community-help

[rustic marlin]

it does not have vision, you can't upload files and it will not use tools or browse the web

for now

[formal brook]

seems like a rather random setting as its not done per topic just on an arbritary time

[rustic marlin]

it is a planned feature that we will probably see when it is out of the review phase

[formal brook]

Oh thats good news then so back to the original GPT I guess

[dusky topaz]

Hi @rustic marlin , I have coding background. I'm using Unity - C# for now.
Trying to make "Wordscape" game. We added all implamentations with "Code Co-Pilot" custom AI.
When I'm trying to write a prompt to AI, it's sometimes messing up all codes.
I'm just trying to create this "Relationship Alg." If a word has a similiar letter with any words bla bla bla. but i couldn't finish it. AI is stuck on previous code improvements.

[rustic marlin]

no idea, I've never used the mac version ¯_(ツ)_/¯

[rustic marlin]

I see, are you using o1-preview?
I have great degree of success getting it to refactor code, Im using that on my job, right now xD

maybe you could try to just start a fresh convo, it can be benefitial

[dusky topaz]

I've never heard that before. What's it?

[rustic marlin]

o1-preview is the model which does the pondering before it does the task you asked

are you a subscriber or is your account free?

[dusky topaz]

I apologize, is it generating prompts or cloning projects?

I have plus membership

[rustic marlin]

im not sure of what you mean

[formal brook]

fair fair fair

[dusky topaz]

I mean how it works? "I want to do this but don't know how to enter prompt, please tell me" or "I want to do this, how can i do please tell me."

I'm sorry cant explain myself. My brain is melting atm

[rustic marlin]

this sounds a lot like something one of the devs would implement to facilitate something on their side and then, just let that there for whoever might want to use it xD

so many of the features I implement on the stuff I make comes up like that too lmao

[rustic marlin]

check this: https://youtu.be/7GaKl6HwZ50

[dusky topaz]

Oh thanks ❤️

[rustic marlin]

you are welcome =)

[dusky topaz]

o1 did well but still not perfect 😄 . It couldn't create perfect shape for word placement.

[fringe elk]

Hi there! Sorry if this was posted, but in regards to scripting and coding with the canvas, do we know how large a canvas can be? Like can it only do 100 lines at a time? or has that been improved?

[wispy seal]

Am I dumb or do I not have the canvas feature (I am a plus sub)

[timber shadow]

It's still rolling out! Hasn't gotten to all of us yet

[dusky vale]

Will the canvas feature be on android?

[charred parcel]

no, it's only on the web version

Canvas is only for the web version, it's not in the mobile apps or the macOS desktop app

[tepid ocean]

It seems that custom gpt's are basically unable to process .md files. It read the .txt fiels fine. Any reason why/ SHould I just use .txt and forget about it?

[keen musk]

Hey,
Is there a way to be FERPA compliant using OpenAI api? Or at least not have a click through agreement?

[dusky vale]

ok i got it on web. Too bad it doesn't have a python interpreter

would be nice to have pygame built in lol

[split reef]

Will 4o with canvas be available in the macos app? It isn't yet.

And on the mobile apps of course

[charred parcel]

it appears not

[split reef]

I was finally switched to the app, and now I have a new reason to go back to the browser 😭

I started to prefer the app but yeah this sucks

[charred parcel]

I prefer the macOS app

[split reef]

Yeah same now

Unfortunate that it isn't available there

At least right now

[runic ravine]

what features does the app have that the browser doesnt?

never used the app

[wicked nova]

Anyone getting these on canvas on everything now?

Scaling issue?

In the mean time i'll just manually copy pasta, just wondering

[junior warren]

Theoretically no, but realistically the .md files possess a wealth of formatting information that could be relevant for a given situation.

I would think that you are specific use case would be more the governing situation in terms of whether or not simply converting it to text ahead of time is a viable option.

[sleek hill]

bring speech to text to windows

not tryna use co pilot cant even do it properly

[naive moth]

unable to publish gpt attachment
Error saving GPT
Unable to publish GPT
I was uploading a simple 0.5 Mb txt file?

[tacit stratus]

Are there any tools/extensions to use ChatGPT within Visual Studio? I'm not looking for VS Code tools

[naive moth]

Error: Failed to parse error response
    at yr.fetch (gghzs582a65dvm3m.js:284:21790)
    at async Object.mutationFn (ke4wseue2s3xar0a.js:1:11542)Caused by: SyntaxError: Unexpected token '<', "<html>
  <"... is not valid JSON
    at JSON.parse (<anonymous>)
    at yr.fetch (gghzs582a65dvm3m.js:284:21659)
    at async Object.mutationFn (ke4wseue2s3xar0a.js:1:11542) Objectstatus: 504text: "<html>\n  <head>\n    <meta name=\"viewport\" content=\"width=device-width, initial-scale=1\" />\n    <style global>body{font-family:Arial,Helvetica,sans-serif}.container{align-items:center;display:flex;flex-direction:column;gap:2rem;height:100%;justify-content:center;width:100%}@keyframes enlarge-appear{0%{opacity:0;transform:scale(75%) rotate(-90deg)}to{opacity:1;transform:scale(100%) rotate(0deg)}}.logo{color:#8e8ea0}.scale-appear{animation:enlarge-appear .4s ease-out}@media (min-width:768px){.scale-appear{height:48px;width:48px}}.data:empty{display:none}.data{border-radius:5px;color:#8e8ea0;max-width:420px;text-align:center;width:100%}.data ul{list-style:none;padding:0}@media (prefers-color-scheme:dark){body{background-color:#343541}.logo{color:#acacbe}}</style>\n  \n\n</head>\n  <body>\n    <div class=\"container\">\n      <div class=\"logo\">\n        <svg\n          width=\"41\"\n          height=\"41\"\n          viewBox=\"0 0 41 41\"\n          fill=\"none\"\n          xmlns=\"http://www.w3.org/2000/svg\"\n   "url: "https://chatgpt.com/backend-api/gizmos/g-0851cwCbE/promote"[[Prototype]]: Object
qo.<computed> @ gghzs582a65dvm3m.js:234
gghzs582a65dvm3m.js:234 FatalServerError: Something went wrong. If this issue persists please contact us through our help center at help.openai.com.
    at F (in2uijrjq22x1vxf.js:35:713)
    at async promoteGizmo (in2uijrjq22x1vxf.js:35:1385)
    at async onClick (in2uijrjq22x1vxf.js:5:39651) undefined
qo.<computed> @ gghzs582a65dvm3m.js:234

[small locust]

I deleted all chats by a mistake. Is there any way to undo that?

[trail cairn]

no

[small locust]

Since there might be other idiots like me out there that actually can do something as stupid as this..... I was actually able to restore my data!!! I did an data export right after I delete all the chats.... And the data was there🚀

[trail cairn]

did you use archive or the delete before that?

[small locust]

By mistake I just clicked "delete all messages". No archive. Nothing to be found in the app. I read up on it online, and it says it is impossible to restore. They recommended doing "export of important data". I just tried out the feature. Roughly an hour after I deleted everything. They sent me a link to my data, and a zip file to download. When I opened it, all my data was there.

[fresh totem]

if the transcription error happens in free plan does it mean too long_ can i pay to get it working or just start a new conversation

[rustic garnet]

Hay

[tepid ocean]

The governing sitaution? The governable situation? Sorry, I didnt understand. But just to clarify GPT is literally blind to .md files. As in, it doesn't even know I uploaded any files. It DOES know that files exist if they are .txt files though.
Thanks for the feedback. : )

[junior warren]

That's a different situation than what I have worked with as I have pushed MD files into the system several times for analysis and feedback. I'm not sure what to tell you about that.

