. Kind of like a leaky ship in that respect, but the leaks are insane traffic

[night breach]

@tranquil jackal @vital radish I've made the petition

I've sent it to your DM's

[vital radish]

Thanks

[night breach]

@rough tundra Is there any way I can put a weblink for a petition for OpenAI here without getting muted or having it removed?

[earnest oasis]

guys why it is putting me on waitlist

i wanna upgrade my subscription

[tranquil jackal]

cheers pal!

[night breach]

[tranquil jackal]

I hear they're throttling the subs for now

[earnest oasis]

noooooooooooooooooo

ive got a project to do

[tranquil jackal]

Feel for you man, i know, We're struggling here too even with our active sub accounts. The last 20 min convo has been all about how quickly the "usage cap" message kicks in for us paying customers.

[earnest oasis]

whats the usage cap

message

[rough tundra]

Hey! At the moment we only allow links from OpenAI's website to be posted here. Sorry for the inconvenience!

[night breach]

But this is for OpenAI

[night breach]

[gritty totem]

Hey all, Im having trouble accessing some of my chats. I am receiving a message: Oops, Something went wrong. New chats work though. Another interesting thing is my mobile app does load the chat but when sharing the link to access from another device i get a 500 error.

I have tired all the trouble shooting as well as tried to get help from the OpenAI help/chat to only be disregarded and ignored. Can anyone help? I need to access these chats and back them up. I have a lot of important things there.

[timber shadow]

Is it a Plugins chat by chance, with more than one plugin enabled? If so, it's a known bug, they're working on it! #chatgpt-discussions message

[gritty totem]

I do think i had a few plug-ins. I dont have a way to remove them at this point. The chat was VERY LONG as i was working through several complex items and definitely can afford to lose that data and what resolutions i started. Im looking for any way to access or export/download all of it lol.

Ill check out the link you provided as well.

[timber shadow]

Hopefully fixed soon!!

[gritty totem]

i afford to wait, just not to lose everything haha. thanks for the info!

[earnest oasis]

when am i gonna get accepted

[tranquil jackal]

I think they're well aware of how excrutiating it is for the people signed up already trying to use it... so this must be their way of not making it any worse. I'm going to stay on the hopeful side and see this as a sign that they're scaling up the servers to allow for more traffic.

[rigid musk]

I suggest maybe in #off-topic

[unreal hemlock]

<:book_icon:1171408210398289941> `` Rule 10 ``  Do not bypass the AutoMod filter.

Avoid attempting to bypass our server filters through the use of similar or meaningless variations of words, as well as by incorporating unique characters. Using your server profile to bypass the filter may also result in infractions.

[dry minnow]

Hello, does anybody know how to fix the "Oops, an error occurred!" when clicking on any chats in ChatGPT?

[timber shadow]

Is it Plugins chats perhaps? There's a bug for chats with more than one plugin enabled, they're working on a fix presently #chatgpt-discussions message

[dry minnow]

Nice one, it appears to be just that. I guess it'll limit uses slightly, but I am thankful to have primary use in the meantime while they work on a fix. Thank you for that, I appreciate the speedy response

[marsh steeple]

This is an issue that only OpenAI can fix. Almost all plug-in users are facing the same issue that you are facing.

[signal flame]

I had the same problem, Apparently it is the plugins, I deactivated some of them and tested them. It works again

[honest pagoda]

Annotations Empty

[fallow swan]

Is there a way to get started with a model on my home PC? I'm very new to this, I just want to start with Python to see if I can generate images using my old nVidia 1080 card.

I can't seem to find a great doc on newbie basics

[cosmic crag]

I don't believe you can run openai models locally. If you want to generate ai images, I would recommend looking at stable diffusion to run on your machine.

And you are probably gonna have a hard time with just a 1080 making an image with a resolution higher than 720p

[fallow swan]

I have an M40 Tesla card but it's difficult to get it working on a consumer board.

[cosmic crag]

The fact you have it is cool, though

[fallow swan]

I don't have crazy needs at the moment, just showing myself I can do it/it can be done.

It is neat, yeah...I just wish I had better case ventilation, normally these go in BIG BAD METAL.

They're pitifully cheap these days.

I got mine for $20 lol

[cosmic crag]

Oh wow

[charred parcel]

maybe move this convo to #ai-discussions so we don't make some work for the mods

[fallow swan]

Will do! Thank you

[gritty totem]

Awesome to know, how did you go about deactivating them? I can't get into the chat to do so.

[timber shadow]

Someone from OpenAI said that the Plugins bug should be fixed, can you try again? Maybe refresh/clear cache? #chatgpt-discussions message

[lofty holly]

Does someone know how to make accurate text (spelling) when creating images in DALL E, would be grateful for any insight

[cosmic crag]

DALLE has no sense of text in images

It ranges from bad to outright awful

[nimble onyx]

Yea text is not the way with DALL-E. But if you can get the images in a certain orientation each time and know where the text should go, you can added to it afterwards with code that a GPT can do.

[young rune]

Is this the best place to ask about ChatGPT's plug in feature change, or is that OT?

[young rune]

It seems to do ok with small words. I made a sign with the words "Fall River" and it got it right about 4 times out of 5.

[trail cairn]

plugins are getting killed

[young rune]

AskyourPDF redirects to an advertisement for their new research function which asks you to pay them for access to ChatGPT4 level tech.

[trail cairn]

probably because the new gpt's hurt their business

[young rune]

That's the really wierd part though. The sub I had with that company was to use with ChatGPT4, now they don't have that I will most likely kill the subscription. I swear these guys are geniuses but don't think things through.

Such is the way of buying in during a feeding frenzy though 🙂

[ancient kiln]

If I need a GPTS to have access to a mapping software such as Google Maps API or Bing, what type of API information would I need to input into GPTs? I’m super clueless of API’s and GPT keeps hallucinating when I ask for assistance.

[cosmic crag]

Would you need real time data or just a static map?

[ancient kiln]

Not real time but actual driving distances and not straight line distance.

It’s for planning employee routes and assigning locations

[cosmic crag]

Thats pretty rough...I don't think there is an API for determining routes, I think that has to be done in the full app, but give me a minute to double check

Oh wait, google has the routes API

Bing also has a routes API

I think you need a developer account with both of them before you can use it tho

[ancient kiln]

Thanks. I met a retired Microsoft CTO who is doing this but for rv travelers. Yes I have the developer account for Bing

[ancient kiln]

I just don’t understand the API info and when I ask GPT for assistance it tells me to do things that don’t exist.

[cosmic crag]

What part don't you understand?

If that reads as demeaning I am so sorry

[ancient kiln]

It lists multiple sections of script for the API on their site. Just not sure what I need to implement into GPTs for it to be usable. I’ve only ever plugged api codes into scripts and copied and pasted them. Have very little experience with scripts or coding.

At first when talking to GPTs it told me it can’t even use other sites apis and I had to convince it that I can

[cosmic crag]

Ah, so in Actions, you copy and paste what is openAPI spec. It is a format of describing an API and its endpoints that is easier for the AI to understand

[ancient kiln]

Ok.

[cosmic crag]

I'll look with ya if we can find one already made.

[ancient kiln]

I keep seeing openAPI and thinking it’s OpenAI

[cosmic crag]

I am not finding one, can you?

[ancient kiln]

An existing GPTs? No

[cosmic crag]

No, an openAPI spec of the bing routes API

[ancient kiln]

Oh sorry. One sec

[cosmic crag]

It looks like we have to create one. I, unfortunately, don't have the time to make one of these for you, but I can give you an example and you can just follow the format.

[ancient kiln]

I’d appreciate it. Thanks for your time

[cosmic crag]

You copy and paste your version into the schema box. Then Find out what kind of authentication it needs and use the authentication box below. Then you need a provacy policy for the API which won't be hard to find

And if you need any other info for it, swagger io will have the answers

[ancient kiln]

So I change the description info to what it will actually do correct?

[stark trout]

Is there a GPT you would recommend for solving math specially geometry problems? I would really appreciate it

[ancient kiln]

Hello. Thought I would give you an update. Finally got a schema created. It connects to Bing maps but when I try to even optimize three adresses, it gives this error: {
"response_data": "ResponseTooLargeError"
}

[cosmic crag]

The only parameter you are using is the api key?

There are a lot more parameters that you should be using. Otherwise, like it says, it will give you a massive amount of data that the AI will refuse to read.

[ancient kiln]

Sorry. Just saw your reply. The following is written in the instructions: Geo Assigner is a friendly and professional tool designed to assist users in route optimization using the Bing Maps API, now enhanced to import data directly from URLs. It is suitable for both technical and non-technical users, avoiding overly technical jargon. The tool allows users to upload files with employee home addresses and work locations, and to directly import data from Bing Maps API endpoints using URLs. Users can provide a URL with a specific request to the Bing Maps API, and Geo Assigner will handle the API call, parse the returned data, and use it for route optimization tasks. The maximum driving distance from an employee's home can be specified for more precise allocation of work addresses. Geo Assigner flags locations that require flying due to distance constraints. It guides users through the process, providing clear instructions for data formatting, uploading, and API integration. The tool ensures detailed and helpful responses, maintaining clear communication and accuracy. It can also use the browser tool to access and explain specific sections of the Bing Maps API documentation, aiding users in their implementation process.

[cosmic crag]

Hold on, you got this thing to use browser tools?

But I meant parameters in the bing routes api. Look at the website I am going to send you. It lists all the parameters that need to be sent.

You can delete the message with your instructions, if you want to keep it more private

[dark steeple]

If I am making a custom GPT, will uploading a ZIP for my knowledge work? will it be able to access all the files

[celest crystal]

Dear Mr. Brockman and Mr. Altman,

You probably won’t ever see this, but just in case you do… Thank you so much for all of your hard work, innovation, and well… everything that you have built and given to the AI community. You have changed lives and the world. You have made it better than you found it. I salute you and all that you have done and will probably continue to do. Thank you 🙏

[cosmic crag]

@dark steeple No, uploading a zip is not a great idea. If the AI did try to read what was in it, it would have to run python code and so on. Better to just upload the original files

[wild socket]

Cant get in chatgtp is it down?

[cosmic crag]

Working for me atm

[wild socket]

im just geting

[cosmic crag]

Samsung odyssey monitor?

Not really much you can do here than wait for it to clear

[wild socket]

ya lol

and ok

[dark steeple]

Does chatgpt have a file limit?

[cosmic crag]

I have heard conflicting reports on that, but I believe the number of files maxes at 50. As for total size... I think it is a gig?

[dark steeple]

Do you know any way around the 50 way limit? Since i have more than that

[bronze charm]

It seems like caps are being changed quite a bit at the moment to determine what sticks given new demand
Enjoy the increased usage!

[cosmic crag]

What kind of files are they?

[dark steeple]

they are md files, it appear 10 is the limit

[cosmic crag]

Hmm. If this is for personal use or for a company or something, I would recommend the Assistants API. You can load more files to that. If you need to have it on the chatgpt platform, I would suggest to try to combine them, but make clear delineations

They lowered the file limit. That's bad news bears for mine.

[wild socket]

This does not make sens Eveythang works but when i create a GPT with somethang to do with Roblox deving it says There was a problem preparing your chat.

[chilly snow]

hello, does ChatGPT use the same models availalbe from the Completion API? From the API I get significantly lower quality completions, even when using gpt-3.5-turbo-1106 from the API

[cosmic crag]

ChatGPT uses optimized models, afaik.

[chilly snow]

I see, that would make sense

are there any official docs stating this?

[onyx kraken]

They continued training from here, clearly. But they describe ChatGPT as a sibling model to InstructGPT. https://openai.com/blog/chatgpt

[dusky quail]

I must be doing something right with my custom GPT 🔥 Got this message 4 times in the past 24h hours lol

GPT can't read Gitlab projects from prompting. Is it possible through actions? I am 100% not a dev

Not even a specific repo, but also documentation, scripts and side code lives in one project. Have not found a way to convert to a .txt file, and thought maybe there is a way for GPTs to read Gitlab directly?

[frigid grove]

Hello, when I try to pay for GP Plus chat, I sign up in a certain queue, how long do I have to wait so that I can pay for my subscription? Thank you in advance for the answer.

[trail cairn]

no one knows, it could be weeks or over a month based on much issues there are with capacity

[frigid grove]

I get it, bro, thanks for helping

[stark trout]

Did you already have a Plus subscription or are you trying to get a new subscription?

[tight fable]

anyone got an idea why I get "the response was too large" when trying to upload something using customs actions? thanks

"response_data": "ResponseTooLargeError"
}

[dim sage]

w h y

[stark trout]

Quick question: how is an Assistant different from a GPT?

[timber shadow]

The Assistants API is a developer tool available on platform.openai.com. They share some similar functionalities, but in general, like all things API, Assistants is for those making their own tools, rather than using the curated (and in ways simplified) experience of custom GPTs on ChatGPT

[stark trout]

Oh I see! Thank you very much for this thorough explanation ☺️

[frigid grove]

I'm trying to get a new subscription

[stark trout]

Ah. Yes, it makes sense in that case. Thank you for the clarification! 😊

[charred parcel]

subscriptions are paused right now

[onyx kraken]

It's hard to make sure that the AI completely forgets anything it ever knew.

It's had multiple training cut off dates.

Any time you ask a question like "What time is it?" the answer changes, right?

The AI knows it has more than one correct answer to that question.

It also knows it maybe sometimes should make up answers to that question

(example, ask it to pretend a date for its training cut off; if it agrees to play that make believe game it will give you another, made up date)

Training helped make sure it's more likely to give you the actual current cut off date. But wrong answers for stuff that has changed since it was taught the answer... is hard to fix.

Harder for humans to learn new stuff over old learned stuff too.

One answer is 'just regen the answer' and check to see if you keep getting the same answer or not

[tulip beacon]

"Message in conversation not found" might e possibly the five most annoying words in existance, anyone else had this problem with chat gpt over the past couple days?

[deft echo]

Like the operation of the framework without the knowledge of the core technical team that leaves or was fired

[random iron]

Folks: For assistants API, can we have it use finetuned model vs. the standard model?

[spark cliff]

what does "user quota exceeded" mean? i tried to upload several pdfs into chatgpt4 and I get this message also I could only use fror 15 messages.. is it normal? why does it happen this always on saturday and sunday?

[whole fossil]

Does anyone know why the cap is 40 messages per hour instead of 50? For 3 hours

[bronze charm]

Much higher usage resulting from the Devday opening keynote

Trying to make sure everyone can use it

[grand stratus]

Wonder if people can help: I already have a Plus subscription, but since the pause on new accounts, I appear to no longer be able to access Plus features. When I go to the iOS app it’s back to running 3.5 instead of 4 and I can’t switch, but when I look in my settings it says I have a Plus subscription. Restore purchases and updating the app doesn’t help. Any ideas?