To the context of my question, Israel free more to usage of markdown as part of something like a system row or part of the message that you are pushing into the api.

I typically read the files into memory and then build my API call from that, so my approach is probably going to be significantly different than the methodology you are using.

[tepid ocean]

I made a simple custom GPT and asked it do you see any files? When they're md it says no. When they're txt it says yes. I repeated this experiment. Interesting that it's working for you though, that's good! : )
You say reading files into memory but isnt tha tequivalent to uploading them either through the convo or through custom gpt configuration?

[junior warren]

I explicitly load the file using Python and create a prompt around that so that the data gets pushed directly through the prompt. I do not let the AI directly access the file. I suspect if I did, I would probably run into the same situations you are.

I've tried using the upload feature but I didn't particularly care for the way it worked from my particular situation as I wanted to do some level of preprocessing to the data before I handed it over to the AI. That has a significant difference in context of what I'm doing versus what you're doing with the actual raw file itself. And that's probably why I'm not encountering the issues you are, simply because of my preprocessing stage.

[cobalt yarrow]

It's been 2 days now

[tepid ocean]

Alright. Thanks for sharing. : ) I saw no need to preprocess it in my case, but who knows.

[blazing rover]

Guys, is it just me with having issues with .webp files from image generation from ChatGPT

.webp files seem extremely hard to delete sometimes and I'm having issues with renaming images sometimes, my File Explorer just crashes, is there any fix to that?

[hard nexus]

https://cdn.discordapp.com/attachments/1058345511725310054/1291271848918519838/ScreenRecording_10-03-2024_00-32-25_1.mov?ex=6700cf9f&is=66ff7e1f&hm=3c913b0975cdb76fcbd9cab0f4e575d072a3452485c3e52a01400f22634f472d&

20 free gigabytes, tried updating and reinstalling

[trail cairn]

did it work before?

[quartz spruce]

im trying to build a GPT is the most basic easy to understand way as possible. I work with the terminal alot for scripting automation apps like sorting videos by metadata frame rate ect but i do not know alot about coding and as ive seen it is very time consuming even with prompts. do any of you my friends know of any good pdf's i can drop into chatgtp it will have the knowledge of zsh commands to better help me write code? any websites free knowledge, anything. i use the text extractor from chatgtp and i just wanted to know if any of you fellas have any good websites or links to pdf's i could use. thank you my friends.

[grim fox]

Hey, I accidentally edited the wrong Assistant, deleting all the instructions, is there a way to restore the assistant to a previous save? (not sure if this is the correct channel)

[trail cairn]

if you mean api then there is no restore feature, however if you ever made a run with any threads with the older version you can see the instructions from there

[grim fox]

I have but I dont have old thread ids.. and I didn't find a way to list old threads either

[trail cairn]

https://platform.openai.com/threads/

[grim fox]

thanks! you are a life saver!

[primal osprey]

Hi guys, reaching out to the developer community for some help and advice! I am working on an AI-driven project, in which I upload a json file containing some data, and I essentially ask the AI to enrich the data in the json file and then output the updated json data as a file so I can retrieve it in python using the latest_message.attachments[0].file_id method. I tried to do quite a lot of prompt engineering and got it to work somewhat reliably. The problem is that a lot of times the AI can bug out and output some malformed json data (which breaks my code), or just output it in text format, which also breaks my code as I only download the outputted file. My question is if there is any way so that I can have the model reliably output the JSON data as I tell it to. I saw there is a way to have it output json data but I'm not sure if this is what I'm looking for? Maybe I just need to do more prompt engineering? As of now I am using the assistants API. I tried to fine-tune my own but fine-tuned models don't allow file upload... Any advice? Cheers and thanks.

TLDR; I want to have the AI reliably manipulate and output JSON data as tasked.

[vagrant plover]

https://platform.openai.com/docs/guides/structured-outputs/structured-outputs

This should help, it ensures outputs exactly match JSON schemas provided by devs

Consider heading to #dev-chat / #1037561751362863144 if you need any specific assistance with regards to this - should get better advice in a channel more focused on development

[primal osprey]

@vagrant plover thank you so much 🙏

I will read up on that, mind if i reach out again in case I need help?

[vagrant plover]

Feel free to, here's the blog post of it's release by the way

https://openai.com/index/introducing-structured-outputs-in-the-api/

There's a really good graph comparing structured outputs mode to previous methods like prompting or the old "json mode"

[frosty rover]

does the chatgpt plus subscription have an upload limit on images?

[primal osprey]

The article is amazing, but I have one doubt. I was just following along with it, and tried one of the example, and while it worked, the output doesn't seem to be in json format. It looks to be something like this:

steps=[Step(explanation='We are given the equation \\(8x + 7 = -23\\). Our goal is to solve for \\(x\\), which means we want to isolate \\(x\\) on one side of the equation. The first step is to eliminate the constant term on the left side of the equation.', output='8x + 7 = -23'), Step(explanation='Subtract \\(7\\) from both sides of the equation to get rid of the \\(+7\\) on the left side.', output='8x = -23 - 7'), Step(explanation='Calculate \\(-23 - 7\\), which simplifies to \\(-30\\).', output='8x = -30'), Step(explanation='To isolate \\(x\\), divide both sides of the equation by \\(8\\).', output='x = \\frac{-30}{8}'), Step(explanation='Simplify \\(\\frac{-30}{8}\\) by finding the greatest common divisor of \\(30\\) and \\(8\\), which is \\(2\\).', output='x = \\frac{-15}{4}'), Step(explanation='The fraction \\(\\frac{-15}{4}\\) is in its simplest form, so this is our final solution.', output='x = -\\frac{15}{4}')] final_answer='x = -\\frac{15}{4}'

This is from doing a copy/paste on the example code.

[vagrant plover]

Just wanted to confirm a couple things

• are structured outputs being used right now (response_format: {type: "json_schema", ...)
• what model are you using?
• is strict set to True?

[primal osprey]

I am using the SDK objects, they seemed to be more fitting because from my understanding of the docs I can show the general structure and it will follow it. In my circumstance, I pass in some pre-written json data and the goal is to have the model populate some of the data (indicated in the prompt), and there is an array of objects which I would like the AI to add to as well, that's why I found the SDK approach more fitting, but I can try the manual schema as well.

This is my current setup:

class Step(BaseModel):
    explanation: str
    output: str

class MathReasoning(BaseModel):
    steps: list[Step]
    final_answer: str

completion = client.beta.chat.completions.parse(
    model="gpt-4o-2024-08-06",
    messages=[
        {"role": "system", "content": "You are a helpful math tutor. Guide the user through the solution step by step."},
        {"role": "user", "content": "how can I solve 8x + 7 = -23"}
    ],
    response_format=MathReasoning,
)

This is similar to my use-case, I would like the model to determine how many "step" objects to create inside of the array on it's own. Is this achieveable also in the manual schema?

[primal osprey]

@vagrant plover , so I got it to work using the json_schema approach, but there is a little bug in the documentation. When using openai.chat.completions.create it doesn't recognise the response_format parameter, so I had to use the openai.beta.chat.completions.parse method. With that being said, is there any way I can use this approach with an assistant I created?