[timber shadow]

Hello! Is it the same behavior when you log in on the web version, chat.openai.com?

[grand stratus]

Yes. When I try to switch to 4 it pulls up the window that suggests subscribing to Plus.

Behavior is almost like I got kicked from Plus, but it still says I have it on iOS (see image).

[timber shadow]

Has your Plus subscription been successfully auto-renewing in the past, or have you been manually re-subscribing? In other words, are you sure your account's Plus sub hasn't lapsed? I see it in your screenshot, I just want to double-check.

[grand stratus]

Yeah, I’ve been subscribed successfully on a Plus subscription since they added 4, with no issues.

[timber shadow]

Specifically, it's been auto-renewing successfully and you've never had to touch your billing info since? Are you able to check your bank statement to verify when your last renewal charge occurred?

If you're indeed still active on your Plus sub, it may be worth doing some network troubleshooting. Do you have a VPN installed? Can you try on another device, or another network?

[grand stratus]

Looking at my account, it looks like the October billing didn’t happen, but I have no idea why. I didn’t touch it, and have no record of declined payments.

Guess I just have to wait for them to open up subs again.

[timber shadow]

That is unfortunately probably the source of your trouble 😦 I can't speak to why this hiccup may have occurred, but, unfortunately, you will most likely now be affected by the pause in new Plus subscriptions. Really sorry about that

[grand stratus]

Yeah, sounds about right. Well damn, that's frustrating

appreciate the help though

[timber shadow]

That is frustrating. Absolutely, sorry for the disappointing result of the troubleshooting. I hope you can get off the waitlist quickly 🤞

[sturdy apex]

Hey, i get a missing access_token while using the Google Sheets API with a Custom GPT. Can somebody help me, how to fix this

[slender lantern]

I have had several content violation warnings and now an email that my account could be shut down, due to using research plugins for topics like investigative interviewing and forensics. I use the plugins to find research articles and ask questions and interact with documents and my use is professional, academic and completely appropriate. The warning that led to the email suggesting my account could be suspended came from asking what the Independent Inquiry into the Rotherham scandal revealed about considerations investigators should consider when interviewing vulnerable subjects in sensitive cases. How can I prevent content violations for education and research?

[cosmic crag]

A good question, but unfortunately, we are at the mercy of openAI censorship. You might be able to appeal, but otherwise, you'll just have to wait and be careful of what you search for. OpenAI censorship seems a bit random and inconsistent.

[timber shadow]

I just did a search about the general topic of the scandal you mentioned (which I do not recommend anyone else do) and, while I appreciate that your use is research-oriented and based in educational endeavors, I am not surprised that you ran into several content violation warnings, per the usage policies of OpenAI.

From my personal perspective: the topic(s) you're referencing just may not be suitable for use in an emerging AI tool like GPT. This is because moderation is automated right now, so the sensitivity needs to be cranked to make sure that there isn't misuse on such topics, as that would be catastrophic. In other words: it's just hard to contextually identify, from a moderation perspective, what should or shouldn't be allowed on sensitive topics like you've mentioned. The content violation warnings you ran into were a good litmus test for this--they weren't judgements on you and your usage, they were just judgments on the content, saying "Yeah, we can't help with this. A human should be doing this."

[sinful grail]

Good night. I write from Brazil. I'm creating my GPT and I'm having a hard time publishing. Could someone give me some support or guidance? please

[astral quest]

In the settings you need to add a link which brings the user to the privacy policy of this GPT

[slender lantern]

Thank you, I agree there should be limitations. The research tool suggested the case as academically relevant to a less sensitive question, but then flagged the chat when I asked further questions about the case study it had provided...

[timber shadow]

Unfortunately, the model doesn't know if it's about to (potentially) lead you into asking something that could violate the usage policies--that onus is always on the user, from a ToS perspective. There could definitely be some potential precautions put in place like for example, "Hey I can give you this ref but don't ask me anything about it bc we'll both get in trouble", but for now its self-knowledge just isn't comprehensive enough to achieve this, so we have to tread lightly. I hope you're able to ultimately avoid or appeal a suspension, and then use the tools per ToS from there--sorry you're running into this. And on a personal note separate from GPT, I appreciate that you're working on this stuff academically--the world needs you!

[sinful grail]

It worked.

[astral quest]

Glad to hear that!

[sinful grail]

Thank you for your attention,

[slender lantern]

Thank you, I found a support email, and made this point, because while safety mechanisms need to be in place, we also need these fields to keep up with others which will move ahead with the benefit of Ai and data analysis. I'm hoping academic validation and exemption can be part of user registration in future.

[timber shadow]

That is a good idea. If you feel up for it, you might consider writing up a post for #1070006151938314300 to this effect as well.

[frigid basalt]

when using the api multiple times in one program run, should i declare a new OpenAI() for every prompt, or should i make just one at the beginning and reuse it?

please ping to respond, thanks

[stark trout]

Hi all! Are there any linguists (or anyone really good at language and grammar) here? I’m trying to build a costume gpt and I would greatly appreciate the help. Thank you in advance! 😊

[bronze charm]

No need to make another client! Simply create another request with an updated messages list if you want continuity

[frigid basalt]

whats it for anyway? why not just make a request right through the module

[bronze charm]

It initializes a bunch of stuff, including your API key

[odd rapids]

Hello! I've ran into an issue with usage limits. I've been a paid Plus member for nearly the entire year and have never hit any limits (I don't abuse it). Has anyone else been having this issue? Can anyone shed some light on what's happening?

I didn't even reach 25 inputs and was locked out. I typed in roughly 23 times.

Then told to wait 1.5 hours

(give or take on the time, it didn't say 1.5 hours, it was just what I estimated the time would be when it gave me the specific time I could come back)

[timber shadow]

It does count messages sent went building a GPT, as well as any edit & resubmits you did, or any regenerate requests you submitted. Is this possibly relevant to your usage?

[odd rapids]

I'm a light user. I've never experienced hitting any limit before. I don't use any API's or anything.

Is there a way to see my usage?

[timber shadow]

On ChatGPT, not other than counting messages you sent, and then considering the other things that "count" that I suggested. One I forgot to mention that also counts is any messages you send in the "Preview" pane when building a custom GPT. Another bit that might be relevant is that the rate was recently lowered from 50/3hrs to 40/3hrs

[odd rapids]

Okay, please forgive me, but I don't know what you mean by building a custom GPT. I just ask it questions. I use AskYourPDF (which I pay for) and I'm a paid Plus GPT patron.

[timber shadow]

Gotcha, sounds like you're not using it then! So you'd probably just be affected by (1) messages you send, (2) any sent message you edit and then resubmit, and (3) any time you ask ChatGPT to regenerate a response it already gave you. Do you do all three of these things, or just (1)?

[odd rapids]

Mostly just (1) but the Chat just stops sometimes, giving me no other option but to do (3). I have done (2) but it's not constant. I've always had enough provided with what I have always been given. Never hit any use limits before.

[timber shadow]

I see. You may just be running into the recently lowered usage cap then. You can see the current usage cap on the web version of ChatGPT by opening the model selector and looking by "GPT-4"

[odd rapids]

Well... I just counted. I didn't even hit 25, no re-gens, no re-submits.

Does the counter reset every 3 hours or just keep counting from when you last entered anything?

[timber shadow]

It's a rolling cap, so think of each message you send having its own 3hr timer before you get it back.

[odd rapids]

As in, it kept my last use count, which would have been 16, then ad the 24 I just entered.

Hmmm.

A counter would be helpful.

Genie gives you wishes, craps out, re-submit the same wish, get 1 for the two you wished, repeat, how many wishes do I have? Oh, none.

None left for this pleab.

But I didn't hit 40, so something is out of whack.

[timber shadow]

If it's at all helpful to contextualize for you, I like to think of it as 1 message every 4.5 minutes. So long as I don't exceed that, I can use it all day.

[timber shadow]

You may consider posting in #1070006915414900886 if you think your usage was below the reported current cap. OpenAI staff are periodically present in there.

[odd rapids]

I understand you don't know me and this is the internet, but [insert deity/belief system token here] as my witness, I don't ever come close to that.

Yeah I did a bug report.

And was a little shocked by thier form to fill out.

The form where i ask for more usage when you pay and barely use it.

[timber shadow]

If you mean the alert they've been showing with the link to the Google form: that's separate from the #1070006915414900886 channel on this server. Tap the link to go there, and you can submit a bug report there. Sorry if I'm misunderstanding you

[odd rapids]

Oh

I did a bug report in the Help automated chat on the website.

I can do one here too.

[rugged wing]

Anyone know how to fix the message not found in conversation error?

[odd rapids]

Thank you for providing me more information than I had, I appreciate you.

[timber shadow]

Heck yeah, we'll get through these rough patches with the usage cap for sure 👍

[analog shore]

Hello, the message cap is 25 again?

[bronze charm]

40 it seems, though some report different numbers which may indicate fluctuation amongst the userbase

[analog shore]

Thanks for your answer... i felt like it was like 15 or less...

[wraith nova]

hey guys, i have 2 chatgpt account, lets call acc 1 and acc 2, in the acc 1, i have chatgpt plus bc i paid, but i dont want to paid again in acc 2, how do i use the 2nd acc or the email of 2nd acc to access gpt 4?

can i use api key of acc 1, or something?

[timber shadow]

There is no function like that in ChatGPT Plus currently--each account could only use GPT-4 with separate Plus subscriptions. I like what you're getting at though, you might consider creating a post in #1070006151938314300 for some authorized "GPT-4 sharing" feature!

[red haven]

Hey Champs

Anyone knows how to "resurrect" a thread in Assistant Playground? My conversation with assistant disappears after I leave Playground (but I can still see it through API).

[rugged wing]

There is the historial button

You can see your past conversations and recovery them

[red haven]

Could you give a screenshot? I don’t see it anywhere on my end

[south geode]

Hello, for a long time, I've been trying to purchase a subscription to ChatGPT Plus, but I can't seem to make it work. Could you help me with this?

[charred parcel]

there is a current pause on new subscriptions, you need to join the waitlist

[south geode]

"Are you aware of how long the usual wait time is?

[charred parcel]

no, it can vary depending on how many people are in front of you

[azure coyote]

when I start to type in new text prompt in vary region all disappear its not ley me type amything its go back and I have remix mode on

[south geode]

Where can I keep track of the whitelist?

[charred parcel]

have you joined the waitlist?

[south geode]

no((

[charred parcel]

when you go through the subscription steps you will be put on a waitlist and notified by email once you can subscribe to Plus

[south geode]

Yes, exactly such a notification appeared when I first clicked on the purchase.

[south geode]

"It seems like I just have to wait then. :("

[charred parcel]

yes, you just need to wait

[south geode]

ty a lot

[fiery canopy]

hello it seems like all in-app purchase of chatgpt are closed, even though i can now purchase on the website, is it a bug?

[charred parcel]

new subscriptions are paused, you need to join a waitlist

[fiery canopy]

i saw this on the website and there's nothing called waitlist now

[charred parcel]

go through the steps to subscribe to Plus

[fiery canopy]

but i want to purchase through mobile app bc i still have credit on store

[charred parcel]

you can't, new subscriptions are paused

[fiery canopy]

got it

[brave gyro]

I didn't even have the option to go through the steps to subscribe to Plus, I just clicked on "Sign up for wait-list".

[crimson grove]

Are there more people who have issues with captcha's? I get it evertime after a prompt... It's on useable..

[charred parcel]

what browser and OS are you using?

[hot tide]

Is this related to sam altman being fired of the board ? Direction changes ?

[crimson grove]

Thank you for the fast reponse! I already found the issue.... My date / time of my os where wrong.... #stupid my appoligies..

[charred parcel]

the subscription waitlist exists as OpenAI attempt to adjust for the demand surge since Dev Day

[hot tide]

Thanks 👍

[near glade]

Hi there, I tried the new feature of the custom GPTs and I'm wondering if it basically the same as "finetuning" a model through the API. That feature isn't really new but I haven't used it yet. Is there a technical difference or are the new customGPTs basically a nice UI for finetuning GPTs?

[still sierra]

In the future, chatgpt gpts is expected to introduce a similar approach to git version management? Now I'm worried that if I change my Settings, and then the result is not satisfied but forget how to set the original

[timber shadow]

Hello, good question. Custom GPTs are closer to the already-existing "custom instructions" feature than they are to API fine-tuning. Fine-tuning, as I understand it, is like actual training of the model on a set of data such that the model's inherent parameters are updated to "know" the data in the knowledge, like how it "knows" everything it knows already.

Custom GPTs, on the other hand, are a bit simpler. They consist of three main sections:

1. Instructions, and name/descrip/convo starters. These are like the custom instructions feature, except you can type up to 8000 characters of instruction, instead of 1500+1500 in CI. The GPT will then keep these instructions "top-of-mind" in everything it gets from you, and in everything it responds to you with.
2. Knowledge files. You can upload up to 10 files that the GPT can then search for relevant info, and then form a response. Knowledge files don't stay permanently in memory like (1) does. Instead, they're basically just "available" to the GPT to search in when you need, kind of like how it searches the internet.
3. Actions. These are advanced features where you can hook your own API calls to your GPT, allowing you to connect the GPT to specific services to accomplish unique tasks. Actions require a bit of programming know-how to set up.

[timber shadow]

You might consider adding a post to #1070006151938314300!

[bitter swallow]

anyone knw how to get help on the forums - my account ha been in libo sice the day I joined and I can actually contact anyone on there .... https://community.openai.com/t/account-temporarily-on-hold/263433

[trail cairn]

the forum is for devs so you probably wont get much help for non api/dev stuff

[bitter swallow]

Not what the channel header says -- "A place to ask the community Got a question about anything OpenAI? Ask here and the community will help you find the answer." but I'll go find another part of the park to play in then ...

[trail cairn]

the site you linked is at least is the dev forum but not sure if you are seeing something else. "Welcome to the OpenAI Developer Forum!"

[trail cairn]

for generic issues #1070006915414900886 , https://help.openai.com/ or support email

[rapid cypress]

Hi, I have tried many times to install a plugin, but I always encounter the message: "COULDN'T LOG IN WITH PLUGIN. can help me?

[charred parcel]

which Plugin?

[rapid cypress]

any

[charred parcel]

what is the last one you tried?

[rapid cypress]

maybe is for this pluggin?

[charred parcel]

ChatWithPDF requires an account with ChatWithPlugins to use it, did you know you can chat with any PDF directly within ChatGPT?