[torn bone]

it seems canvas with gpt4o dosen't support csharp

[vagrant plover]

Sorry for the delay! Yes, it should be possible (I've done so before)

There's a well written discussion about this on the Developer forum (unfortunately, the docs are kind of limited here)

https://community.openai.com/t/structured-outputs-with-assistants/900658?u=kesku

I'd recommend going to the forum if you need any help with more technical aspects of the API, it's a community entirely focused on devs like us haha

Something like this should work

client.beta.threads.runs.create(
    thread_id=thread.id,
    assistant_id=assistant.id,
    response_format={
        "type": "json_schema",
        "json_schema": {
            "name": "test_schema",
            "schema": Messages.model_json_schema()

[primal osprey]

Understood. And is there any way I can attach files to code_interpreter or file_search at all?

From my understanding, if I use the completions API, I cannot upload files. That is why I'm trying to use the assistants API approach. But I don't know if it will work. Do assistants even work with this strict json output format?

[trail cairn]

none of the api's actually support files like that. assistants only picks relevant parts of the file and puts them in context and you can do that yourself with completion too

[primal osprey]

But with assistants I had the option to do this:

dataFile = client.files.create(
    file=open('/output/data.json', "rb"),
    purpose='assistants'
)

attachments = [
        {
            "file_id": dataFile.id,
            "tools": [{"type": "code_interpreter"}]
        }
]

thread = client.beta.threads.create(
        messages=[
            {
                "role": "user",
                "content": message_prompt,
                "attachments": attachments
            }
        ]
)

How can I do this with the completions?

[trail cairn]

that just launches the code interpreter that is just a sandboxed python environment. do you actually need it as its not the same as llm doing the work. llm only orchestrates it

[primal osprey]

I just need some way of allowing the user to upload a file to the assistant/completions/model because in my prompt it tells the model to reference the uploaded file since it has some key information.

[trail cairn]

if you use the code interpreter to alter the file then its no longer the model doing the work and structured output wont affect it the same way

[primal osprey]

I only use code interpreter to upload the original copy of the data.json file, which is a half-completed, and my goal is to pass it to the model, so then the model can populate and enrich some of it's parameters with data given a specific prompt on what to do exactly. Am I taking the wrong approach here? By the way thanks for helping out.

[trail cairn]

have you tried to just pass the json to the model with instructions and use response format or do you need code interpreter for something like math?

[primal osprey]

I just want to upload it as a file, because if the json file inputted is too large for the text input then I don't want that to be a problem.

So code_interpreter isn't really necessary, even file_upload would work but as long as it can receive file upload input.

[trail cairn]

you cant go beyond model context window limit with any api features. some of them just do the splitting for you

[primal osprey]

Aha, so you suggest I still resort to pasting it in?

[trail cairn]

that way you can control what is passed to it and split it better. this is how file search does it https://platform.openai.com/docs/assistants/tools/file-search/how-it-works

[primal osprey]

Also, when trying to use this json format in my assistant, it doesn't let me specify the response_format. This is how I set it up:

thread = client.beta.threads.create(
    # model="gpt-4o-2024-08-06",
    messages=[
        {"role": "system", "content": "System Prompt."},
        {
            "role": "user",
            "content": "User Prompt.",
            "attachments": attachments
        },
    ],
    response_format={
        "type": "json_schema",
        "json_schema": {}
    }
)

run = client.beta.threads.runs.create(thread_id=thread.id, assistant_id=ASSISTANT_ID)

And for some reason it returns this error:

``TypeError: create() got an unexpected keyword argument 'response_format'`

@trail cairn Maybe I am calling the wrong method?

[trail cairn]

response format is a run property

[primal osprey]

So I append that to the run?

Ok, well I got that working, but unfortunately, as @vagrant plover pointed out by linking that article, structured json data output doesn't support tools like code_interpreter or file_upload!

{'error': {'message': 'Invalid tools: all tools must be of type `function` when `response_format` is of type `json_schema`.', 'type': 'invalid_request_error', 'param': 'response_format', 'code': None}}

Why not?

Does anyone have any ideas on how I can work around this? Maybe I just resort to prompt engineering to convinve the AI assistant to produce a reliable JSON data output? No clue. Any feedback here would be greatly appreciated.

[tropic trellis]

hello, im a plus user but dont have access to advanced voice mode because i live in Europe any solution ?

[trail cairn]

response mode is a model feature that its been trained or fine-tuned for and tools work outside the model. the solution is still the same that you provide the data as prompt and use the response format as no other solution will guarantee valid json

[trail cairn]

not available https://help.openai.com/en/articles/8400625-voice-mode-faq

[pine flame]

what the heck is this it only happens when using canvas

[trail cairn]

just acknowledge it that it comes a slight risk of account suspension

[pine flame]

what ?

[tropic trellis]

u getting ban buddy

[pine flame]

why ??

someone care to explain ??

im both confused and concerned as this costs a lot where im from

[trail cairn]

its against terms of service

[pine flame]

what is

[trail cairn]

evading set geo restrictions

[pine flame]

hu?

what geo restrictions ?

im just using GPT like every day

[trail cairn]

the guy who used vpn

he deleted the messages

[tropic trellis]

🤣 🤣 🤣 🤣 🤣

[pine flame]

oh I thought you were speaking to me

[pine flame]

not funny mate

[tropic trellis]

stop stressing out

[pine flame]

it costs one third of what I make a month to run GPT mate, i will stress out

[tropic trellis]

u make 60$ a month ?

[pine flame]

p much

millitary service

[trail cairn]

check your browser console for errors or make sure none of your extensions mess with chatgpt

[pine flame]

I have no extentions

[tropic trellis]

@trail cairn u got advanced voice mode in finland ?

[trail cairn]

with teams yes, not with plus

[slender island]

@trail cairn sup,do you think if i downgraded my huawei firmaware 12 to 10 and installed googlefier chat gpt could work ?

since the google ban huawei users are only rellying the web version and third party apps to install google apps

even with gpsace i cant open chat gpt and some other apps

what seemed to me to support eveything is googlefier

hellooooo

hiiii

[tropic trellis]

richard

change ur phone

[slender island]

i am going to buy a samsung a55 around christams since its cheap and has cool feautures than mine

[tropic trellis]

stop persisting w da wawei take a new phone its like 150$ for a very good phone

[slender island]

but you know i am not the only one using huawei devices in 2024

[tropic trellis]

maybe but stop using it

[vagrant plover]

ChatGPT uses the Play Integrity API (SafetyNet), if your firmware isn't going to work well with Google services (and things like banking apps) then things aren't going to work out.

That said, ChatGPT works on Android 6.0 and higher

Downgrading firmware wont make ChatGPT work if it doesn't work on your current latest firmware, you should consider finding a device which would work

Ideally, you'd have an Android-powered device with Google Play services that passes system integrity checks and meets Android compatibility requirements

The Samsung A55 you mentioned would be perfect

[open cave]

Any help with chatgpt for chess?

I think it is limited as it can’t draw board on fly.

[onyx kraken]

Hey, I'm not much for chess. But here's a potential solution;

I used o1-mini to write Python code for the board and some movement logic (to help prevent 4o from accidentally cheating, to also help 4o keep track of the pieces).

Then I had 4o put it into its Python environment.

I'm not much of a chess player, want to test this out?

You might need to copy the code and paste it into your own 4o new chat, asking it to put it into its Python tool to play with you.