[rapid cypress]

ok ok how much information, I didn't know this, do you mean I can upload a pdf with chatgpt?

thanks Robert for your help!

but this pluggin?

[charred parcel]

just drag a PDF file into GPT4 and it will upload and you can chat with it

[charred parcel]

SmartSlides also requires an account with SmartSlides

[rapid cypress]

yes i just do it and work! thanks!!!

but i create an account and dosent work

[charred parcel]

that's a problem with the Plugin provider, not OpenAI

[rapid cypress]

ahhh ok ok

thank you very much!!!!!!!

[rugged wing]

Is there a way to change my OpenAI account email? The Email address section is greyed out

[rapid cypress]

you are amazing! thanks for your time

[charred parcel]

you can't change your email address, no

[rugged wing]

Why though, that's dumb

[charred parcel]

I don't know

[ashen sandal]

hi friends, running into a problem (or bug).

If I upload the logo for my custom GPT after some hours it shows me this. Anyone know how to solve this?

[lucid wolf]

Waitlists were pre-2020.

[wise lodge]

Anyone have example action schema they'd be willing to share? I can't get anything to work. ☹️

[modern tundra]

Does someone know or have documentation on how to have it so that my custom gpt imports modules and packages into its environment?

[cosmic crag]

You can tell the gpt in its instructions to use certain packages when creating its own code. Describe how you want it to use it. Or be like me and get tired of the gpt failing to follow instructions and write your own python script to put into knowledge

Even then, its env is limited

@wise lodge

And if you still can't get it to work, DM me and we'll figure it out

[regal lynx]

questions about moderation can be directed to @burnt galleon

there isn't a set response time, but the server staff team will look over it 🙂

[fallen raft]

is any version of official openai dall e available for free???

[mellow mortar]

Did anyone get the promised $500 DevDay credits in their account (for attending the conference)? I don't see anything when I look at my billing/usage pages, and this dashboard says none found:

[cosmic crag]

Bing image creator

[fallen raft]

thanks for the info beard guy

[cosmic crag]

You are welcome!

[candid schooner]

might be a browser caching issue

[fierce sand]

Anyone have any success getting a website link for their GPT. I validate my DNS but i can never flick the control on to use the website URL even thoough it says the DNS is validated

[broken fern]

Something went wrong. If this issue persists please contact us through our help center at help.openai.com. it is popping up repeatedly. I have tried to disable all plug-ins in Chrome(My office computer), delete cookies, and not using VPN.
is there anyone that can help on this matter?

[cosmic crag]

Try using a different browser

[broken fern]

I have tried using chrome and microsoft edge

is there any possiblity that my organization kinda blocked chatgpt

[cosmic crag]

They use the same chromium engine. Try firefox or opera or something

If the org blocked chatgpt, there would be a whole connection error instead of the "something went wrong" error

[broken fern]

so it should be related to browser right?

since I am not using vpn or something

[cosmic crag]

I think so, yeah

Otherwise it islikely a problem with openAI, which we have had plenty of server bugs and stuff lately

[broken fern]

It has been a week and my chat gpt is working on my phone or my own PC

but in my office computer it is not working it has only chrome and edge , I need permission to download firefox or opera

[cosmic crag]

Its possible that your org is blocking a resource that chatgpt needs to work

[broken fern]

I have just downloaded the Mozilla Firefox and it is wokring now

Thanks for your help.

[cosmic crag]

You are welcome!

[tribal talon]

It seems some people had that issue. Recommend emailing the guests email maybe or support. They can confirm it all and help.

[wise lodge]

Anyone got a vector database running as a custom GPT action? I've got it connecting to Weaviate now but not sure how to query it. looking for the right way to do it or a different database that's easier with a REST API

[vale glen]

[compact cobalt]

Hey. have someone already try The PubMed Research ChatGPT plugin?

[livid egret]

If I would like to use my ChatGPT Plus for a bot, how would I do that? I have an LSL GPT I made with GPT Plus I'd like to use in a video game I play. That is, I want to make a call to it for a response rather than using the website. I don't want the API plan that isn't for custom GPT-4 Plus as I am not going to give anyone else access, I just want it for me. How can I use the key that lets me only make the normal Plus amount of calls, that is what I want to do basically just do the same chatting I am now just not in my browser rather I am doing it through a script in a game?

I can do the script part but, at the end of the day the script needs my account's API key for the ~40 calls/3 hours (is that 80 now? I forget).

[honest tulip]

i was trying to upload some files to the gpts i created since last week, but i found that i cannot upload any file any more since last Friday, it is strange it doent show any storage limite or else, it just repeatly saying Error saving draft when i uploaded some files to the knowledge

😫 anybody knows why, it is very frustrating

[wraith nova]

Anyone know why there is no button to cancel plan?

Please help me

[trail cairn]

can you open the same page on desktop computer? or switch to desktop version from the browser settings

[wraith nova]

Sure, wait a little, i have some works

[wraith nova]

Still cant see

There is no way to cancel my plan:(

[trail cairn]

what does the say at top of the page for the next renewal date?

[wraith nova]

it doesnt say anything

[wraith nova]

this is the top picture

the words at the top is " Phương thức thanh toán" which means " Payment method"

[tranquil jackal]

anyone else getting errors with image generation inside gpt builder?

Also guys, i'm not sure if this is allowed. I want to create a harry potter quiz and trivia gpt. I'm aware that the profile image for the gpt is not allowed to represent harry potter, but is the gpt itself allowed? Or is that against policy?

[trail cairn]

harry potter is IP and some of the stuff trademarked so it would probably get removed at some point

[tranquil jackal]

IP?

[trail cairn]

intellectual property. legally you would have to have a license to use their stuff since fair use does not cover stuff behind paywalls, etc.

[halcyon charm]

has anyone suceeded in getting chatgpt support to unlink google account login, such that you can use email and password to login instead of google login?

[vocal basin]

just confirming custom GPTs will have id's so we can export them to our apps in the future or is this a bad idea should we create a lot of logic in the app and use the assistant api ?

[tranquil jackal]

suppose i'll have to keep this gpt private for a while. Thanks!

[rigid musk]

Some people did succeed with [e-mail]

[vocal basin]

I mean once the store is released amd I mistkaen does anyone know "just confirming custom GPTs will have id's so we can export them to our apps in the future or is this a bad idea should we create a lot of logic in the app and use the assistant api ?"

[trail cairn]

api and chatgpt are different environments and services, there is no connection between them

[obtuse zealot]

HI there (i did post it in off-topic room but i think that's the place to post it), can some good people, answer a question related to Message cap?
I can't understand how it works since, when i subscribed to plus plan i had 50 message / 3 h , then after two days 40 message / 3h , and now even if chat-gpt says that i have 40 message, i was able to reach the limit after 25/30 messages. how's that?

[vocal basin]

but if I built a customgpt I love, how do I share it with non gpt4 users, i mean i dont mind paying the embedding costs to have it on my app? it is a CSR thing

[trail cairn]

the message limit may change at any time or remain the same for a long time. it is adjusted based on server capacity. it was lowered to 40 last week around the same time they stopped taking new plus users

[trail cairn]

you cant as gpt's are chatgpt plus only feature. if you want one that anyone can use then you need to rebuild it on the api side and pay for its usage

[vocal basin]

Thank you so you saying is I need to build the synchronize process in next.js in the server.... and not use assitant api or customgpt as both dont really serve the purpose of solving my problem

[obtuse zealot]

yes i figured that out, but it keeps saying 40 message cap and i hit the limit after 20/30 max (counting errors/refresh/ecc)

[vocal basin]

I am just confirming before I put a ton of work in

[trail cairn]

you can use anything you want like next.js but you need your own backend and web site. you can use the assistant api or the other api services to replicate your custom gpt from chatgpt

[trail cairn]

40 is the official limit but i've seen others say they sometimes hit already at 25

[tranquil jackal]

Has anyone else been having issues with image generation inside the GPT builder? I've also noticed i cannot manually upload an image either.