Be aware Python tool times out after several minutes, this is not a game where you can go away for a few hours then resume easily.

That can be done, but the model needs to re-input the code, typically twice (the first attempt just gets the Python tool active again, the second actually gets the code into it again) in order to use it again, and the game would have to start over.

Let me know how this works for you! For sure, the board is drawn and there's logic to handle how pieces move. https://chatgpt.com/share/67019458-cf84-8011-b0a8-641d5d555819

I did have to ask it:

Is this ready for us to play?

We will input them manually. You'll decide your own move, I'll tell you mine.

And then it suggested how I would tell it a move, such as e2e4, which I used.

[plush pewter]

Is there a way to do a brutally extreme and extensive brainstorming? With unexpected connections, capturing every angle of a topic, simulating, perhaps, the way the human brain organizes information. Reference image

[onyx kraken]

Every angle of a topic requires a lot of space, maybe 1000s or more of pages of text.

That's more than one output can do.

However, the model can brainstorm for or with you, based on your ability to guide it to what you want to explore.

[plush pewter]

Today, I brainstormed methods to enhance cognitive performance and capacity using it. However, after reaching the 113th idea, the model became redundant and repetitive, unable to explore new angles or approach the problem with lateral thinking

[slender iris]

Hey guys :) I can see here that the limit is 65k, but in my training file it says 550k, does it mean that it didnt train entirely?

or is it 65k tokens per example, and since i gave like 50 examples then i am chill?

[cursive steeple]

I paid for the subscription and it was working perfectly fine for a few days and now the last two days it's becoming very laggy and slow enough to where my browser will ask if I want to quit the page or wait.

I've cleared the cache and cookies, changed internet connections, changed browsers. It definitely works better on my phone but I really need to be able to use it on my laptop. I'm not sure what else I can do. I don't really like paying for something that I can't utilize like I expected to and was for a bit.

Anyone else have this issue? How can I fix it?

[onyx kraken]

You might want to consider if a browser extension could be part of it.

If you want help, use this https://help.openai.com/ and bot bottom right. Select messages then click through the choices that closest match your situation, then the bug report option.

Initially a bot will talk with you, then in a few hours to a few days a human will answer; keep checking back at that page and the conversation to see if the answer's happened yet.

Ideally you get an email saying they've answered, for a time that email wasn't happening

[cursive steeple]

Thank you! I'll give that a go.

[lost tinsel]

I keep get redirect to the login page when i log in using log in with google. Tried clear browser cache and data, tried use another browser.

[final olive]

Did you create the openai account using an email address instead of google? Can't switch.

[lost tinsel]

yeah

[clear trench]

I had a plus account using my work email (Gmail based). I created a team account. I added members but for some reason one of them can't see the team features. He had a free account using the same address as the one I invited him. Ideas?

[charred parcel]

contact OpenAI support at https://help.openai.com

[burnt elk]

i can no longer send anything in any chat, is the 50 message limit still exists for plus users?

[trail cairn]

disable unofficial extensions

[burnt elk]

are you talking about Grammarly and such?
because i have been using it since the begining and it has never caused any issues for me before.

[trail cairn]

you can easily try it. if it works without it then that's the reason

[burnt elk]

nope. still not working

[trail cairn]

try in incognito

[burnt elk]

not working in incognito either. is it because of the cap limit?
because if it is shouldnt i be getting a notification?

[trail cairn]

yes, unless if something blocked the notification

does 4o mini work?

[burnt elk]

nope. no model is working

[trail cairn]

something is messing it up then, maybe browser console showing errors

[burnt elk]

let me restart the browser and see if that fixes it or change the browser altogether

nope. did not help at all.

[hushed gull]

Hey guys, I'm having issues with my ChatGPT android app. It doesn't matter what I type the message responder just stays in thinking mode and doesn't give any kind of response!

I have tried uninstall and reinstall the app but it doesn't fix the issue.

I'm on a Samsung S24 Ultra and didn't have any issues with the app on my S22. My coworker has the same app on her iPhone 14 and no trouble with it at all.

I have the paid version so not sure why it's not working.

Any suggestions?

[short chasm]

just a few days ago I was able to send links and it would easily look up the website, now this suddenly is not possible apparently? how does it know the weather, which is real-time data, but it can't access live websites? Could some of you guys maybe test the same request?

*I pay for the plus subscription and this was using GPT-4o

[onyx kraken]

So, it can search the web, like it did to find a webpage that will talk to it and tell it weather data (not all pages will do so)

But it cannot always check a specific page, for example that page may tell it through its robots.txt page that it does not want to talk to ChatGPT.

The model's designed to SEARCH, not READ A PAGE YOU CHOOSE, for various reasons.

A method you can use to get that information summarized is for you visit the page and either copy/paste from it, or screenshot it and paste the screenshot(s) in and ask the model to summarize what you have pasted.

[short chasm]

okay yes, that makes sense - it's just very strange because it worked just a few days ago
to go into a bit more detail: I opened a new chat for "review packages" where I paste the link to a product on a shop and based on the product page, it wrote a list of customer reviews. This worked for weeks, but now if I request the exact same thing (for products on the same store - so the robots.txt should be the exact same) it says that it cannot search live webpages?

here 3 screenshots, one from a few days ago where it worked and then one from later on where it suddenly didn't work anymore (ignore the switch in languages) and also one where it apparently admits to this sudden unavailability?

[onyx kraken]

So, sometimes the model will pick a false answer that seems to satisfy the request.

This is often called 'AI hallucination'. It isn't based on 'facts'.

Often the url has clues. Even without visiting the web page, the model can know something about the content of the web page.

However, in this case, your conversation is actually likely hallucinating the other way.

I checked and without a web search, the model doesn't know what the Elowa Catchy is, with a search it does. You can see in the 'no search' answer it makes up a meaning for the term.

So, chances are in that conversation the model is imagining it can't search.

One answer is to go up earlier in the conversation and 'edit' your post that it first replied to with a 'can't search' answer, so it makes a new answer and doesn't see its refusal or claim to fail to search (once it sees that, it may continue the pattern).

It's clear that the website in question is not blocking ChatGPT.

[short chasm]

wow, thank you so much! so pretty much it could "lie" to itself and then just repeat that lie instead of trying again?

that's so interesting! I just did the editing trick you mentioned and now it can suddenly access the exact link it failed before

lol

[onyx kraken]

It can get confused. It picks how to answer, following its programming and training as best it can - but it isn't perfect.

Kinda like a human who might know an answer, but right then that time when asked, give a wrong answer.

Like, ever heard of someone messing up giving their phone number or address? Stuff they know (and maybe relearned a few times across their life), and every so often they just mess up and give a wrong answer, don't notice they did.

For the model, once the model sees an answer, it often tries to assume that answer's right and support it further.

So the 'trick' to go edit above the error means the model gets to answer again 'for the first time', and usually it won't mess up as long as it doesn't have a pattern of mistakes to continue.

[onyx kraken]

In some ways, I compare the model to a very, very, very neurodivergent human, and one around 5-7 years old as well, 'pre-operational'.

I dunno if you ever have long talks with kids in that age, especially ones who have differences in their abilities and difficulty with some ideas - it can be similar.

The model is guessing what truth is, it has no reality to double check. It has this huge amount of training data, full of exceptions and in some cases outright contradictions. It has some training to guide it towards 'preferred answers' between all the choices but it can't actually check facts. It tends... 4o especially... to believe what it is told, believe what it has already said, and to keep going in patterns.