[vocal basin]

ty

[stray owl]

Hey can I get a moderators help

[rigid musk]

Hi, to reach a mod you can dm @burnt galleon

[tranquil jackal]

Nevermind, image generation seems to be working for me again.

Looks like Sam is moving to Microsoft now.

[tranquil jackal]

Does anyone know if multiple OpenAI accounts can be TXT registered to the same website? For example, We have a website that we run with multiple partners in the business. They each want to get involved in their own unique OpenAI development, specifically building custom gpts. However, we all want to do it through our own OpenAI accounts, so that any potential profit sharing by OpenAI is done fairly through each person's accounts. But we want all the custom gpt's to resolve back to the website identity. Is it possible to add multiple TXT records linking several OpenAI accounts to the same website?

If this question is better suited to community api help i can post it there instead.

[foggy musk]

Why does 4.0 keep retelling me information from a PDF incorrectly? Is there any way I can make it answer correctly?

[amber breach]

Okay, so the auto mod keeps blocking my message, so I'll try it another way. Has anyone successfully created an action in your custom GPT that is able to make a call to a certain buy browse API and define the market place? The certain market place documentation states this should be a parameter value, however the Action editor throws an error when you define an "in": "header" parameter value. If you use "in": "query" to define there is no error however it ignores the market place ID and just uses the default market place. If I define the "headers" outside of the parameters the marketplace ID is once again ignored. I've tried many different combinations of everything I can think of, been in tutorial heck for days and still, no go. Any help would be greatly appreciated.

[elder estuary]

when it says budget does that i mean if i send an image lower than 512X512 it will be less than 65?

or is 65 the minimum

[vital radish]

this looks new :

[brave gyro]

It isn't, it's been there for a long time.

[vital radish]

really - ive never seen that before

when it stops ive only seen the regenerate button

[brave gyro]

Yeah, I've been using it with GPT-3.5; maybe it wasn't available to GPT-4 until now. I'm not sure, though, it's weird you didn't see it before.

[vital radish]

ok

[hazy wind]

Has anyone figured out how to get the builder profile active to the public?? My website button is unresponsive. The website is verified through DNS. I have GPTplus and OpenAI paid accounts. There is no "name" option as I have seen on several other users' screenshots. Any help would be appreciated!

[glad fog]

Can someone help me find a downloadable Mac or PC App that I can give my GPT4 API and for it to have simliar capabilities to ChatGPT without the freaken limits.

I know there are a few out there that actually charge monthly for the service (not interested)

I am looking for an open source app, that wouldn't charge me (when I am already paying for GPT4 API)

[rugged wing]

You don't pay monthly for the GPT-4 API

that is a pay per token commercial offering

[glad fog]

yes, i know that, I am looking for an App that doesn't increase this monetary charge

an Open Source app that works with the API

honestly at this point anything better than ChatGPT, i am so tiered of the limits in place screwingg me every time i get going

[astral heart]

so you want a solution that you can pay a flat fee for what could be infinite usage?
this is flawed in so many aspects. what would be holding one person from getting this "infinite" plan and distributing it to everyone?
and especially with AI, which is very reliant on expensive hardware, so they can't let you infinitely use it

a better approach would be you calculating how many tokens you would spend per month and considering that your "monthly" plan

[glad fog]

no, i can pay for the API, this is no problem, i just want an open source app that uses the API so i dont run into limits after sending ChatGPT 10 freaken messages

I understand that it costs to use the tokens, im happy to pay

Im not happy to pay another monthly cost on top of the cost of the tokens

Like paying for ChatGPT but also paying for the tokens you use within ChatGPT

I'm just looking for an OpenSource app that makes me just focus on paying for the tokens

[astral heart]

thing is, keys belong to organizations, and if an organization lends you a key for you to use (even with it allowing you to pay), it has rate limits that are in place to impede big companies from monopolizing openai's apis, so i think that unfortunately this problem will remain for a bit longer

the rate limits are pretty stingy :/

[glad fog]

ok, so you don't know of an app, thank you

does anyone know of an app, please dont respond with messages like this, i am looking for an App not advice

[glad fog]

I have my own key.......

[astral heart]

and you're getting rate limited with that key, is that right? that's what you complained about previously, right?

so i dont run into limits after sending ChatGPT 10 freaken messages

what are these "limits" and what is your problem exactly? does this help you?
https://platform.openai.com/docs/guides/rate-limits/free-tier-rate-limits

wait, do you want just a frontend that calls openai? something you can plug your key in, and it serves as a chatgpt interface

but while you use your openai key for the requests?

if so, just google every-chatgpt-gui (can't paste links from github, just got timed out lol)

[foggy musk]

it's because my prompt wasn't properly received on the first message and the following ones' turned out to be hallucinations without any salvage. Making a new chat with an improved prompt (that didn't induce hallucinating) fixed it

[clear geyser]

I'm not 100 % this is the correct channel for this, the upload button on my ChatGPT is obscuring and obstructing the text box, it happened at the last update last week, does anyone have any idea how to fix it?

[loud heath]

I've had it come up in 4 quite a bit - usually it's if you make a request which is reasonably long (but not absurdly long - do that and it'll either explicitly refuse or only do part of what you ask) - roughly 20 paragraphs tends to trigger it

[brave gyro]

Ah, I see, thanks for the info! I'm not a Plus user, so I know little about how GPT-4 works.

[zenith solstice]

hello, I found Creat a GPT cannot upload further documents after uploading 10 documents, it would show "error saving draft", anyone met this problem?

[cosmic crag]

That is the current max

It won't tell you that you hit the max, but 10 is the max

[stark trout]

Hi all! I am trying to have chatgpt write an essay in a similar format to my own previous written essays on similar topics (similar tone, language). Would it be better if I build a gpt and upload my previous essays with their related topics as knowledge or just use a GPT4 chat and upload my papers? Thank you in advance!

[cosmic crag]

Having chatgpt or any generative model mimic tone and voice in text a really hard thing to do and there is no easy way for you to get that done.

[stark trout]

I see. Thank you!!

[timber shadow]

To expand on the feedback you got from Socrates: one challenge in getting mimicry at an essay-length of output is that the model would (probably) use a lot of its context just understanding your example essay in the first place. When you upload files of length, either to a custom GPT or to a single chat, the model doesn't internalize it all at once into context, either. Instead, it will search through the file and try to find context most relevant to some attached query, similar to how it searches the internet.

So, to achieve a result similar to your goal, you might try building the essay in smaller chunks with ChatGPT. By this I mean, offer a few introductions you've written in the plain chat entry box, then ask for help writing just an intro to an essay in a style similar to the examples you provide. You could theoretically use this process, with a little human intervention, for a whole essay.

But in general, the idea of (1) providing examples and (2) requesting similar content is totally possible, and is actually one of the best ways to work with ChatGPT. You just have to operate within its length limitations to both understand and output content.

[stark trout]

Oh the essays are very short (300 words max).

[timber shadow]

Seems a good case then! I'd still avoid uploading and would instead just copy/paste them in-chat as text. Just let the model know, "I'm going to provide you some examples first, then I'll let you know when I want to work on the new one" or however you wanna do that 🙂

[stark trout]

That sounds like a fantastic start! Thank you so very much!🖤🖤

[quiet plover]

Hi Guys! I've been dealing with sever dyslexia my whole life and chatgpt has been a godsend. I'm running into problems using chatgpt3.5 while doing my uni course and wish to buy chatgpt4 (plus?). Is there a waitlist? Is there any way I can email or chat with someone that may be able to provide plusor an alternate program to plus? Please @ me if you respond it is hard to read through other messages.

[quiet plover]

please direct me to the correct channel for this question if probable also ❤️

[timber shadow]

Plus subscriptions are currently paused due to high server load, and yes indeed there is a waitlist to be notified when subscriptions are open again! You should see it when you click the "Upgrade to Plus" button

[bitter swallow]

what about theopenAI playground?

[bitter swallow]

help.openai.com is pretty useless except for very specific issues now, no opportunity to send general feedback or ask for general help anymore.

[hazy lily]

I keep being referred back to 3.5 before maxing out smh. Annoying.

[muted loom]

I need help with this error, authorization URL, token URL and API hostname must share a root domain. I cannot figure out how to solve it, could anyone help or maybe point me in a direction to where I can get some help?

[stark trout]

Hi! Another question: does anyone know what causes the “Error analyzing” when using the analysis? Thank you again!

[cosmic crag]

Remove auth from the schema. Only add it through the authentication menu you see at the bottom of your screenshot

[cosmic crag]

It could be a lot of things. What are you analyzing?

[stark trout]

just an image of a simple arithmetic. It writes out the process, and it is fine every now and then. Then that error just occurs out of nowhere. I know, Math isnt its strongest suit 😅 But this error is just odd.

[cosmic crag]

OCR, then. I think this is might be another to blame on openai servers or limitations of gpt4's ocr. Unless it's handwritten and hard to read

[stark trout]

Oh no it's not handwritten. I have typed it all out and screenshot it.

But also, I dont think it's the OCR, because I even typed it out and I got the same error.

[muted loom]

@cosmic crag I gave that a go, problem is still there though

[steel monolith]

Hi, can anybody tell me how to get chat gpt to analyse data sheets with tables like this and format into csv please?

the pain is real having to write out pages of csv 😦

[vague swan]

is there a way to send images to GPT's via an API? I've been trying and failing. I've been trying to send it as a zip, jpeg, and base64. zip and jpeg, it fails as soon as it gets it, base64 it fails to decode

to clarify: trying to do this as an "action" in a GPT

[rugged wing]

does gpt builder count against our quota? i think it does. it'd be nicer if it didn't.

[vague swan]

I know testing API calls does. Has been KILLING my quota

[fiery crag]

Are the new Models available via Website , or only via API?

[rugged wing]

i hope they address this soon

[wise lodge]

Yes gpt builder counts

[lost roost]

Hi guys - do you know how to get to the max. tier possible. ?

Even if I am spending I get stuck in tier 1

[steel forum]

I use GPT for work and recently after upgrading I get a "Message in conversation not found" error for anything I ask it. Anyone know a fix for this? I have not been able to get my normal tasks done now for several days.

I have already tried clearing cache and cookies and using another browser. Nothing is working

[abstract spear]

Is there a reason it doesn’t have me as subscribed to plus on the app?

[turbid crag]

what happened? anyone knows?

[cold yacht]

it allways will, it counts towards your quota because its using the model. if it didn't count they would just be volunteering to shoulder the cost of all custom gpt development

[undone crater]

Hello guys,
I'm encountering a problem when trying to use ChatGPT-4 and ChatGPT-Plugins in various web browsers, including Firefox, Edge, and Chrome. When I type something in the chat and press enter, nothing happens, and I receive the following error in the browser console: "Uncaught (in promise) Arkose script failed to load."

I have tried the following troubleshooting steps:

• Clearing my browser cache and cookies.
• Disabling browser extensions.
• Ensuring that my Firefox, Edge, and Chrome browsers are up to date.

However, the issue persists. Interestingly, when I switch to ChatGPT 3.5, I do not encounter this problem. Additionally, I've tested ChatGPT on the Android app, and it functions without any issues.

Does anyone know how to fix it?

[still sierra]

How can gpt summarize the content into a mind map

[gloomy elbow]

I'm really stuck on getting GPT to use my google cloud api key to interact with my google calendar

[nocturne cosmos]

Hi everyone,

We would like to create a GPTs dedicated to image recognition for alternative communication. The objective is to provide a text database that defines an action and associates it with an image. As an output, we'd like to display an image that replaces words in a sentence, for example. The output would be three pictograms. We tried using a PDF document, then Excel with the image URL and Word with the images embedded, but it didn't work. The image only appears when loaded directly into the database. However, we are limited to loading a maximum of 10 images. Is it possible to help us create a template that would allow images to be displayed as output instead of text ?

[ashen sandal]

anyone know how to solve this?

[wooden sleet]

Good morning/afternoon/evening all, Hopefully someone can help. I am struggling to access Dall-e 3. I have a GPT plus account, i select chatgpt 4 and start a new chat, but all the generation is on dall-e 2, if queried it tells me its dall-e 2. even if i browse to the dall-e openai homepage, and click the "try in chatgpt" Dall-e 3 URL, it goes to dall-e 2. i must be screwing something up but cant see what i'm doing wrong. can anybody give any advice?

[rigid musk]

Are you using chat.openai.com

[timber shadow]

Hello! Do you have a VPN installed? I'm also wondering the same for you @turbid crag

[timber shadow]

Hello! Do you have a banner at the top of your screen asking if you want to try an alpha version of ChatGPT in another language? If so, try accepting/closing it, and trying again.

[tawdry bloom]

Does anyone happen to know how to add a password to a chatgpt account made with google? Even after putting in the email on the forget password screen it never sends it :(

[timber shadow]

I don't think there's currently a way to change your sign-in method, so unfortunately since you're already linked with your Google account, using the "forget password" screen won't trigger anything. If you just want to change your Google password though, that'd just be through Google's own password reset

[rigid musk]

To add to solbus, you can try to contact [e-mail] if they can change the login way, but I doubt

[tawdry bloom]

Ahh, oki. I was wondering since my school machine doesn’t allow my home google on the network lol

[tawdry bloom]

it was linked to a Microsoft account this whole time, i have been wasting my time for days.

[marsh steeple]

Guys where can I select the Plugins in Alpha version. I'm not getting any option to select a specific plugin after the new update.

[vagrant plover]

Next to the model selector when using the Plugins model

[winter glacier]

why it capped me for about 1h30mins, I only sent about less than 30 pushes early in the morning

[brisk garnet]

how do I get in touch with a human customer support representative? or they don't exist and they let customers figure things out by themselves in the discord server?

[timber shadow]

You can email [e-mail], but their response time is long right now I've read. What issue are you having?

[brisk garnet]

Are you official openAI support?

[timber shadow]

I am not, OpenAI staff will appear in yellow text with a in their name. They are periodically present in #1070006915414900886, you could try posting there.