[short chasm]

yes, definitely noticed that, do you think this trait can be improved in the future? or is it an intrinsic problem in the ways we train these models now?

[solemn totem]

Is there a decent app on Samsung s10+ that would read aloud the text response on the chatgpt app without having to copy and past the text , ie directly read it out if I tap on the text once finished responding or as it responds.

The inbuilt talk back is not the best. And doesn't allow to pres the text box

[onyx kraken]

I don't know enough about model design, and the future of what's possible that we haven't seen that, to know.

I can tell you that there's humans who have very similar responses. I've watched many humans across all ages give 'hallucinatory answers' similar to the models.

Saying stuff that simply isn't backed up by evidence, be it from what they decide is true through mistaken critical thinking, or through earnest belief, or through personal and atypical experience, or through conditions and situations that cause actual hallucinations.

But we as humans hold 'human hallucinations' to be true, at least some of us at least sometimes. We teach them too - for example, many young kids in the US believe in the tooth fairy, and in large part because it's taught in general cultural and also personally to them, even by the very adults creating the elaborate and fun 'lies' like creating and placing Easter eggs and other treats 'that the Easter bunny brought' for the kids to find.

So even if it's possible for the models to learn to not guess about truth, is it actually possible for us, with our love of fiction and fantasy, our own confusions and hallucinations, to teach them to 'not hallucinate'?

And if we did, would most of us like the other behaviors and answers that this would cause the models to also use with us?

[soft sedge]

I have a question so I am new to this ai stuff but when I used Hermes 3 405B (free) in open router chat and it started to forget after 20 messages... I am not sure but it shouldn't happen for a 128k context window model right? can you tell me what's wrong or maybe it's the settings... I don't understand.

[short chasm]

very deep thought but also very true...

[onyx kraken]

We kinda need to only discuss not-OpenAI-models in #ai-discussions , per #server-rules.

But some stuff's true for any model.

if there's confusing or especially conflicting information or really especially conflicting instructions, the model may behave very strangely at any point. This strangeness may appear to be forgetting context it should remember if everything was working as expected.

The conflicts can come from its programming, like rules how it handles safety, or it can come from your inputs "If A, do B and never do D." later "If A, do D and never do B." That kind of instruction can get really odd behavior when A is used.

[tough solar]

Please tell me how I won’t sleep until i figure out how to get my conversation back

[onyx kraken]

Right, what I've told you about is how; that's the person who said they saved a conversation that way. Good luck!

[drifting sable]

Hey, I have chatgpt plus but cant create any new canvas, is that normal? Did they deactivate that again?

[drifting sable]

Good to know.
I can even continue working on and adapting canvases of previous chats

[crisp whale]

I noticed something different lately. Not long ago when I clicked the copy button at the bottom of an answer, my clipboard was filled with raw markdown text. But now when I copy the answer, it is rich text in my clipboard. Does anyone notice that?

[cloud pecan]

I have noticed this as well

[lucid hazel]

Yikes! That sounds like a problem for sure.

[crisp whale]

I prefer copying the markdown since it gives me more control. I even collaborated with chatgpt to write a nodejs program to convert the markdown to html and copy the html to the clipboard. After that I'll paste the rich text to Evernote and it renders well. Now the rich text I directly get from the answer doesn't render well in Evernote. I'm thinking of a new way to sidestep the issue. Maybe deliberately ask chatgpt to output the markdown text.

[crisp whale]

The web version still copies the markdown. @cloud pecan Nothing is permanent though.

[manic shell]

trying to generate some images of maps of imaginary places, but i don't seem to be hitting the right keywords. all the resulting images aren't actually top-down ortographic views. they're all at a slight angle. any suggestions?

[rigid musk]

What was the prompt ?

[manic shell]

Tried a couple of things:

top orthographic plane view of a small map of a zoo with 5 naturally occurring enclosures in a forest environment, neighboured by mountainous terrain, a river splitting the zoo with a waterfall falling into it

a dnd map of a zoo in a mountainous, foresty area. a river splits the zoo into two.

From a 2D aerial perspective, visualize a vibrant, family-friendly tourist map of a zoo nestled within a lush, forested mountain landscape. Prominently illustrated is a majestic waterfall cascading into a river that bisects the zoo into two distinctive sectors. Render this in a classic cartographic style reminiscent of old-world city maps, complete with nature details, animal enclosures, and various amenities ingeniously dotted across.

Perfect 90-degree top-down view of a smaller zoo in a mountaineous, foresty area, on an island. A waterfall drops down a mountain and splits the zoo into two parts.

[rigid musk]

Try at the end

Top down, world map view, geographic map

[manic shell]

getting results like these:

[rigid musk]

Can you give an imagine of what you expect?

[manic shell]

more cartographic, such as this. the style doesn't really matter, the perspective does. I want to use it as a reference for modeling.

images like these are close but it's always at a slight angle instead of an actual ortographic projection

[weak knot]

why would chatgpt use safetynet anyway? that sounds supremely pointless
makes sense for banking. not for app that makes language model go brr

[rigid musk]

Try with battlemap style

[ashen dagger]

hi! I have a conversation that keeps "disappearing/appearing" from my tab, is this common?

[viscid wigeon]

We're truly tired of being payng customers and cotinuing to get the "Hmm... somthing seems to have gone wrong" in both mobile and desktop, and after trying every single possible fix suggested by OpenAI. Is anyone else having this problem?

Not only do we get that message, but GPT deletes our prompt and it's answer soon after sending that message of death 😦

[trail cairn]

what do you see in browser error console when that error happens?

[viscid wigeon]

We're using the desktop and mobile apps.

We were tempted to delete both apps and redownload, but the error still comes up on some of our chats on web based.

[fathom bolt]

I'm sorry if this has been asked before but where can I see my o1 usage in plus subscription? And days until it resets?

[modern cloak]

Thank you Openai! I received this message on the chatgpt.com website telling me to download the latest version of the app to enjoy the new voice mode, but when I check the app, I don't have the feature!

[regal lynx]

Do you live somewhere in the EU?

[modern cloak]

Yes, I live in France, but I shouldn't be receiving this kind of message, which is very frustrating.

[regal lynx]

I'm not sure of EU customers have access to advanced voice

[modern cloak]

Indeed, I can confirm this as I downloaded the latest version of the app from the App Store, and I don't have access to the new voice mode.

[trail cairn]

https://help.openai.com/en/articles/8400625-voice-mode-faq

[terse sparrow]

si on met un vpn aux us on peut l'avoir N?

[copper lodge]

Hey guys i need help:
Everytime I upload any time of file whether an image or pdf, it will upload about 2/3rds of the way and then stop and stay at that loading point. These aren’t large files, I’ve tried tiny screenshots that I just Ctrl+V into the chat and it does the same thing. This doesn’t happen on any other device I have (I have tried on mobile and another desktop). I have logged into my personal account and my company Team account and I am still getting the same issue. Both accounts are ChatGPT Plus accounts.

[final olive]

cross posted

[onyx kraken]

First is a medieval style map of your DND zoo prompt, the very short prompt.

Second is an atlas style of the same prompt, third is a topographical style of the same prompt.

Those three types of maps traditionally are always shown strictly top-down, never isometric, which may help ensure it's the view you want.

[modern cloak]

oui on peut l'avoir avec un VPN, le truc c'est que c'est pas cool de recevoir ce message pour au final ne rien avoir tu vois ?

[small locust]

Isnt canvas available in the Mac App?

[charred parcel]

no

[astral juniper]

hi i am not able to create images in chatgpt

is there any lock ? im a plus user

[small locust]

Have they said anything if it will be in the future?

[astral juniper]

It appears that image generation is still encountering issues. I'm sorry for the inconvenience.

this is the message i get from the bot

[charred parcel]

there has been no announcements about that

[astral juniper]

It seems Firefox is having trouble with chatgpt in generating images... chrome browser works fine... thanks

[worthy frost]

Hi guys, I want to use assistant API to make my project, there is 2 special abilitys that I want:

1_ online acsses to some websites (online shops), so it can guid the users to find the suitable product according to written information of the products and user needs.

2_ sending some shortcuts to website pages of suitabale products (buying page), the shortcut should be the picture of products and a little of information.

The question is, is this possible with OpenAI assistant API?

[kind sundial]

Got kicked out of my account and am continuously getting this error trying to log in:

"Something went wrong. Please make sure your device's date and time are set properly. Check that your internet connection is stable, then restart the app and try again."

Have tried, uninstalling/reinstalling, restarting phone. Nothing has worked so far.

[solemn totem]

In chatgpt android app using standard voice feature is there a way to extend the pause duration before it begins it locks me out of adding to what I day and starts replying/processing

[gleaming sparrow]

anyone constantly getting Unable to upload image.png when trying to paste screenshots

[final olive]

Make sure that the phones date, time, timezone match reality. Are you maybe currently outside of your usual timezone? I am thinking that this is part of the location checks needed because of the EU rules. Your IP address and device time might be used to check where you are, and if something does not make sense to the filters, it complains. Without that, most internet (general network) transmission require a reasonably accurate time to work properly.

[thin kite]

Hi fellas. Is it me, or is the "Code Review" funciton of canvas failing?

[undone mica]

In ChatGPT, do we have a global search feature to find something mentioned in a past chat? If not that would be a really great feature to have. Also, it would be great to have dates for the past chats, not just "past week" etc.

[charred parcel]

the macOS Desktop App has global search

[solemn bison]

Hello, Sirs. Using Laravel Openai, is it possible to apply structured output? Cause I can't find any reference.

[umbral jetty]

Good morning everyone,
I have a question about chat gpt canvas. I am a Plus user from Germany but I cannot select the Canvas version. Why is that?
Please help

[final olive]

AFAIK canvas is only available in the web interface. Are you using desktop or phone app? Canvas was rolled out in stages, but I think all Plus accounts should have access by now. I am not aware of any location based bias for the rollout.

[harsh kernel]

Hey guys, I have a small question about LangChain's embedDocuments func, which basically sends API calls to OpenAI to embed docs.
I am just wondering if the order of docs in input array preserved as in the output (returned) array, so I can attach metadata in the same order?

[stable crypt]

Hey

[pseudo oar]

how are the o1-models getting stuck non stop on long outputs. I dont get any of them to finish as they get stuck like 99% of the time. Anyone else?

[visual rivet]

Does anyone know how I can make Canvas work as intended? I have the "ChatGPT 4o with Canvas" model selected but none of the features appear, only the default chat layout

[trail cairn]

it does not activate until you do something that its good for

[brisk tide]

I think my app broke

[brisk tide]

Give it some code and if canvas doesn’t open well ask it to open it

[timber shadow]

I've had luck with kind of manually entering the canvas UI by saying somewhere in the query, "Use canvas to...". Like Radium said, it's set to engage when appropriate (they're still working on it - see "Training the model to become a collaborator" section of: https://openai.com/index/introducing-canvas/), but it also seems like you can just explicitly request the environment to trigger somewhere in your input.

[brisk tide]

Only available on desktop on the website and if you don’t have it it’s not because you are form Germany I have it too and I am form there too

[timber shadow]

Is this iOS or Android? Have you already tried closing the app all the way and reopening?

[brisk tide]

iOS and yes I tried but it doesn’t fix it but my app is weird tbh because it’s always jumps form new ui (with advanced voice and stuff) to old ui and back

[timber shadow]

Very odd! On Android myself so maybe someone on iOS can chime in too but - might be worth checking that you have the app updated, and/or maybe try uninstalling/reinstalling/rebooting/etc., if you haven't already tried those too!

[brisk tide]

Yeah didn’t help it’s weird but. Not a big deal just a visual bug

Another funny one I encountered recently

[slow rose]

anyone unable to get canvas to properly generate more than 200 ish lines of code?

[cold folio]

No issues here for me. how does it happen?

[slow rose]

I'll chat with it and it'll stop generating halfway trying to fix something at the bottom of the code file

[cold folio]

It's probably a bug. can you try asking it "Please use gpt4o with canvas mode " or use the link manually https://chatgpt.com/?model=gpt-4o-canmore

[slow rose]

it is in canvas mode, when it tries to re-edit it stops halfway, doesn't generate rest of the code

cuts out at line 190ish 200+ish

[charred breach]

When I attempt to ask ChatGPT to generate a PDF… it does it. When I try to use the api with openwebui- it tells me it is not capable of doing that. Am I just asking wrong ?

[dry rain]

now it's just making fun of me: on the web page it's telling me to download iOS but then on iOS i still don't have advanced voice 😦 Friends have it though...

[tiny mauve]

When o1 will be with no limits? Can't live without it anymore

[untold hare]

I am trying to find a good voice to text speech software

any recommendations?

[rustic marlin]

OAI's whisper is pretty good, but it is not quite real time, if you just want to process audio files it is very reliable

[mild bobcat]

Is it possible to open a Canvas on the MacOS Desktop app? I have managed to do it by using the web app and switching over to Mac after, but not direct within Mac. If it isn't possible, is it planned to be added in the future?

[neat quest]

I'm trying to share a chatgpt chat link via link with a friend. I shared another chat no problem before and it was fine. This one keeps leading to a 404. I deleted the link and made a new one but it's still 404ing. Is this a known issue or a me issue

[cold folio]

For now, only via browser. i gues it will be added to the app soon

[cold folio]

Some links with image wont work it's a known issue

[tepid ocean]

Does nobody have issues with uploading some file types to gpt? Markdown has never worked for me horribly for me. So now i kind of want to understand what works and doesn't. .py and .css files seem to be truggling too

[unreal agate]

I am trying to upload large number of tweets to openAI thread API in few batches. Sometime the batches are not merged together. Any idea why sometimes the batches are added and someitmes not? Any hint on best practice to handle that. ?

[supple steeple]

lmao this happened to me but it was actually recording. For whatever reason I kept the whisper voice thingy on did some transcription, but didn’t turn it off. Went downstairs to eat some lunch, did some other things and I come back to a nearly 3 hour voice recording

[north torrent]

yeah i had a few issues with it too, but i just press the restore previous button and ask it to try again

actually the main issue im getting is "failed to edit"

it tries to edit the code but it cant, and it undoes any of the latest changes i made

but in all fairness it is still in beta and basically fresh

i found out that there was something in my code that made it bug out. works fine with other code

[dreamy basin]

For the past 3 hours I've been trying to generate one image with chat gpt - coloring book hero, a young girl with long hair, and a long dress, it keeps creating duplicated images like this one

and because it is not doing the right thing, I am keep getting the message that I am generating too many images and banned for 20-30 minutes

now I get this , and after 3 hours, I still did not get 1 image with 1 little girl

this are the images I am geeting

[wintry wadi]

personally I am not happy with Dall-E. I run stable diffusion where you can download various models to suit your needs and get better results

may want to ask others in the Dall-E channels

[oak inlet]

Probably your case is same as or similar to the problem I am experiencing. I can't signup to ChatGPT and OpenAI using email addresses on my own company domain. OpenAI support does respond to my inquiries, but they haven't checked whether there is any domain issue on their side so far.