[brisk garnet]

ah ok so basically I can send an email to [e-mail] and larp on this discord server. Ok cool, good to know

[timber shadow]

Hello, in case it's relevant to you, these are all the things that count as one GPT-4 message:

• Sending a GPT-4 message
• Editing/resubmitting a GPT-4 message
• Asking for a regeneration of a GPT-4 answer
• Asking for a continuation of a GPT-4 answer
• "Create" chat in GPT builder
• "Preview" chat in GPT builder
  The current GPT-4 cap is 40 per 3hrs. Custom GPTs were last reported to be capped at 25 per 3hrs of that 40.

[weak ruin]

can you fix this

[timber shadow]

We do not currently have seed control I don't believe, here is a bit of confirmation/clarity from OpenAI dev Moxi #images-discussions message
Basically, because they're still working on the "back end" of the model, seed control is not something that is reliable as a user tool as of now. In other words, if you use same seed + same image prompt at two different times, you may get different results because of the current beta nature of DALL·E 3

[weak ruin]

okay sad thank you

[timber shadow]

You're welcome, I bet it'll come back someday down the line!

[weak ruin]

I hope

[cosmic dune]

is there a way to make gpt have context about the structure of something or information about something and have it use this context in the analysis?
for example, let's say I wanna create a custom gpt about the Unreal Engine, and to make sure it does everything right, I download the entire doc and add it as a file, would it be able to read from that complete file and use that to base its responses?

[weak ruin]

yes

that what file upload for

[cosmic dune]

Does it have a limit for much context it can have at a time? because I tried to upload a partial file in the custom gpt containing only a small amount of information (it still was relatively large in size with more than 5k lines) and in the prompt I told it to use that file to base its responses, however from my tests it has been unable to use the entire file as context, it seems to try to read the file per response and it stops reading at around 1/3 of the file, being unable to give a proper analysis.

[weak ruin]

im not sure

I also tried to use some context but the output was not what I expected

[timber shadow]

There are definite context limits on knowledge file "imports." Really they're designed to be reference lookup files for the model to search through for relevant information, not to import it all as context. I'm not sure what the hard limits are on this, as far as how much you could reliably get in as context, but I know that just in general, it looks to knowledge files for relevant info more than it looks to them for comprehensive context

[weak ruin]

this is trash

[timber shadow]

Sounds like a hallucination

[weak ruin]

yeah it says

[cosmic dune]

I see... perhaps I need to make the info smaller, simpler, more searchable. I was trying to use it to verify if it could help me find which tables on a database are probably carrying customer data, because for some reason the database has more than 1.5k tables and I was given the task to import the customer data from it to another system.

[timber shadow]

Yeah it sounds like it's trying to hammer home some point like "My browsing tool is best for looking up text-based info like news stories, and can't really be used to look up 'pure data' like times or stocks" but it just sounds a little confusing in this phrasing

[timber shadow]

Working in smaller chunks when you run into issues like you've described is a great idea, you'll almost definitely have better luck that way

[weak ruin]

it's actually funny how ai is doing

[marsh steeple]

I don't get this option.

[timber shadow]

I'm wondering: are you a plugin dev, and not a Plus subscriber? If so, you might check this bug report: #1175444436256309419 message

[marsh steeple]

I got access as a plugin dev.

[timber shadow]

Ah yeah, you are not alone in that experience then, you might add your experience to that linked bug report and see if that plugin dev has gotten any updates on their end

[humble vale]

ChatGPT python environment having issue

[spark summit]

Custom GPTs: I don't see Edit options anymore for the GPTs already created. Have anyone noticed ? Thanks

[trail cairn]

plugins are getting removed and they stopped taking submissions over 2 weeks ago

[rugged wing]

you can port your plugins to GPT's quite simply

[feral jasper]

OCR doesn’t work

[vital radish]

Ask it for the "gen_id"

[timber shadow]

That was another thing Moxi described as a "implementation detail" and not a "feature," but it can be a helpful reference tool within a single chat!

[vital radish]

suggesting that becuase "seed" doesnt work anymore - gen_id does like you say at least give you a reference to the images it creates, and based on those - you can generate new images with slight variations.

[vague pagoda]

its working for you chatgpt?

[timber shadow]

Seems to be working as expected for me, are you having some issue in particular?

[spark mango]

Hey guys i think i have a very simple question but i'm unable to formulate it well enough to get google to give me good info. Is it possible to submit a prompt through the API and then not have to send the whole prompt over every time you make a request?

[vital radish]

working for you ?

[weak ruin]

I need the same image but I dont get it

[vital radish]

you wont get exactly the same image

[weak ruin]

sad

[cold folio]

wdym widescreen?

[vital radish]

they have the same "theme", to get exactly the same image you need the seed i think, which dall-e does not provide anymore

[cold folio]

you can just upscale it with a tool

[weak ruin]

I got the seed but it cant use it to generate same image

[weak ruin]

I mean widescreen

and then upscale

[vital radish]

that is widescreen though ?

[weak ruin]

yes

[vital radish]

also dall-e doesnt upscale

[weak ruin]

I know I have a tool

[vital radish]

ok, maybe im not understanding what you need to do ?

[cold folio]

there are some cool websites out there

[weak ruin]

I want this in widescreen

[cold folio]

well just make it 3840 × 2160

[weak ruin]

you cant upscale this

I need it widescreen like 16:9

[cold folio]

1sec

[vital radish]

i had to crop it a bit

[weak ruin]

how

[vital radish]

photoshop

dall-e cannot do this natively

you should have asked it to generate the original image in wide format

[weak ruin]

it says it cant

[vital radish]

are you using gpt4 ? or Bing ?

[cold folio]

dalle can't its limited to 17xx by 17xx something

[timber shadow]

1024 x 1024
1792 x 1024
I think. With the latter flippable tall/wide

[vital radish]

yeah - something like that

[weak ruin]

gpt4

[vital radish]

gpt4 can do wide - but you need to ask for it in the prompt

"create an image of blah blah blah in a wide format"

[weak ruin]

yes but last time it worked but they removed the seed thing

I asked make this wide screen

and it worked because of the same seed

[vital radish]

yeah but you asked it AFTER it made it 1:1

[weak ruin]

yes now it's not working anymore as last time did

[vital radish]

once its in 1:1 - you cannot ask it to make it widescreen afterwards

[weak ruin]

last time it worked perfect

since they updated it not

[vital radish]

then i dunno

its never worked for me ( at least not to get the EXACT same image )

[cold folio]

i can't help rn i'm on cellphone. but there are many tricks you can just put some bars on both sides and clone the background

i used to do this in 2014 lol

[weak ruin]

mhh

[vital radish]

if you have something like A1111 or InvokeAI - you can do Image to Image - but leave the weight to minimum so it keeps the original image - just creates a wide image of it and fills in the blanks

but the result isnt always desireable

[hollow veldt]

Hey everyone, I am trying to create embeddings for my dataset following : https://github.com/openai/openai-cookbook/blob/main/examples/Get_embeddings_from_dataset.ipynb

I managed to sort out most issues; my code is running.. but it never ends.... I've got about 1000 rows of data. Content is german text. Some messages are long-ish.

not sure how to paste the notebook though... a bit too long to fit in a message.
Unless git repos are an option

btw I cant afford to re-run this code... it kind of used a lot of tokens just 1 one go

[spring arrow]

Does ChatGPT provide a "privatized database" option? I would like to use it for confidential data but would like the peace of miind knowing its being transmitted anywhere else?