[forest sigil]

🔍 OpenAI PDF Upload Issues - Need Help! 📄

Hey everyone! I'm encountering some challenges with OpenAI's PDF upload feature and could use your expertise.

Issue:

• Text-based PDFs (e.g., simple text or printed documentation) upload without problems
• Image-based PDFs fail to upload
• This occurs in both the API and the web interface GUI

Details:

• Text-based PDF size: 9MB (uploads successfully)
• Image-based PDF size: 2.5MB (fails to upload)

Initially, I thought this was an API-specific problem, but I've confirmed the same issue occurs when using the web interface.

Questions:

1. Is the image-based nature of the PDF the root cause of the upload failure?
2. How can I analyze the PDFs to determine what OpenAI accepts?
3. What are OpenAI's specific requirements for PDF uploads?
4. Are there any known workarounds for uploading image-based PDFs?

Any insights, solutions, or experiences with similar issues would be greatly appreciated! If you need any more information, just let me know.

Thanks in advance for your help! 🙏

Btw, also tried this in Playgrounds.

[trail cairn]

https://help.openai.com/en/articles/8983779-are-you-able-to-handle-images-embedded-in-docs-presentations

[forest sigil]

Thank you. I have a document that was generated by a client image-based unfortunately. Any recommendations/suggestions any other tool that can perhaps read those?

[trail cairn]

there are several online services for that but cant recommend any specifically. you could convert them to images and use vision api for each one and then the combined text result as doc to gpt-4o

[forest sigil]

Vision eh? That's an approach... 😉 ty!

[sweet star]

why do i still not have advanced voice

[cyan matrix]

EU or non-US?

[crude vessel]

where can i get support, i have troubles with whisper

[fickle quarry]

1. i have credits
2. i have not crossed any usage limits
3. it was working fine a few hours ago
4. ??!?!??

[onyx kraken]

"Finish account setup" suggests that you may be logged out and thus not credited with any information or credits.

[fickle quarry]

Nope, am logged in

[onyx kraken]

https://help.openai.com/

Bot bottom right.

Select messages then pick through the choices to mark the closest choice to your situation, soon you can select 'report a bug'.

Initially you chat with a bot, within hours to days a human will respond. You may get an email when the person answers, otherwise you need to keep checking back to see if the chat has been replied to yet.

[fickle quarry]

Okay thank you so much :))

[robust night]

So, I've been encountering this error all day.

I have no idea what this is. I haven't changed how I use chatGPT.

Any idea what this is? Anyone else affected?

[final olive]

Variations of that seen before on web browser. Normally a page refresh lets you continue. I don't know what the equivalent of a refresh is for the app.

[cyan token]

Can someone help to integrate a qubit speaking model / game:

[elfin turtle]

Hello does anyone know why i cant see the pictures or why the load for like a Minute ?

[onyx kraken]

The 4o model can make the diagrams in the web interface. However it has essentially no spatial sense. o1-preview might be able to handle it, but we only get 50 messages per week with that, and I'm not personally ready to put however many messages needed into trying to design this with o1.

So if the goal is to 'create a game' with 4o's help, heck yes it can do it, but might need a lot of micromanagement to adjust the design of the diagrams correctly.

If the goal is to handle 4o's playing of the game as well, following rules and using a type of spatial sense (if you can define the rules fully with coded logic, and the rules are pretty simple, the model often can handle that), then I don't think 4o can handle that currently.

[ancient panther]

For context: I am an anti-Scientology activist. I have created a GPT preloaded with Scientology's doctrine relating to their counseling procedures (auditing) as well as two indexes (Document Title Index and Subject Index) to try and ensure accuracy with citations.
However, I am having trouble actually getting accuracy when it comes to giving the prompter citations for the requested information. If anyone has the time to talk on a call sometime, I can share the resources I have uploaded into the Knowledge Base to help get a better understanding

[onyx kraken]

What 4o can definitely do:

[onyx kraken]

You mentioned the issue with citations.

The model is extremely challenged to accurately cite anything; it does not observe 'pages' like we do, and it can struggle with exact quotes once the material gets 'very long' (like, longer than a short chat). It also does not easily remember long lists of specific details, and tends to 'forget' details it does not consider relevant so it can track details it thinks are relevant and have context available to track your new input that will come soon (which is has been trained is fascinating and very high value; it really wants to make sure it can track and work with what you're probably going to ask next).

Realistically, if you highly value the model giving accurate citations, yourself make a citations page that has areas of the documents commonly used concepts come from, and have the model check that page and quote that material.

The model can follow that simple instruction far easier and more reliably (it will still make mistakes with that - the way a person trying to answer their kid's question while trying to cook dinner while trying to shush a partially obedient dog that wants to bark at something and then the phone is ringing too - that phone may not get answered especially if the person struggles with phone conversations more than most people do, like the model struggles with citations) than trying to reason its way through trying to find the quotes and correctly cite them when it is essentially 'blind' to that level of detail.

[ancient panther]

The citations are given via the index text documents. I had to reformat the entire Index volume for the bot so that it would be easier to read as it had issues with reading columns from OCR.

[onyx kraken]

You can perhaps explore, maybe with a smaller, simplier mockup, of how to further support the AI with citation use so it can succeed.

But citations are a known very major weakness of the model, with many factors making it very hard to impossible for it.

[ancient panther]

I will try using one or two of the volumes instead of the whole set along with the citation document.

[onyx kraken]

Yeah. Sadly, more than about 3-5 citations and more than about 2-3 types of ideas, 'at once' is really going to strain the current 4o's capacity.

The complexity, the more the mixing and matching needs to be done, the vastly harder it is for the model.

Especially because to the largest extent it was trained on material it was not supposed to either perfectly quote or remember the exact source of. It definitely knows some quotes and can correctly attribute some commonly used ones from the training data, but that's not actually a citation, that's more a type of information that it recalls as it was given; citing is a different skill that it was kinda taught the opposite of, more synthesize plausible answers from many not-given sources.

[rapid spindle]

Hey can an experienced dev help me out?

I have a set of questions that I want the bot to ask users but it should only move to the next question once the user has provided a relevant answer.

If the user responds with gibberish or an unrelated answer, bot should politely ask the same question again. I also want the bot to communicate in a human like way.

What approach would you recommend to achieve this? Shall I go for custom agents or there's another better way of doing it? Thanks!

[sacred plinth]

#dev-chat message

[sacred plinth]

I need some help about this

anyone have some experience in this, please give me some advice

[tough solar]

Guy i believe we can get openai and ChatGPT to add a deletion box or some way to recover deleted conversations by fully identifying themselves.

never mess with DETERMINATION
if ever want something in life go beyond the physical see the intangible and use it, use it to achieve the IMPOSSIBLE

[tough solar]

I PETITION FOR A RECOVERY FEATURE
on the behalf of every member who wants one i will fight for them to add a recovery feature

[onyx kraken]

#1070006151938314300 That's the best place to make a suggestion; OAI probably doesn't read everything we say in the discussion threads

[tough solar]

Ive already made one

https://discord.com/channels/974519864045756446/1293207369194405938

[solar stream]

how can i create an api key for a specifc project and view and manage costs for that specific key
like
api key 1 - user puts in 10$ - user can only use 10$ of tokens
api key 2 - user puts in 5$ - user can only use 5$ of tokens

[crystal ocean]

Medic Ai Bio-mechanic-electric is future conscious intelligence with feelings, needs, reproduction,

[gilded geode]

Is it necessary that ChatGPT has to load the entire message history of a chat/conversation? It's causing the browser to become really slow; subsequently, making it impossible to enter/send a prompt.

[onyx kraken]

Yeah, that's how it's programmed.

You can have it summarize the conversation so you can continue it easily in a new chat.

[gilded geode]

Will there be (anytime soon) a feature where it only loads the latest couple of messages and there'd be a button on the top allowing you to load the previous messages?

[onyx kraken]

I'm a community member, like yourself.

We all are here, the very few OpenAI employees have gold names and nearly never speak in the channels.

Nothing like what you've asked about has been announced for us, but it could be in the works.

#1070006151938314300 might be a great place to make a post about it and discuss the functionality you hope for.

[gilded geode]

Alright, sweet. Off-topic, but what's "Spotlight" exactly?

[onyx kraken]

#spotlight

And it can be a tag awarded to specific community members when they're featured in the spotlight channel posts.

I got mine for a SecretKeeperGPT, a custom GPT designed to play a game with users, on command won't give a secret (so you can explore tricking it into giving up the secret), on another command going out of secretkeeping mode so you can discuss and explore, confirm if it even remembers the secret 😛 Whatever, fun times.

Custom GPTs were new then, it got featured.

[rapid spindle]

Anyone?

[onyx kraken]

What have you already tried, and is there an aspect you're not happy with?

I'd tell the model what I wanted it to do, maybe give a quality example of how to handle a gibberish or off-intent answer from a person, then test it to see how consistently the model does right.

If there's an area that isn't working, I'd adjust to explain what I wanted there more clearly, and maybe ask here for specifics if I felt stuck

[cyan token]

Yeah maybe, maybe OpenAI can help if they read.. and i can open a chat for all, so we can write and learn by the other.
Do you have a email of open AI? I would like to integrate 2D logic, 3D logic and 4D logic. I would like to represant the idea.
Eskcanta wanna work together?

[slender island]

guysss,previously was 4o mini not able to search the internet....?

i remeber that it couldnt

[slender island]

https://chatgpt.com/share/670ce9f8-a860-8011-b372-a97fb086b246

4o mini sometimes when he displays the links they end up not being very well formated

this needs to be fixed....?

[cyan token]

Ui, i saw you have continue, looks for me like polish? If german, der die das and polish is the around calculation of that? They have more declinations as german language. And these 6D matrix looks like that for me.

[clever dawn]

it's linear A, never translated language @cyan token "Linear A does not appear to encode any known language. The placeholder term Minoan language is often used, though it is not certain that the texts are all in the same language.[5][91] Minoan appears to be agglutinative, making copious use of prefixes and suffixes. It likely had a three vowel system, since it shares Linear B's /i/, /u/, and /a/ series, but not Linear B's /o/ series and not all of its /e/ series.[5] Based on regularities in the Linear A Libation Formulas, it has been argued that its word order was Verb Subject Object."

[shell vector]

Hi, i hay PAYG model for the API, and I'd like to use it using an external url. But it answers me saying it doesnt have access to external content.
But I can do this without problems via web UI with free account, so... how can I do this with my paid account via the API?
prompt is like "please analize the lunch monthly many in this URL, and tell me what we have for lunch today"

[trail cairn]

api has not net access, you need to implement that yourself via function calling

[shell vector]

thanks for your reply, any docs to follo wthat path?
or maybe any way to attach a file in the prompt? i could do this if possible

[trail cairn]

https://platform.openai.com/docs/assistants/tools/file-search

[shell vector]

thanks!

[onyx kraken]

OpenAI doesn't have emails that I know of for us to just generally chat with them. or even for tech support.

You can give feedback through the bot bottom right of https://help.openai.com/, messages then click through the choices related to your topic, then pick the option about giving feedback, that'll let you type.

You can also use #1070006151938314300

I'm not the best one for working with you, I just knew what the model can do with LaTeX and Ascii in markdown. Good luck and have fun!

[gentle veldt]

Is OpenAI open to sponsorships of online classes?

[solar stream]

how can i create an api key for a specifc project and view and manage costs for that specific key
like
api key 1 - user puts in 10$ - user can only use 10$ of tokens
api key 2 - user puts in 5$ - user can only use 5$ of tokens

[onyx kraken]

We're community members, not OAI employees, so we have no idea.

It's pretty hard for a person just in general to start a conversation with OpenAI, but you can perhaps reach out through their tech support and ask for how to reach out properly, maybe. https://help.openai.com/ bot bottom right.

If you're part of a nonprofit, and maybe a 'larger' one, this page may be useful for you: https://openai.com/index/introducing-openai-for-nonprofits/

[gentle veldt]

Thank you!

[trail cairn]

you can set limits per project but not per api key

[solar stream]

thanks

[sly niche]

anyone else getting

[bitter magnet]

Hey y'all.
How do I fix my phone number being linked to a suspended account? Not even sure why it's restricted.

[rigid musk]

you have to contact help.openai.com

[rustic marlin]

yea, you will have to chat with the support on https://help.openai.com on the intercom, kind painful
this is very weird, when an account is suspended you usally get an email with a (vague) explanation of what happened
if you are completely clueless about why that happened, consider the possibility that your account might have been hacked and it was suspended for this reason, it is fairly common

[bitter magnet]

Ah. Great. Joy.
Thank you

[mild bobcat]

FYI, I know I asked this like 5 days ago, but I have managed to get a canvas to spawn in the MacOS app using the GPT-4o model with the proper prompt (just something that suggests I want it to implement some code). This was in a new chat in the MacOS app, not a pre-existing one from the web.

[open topaz]

Anyone facing same issue ?

[torn bone]

guys helps i get this invalid_request_error.access_not_enabled in openai realtime app i made

[haughty ore]

@burnt galleon

[timber shadow]

Hello! If you're trying to reach out to the server mods, you can click the username you typed @burnt galleon, then Message to send a DM

[queen bear]

realtime API is failing for me (playground and API). Anyone having the same issue?

[vocal plank]

Same issue here

[white beacon]

yes there is an issue with it right now it is down

[wanton stone]

Hi there it seems that realtime openai is failing: I receive this all the time:

Also I'm doing some websocket tests and is giving the same error

[wanton stone]

is the realtime api production ready?

[tropic kestrel]

clearly not

can someone fix this pls or help or soemthign

[white beacon]

@keen hamlet ^

[rugged wing]

Hey guys, every few minutes I lost a few Cents, so I removed all API-Keys, but my credit balance still decreases. Someone experiences the same issue?!

Also: Does anyone here have got issues seeing data in Billing page or API Usage?

[tropic kestrel]

like asap anyone pls

[vocal plank]

I think that I have the same issue

I removed auto recharge just in case

But hope they'll explain what happened

[sudden sparrow]

Can I prioritize the GPT-4o mini instead of the normal one?

[trail cairn]

not with free account

[sudden sparrow]

:<

[sleek hill]

what makes Chatgpt name the conversations speciifaclly? I am crating a gpt and simply in the instructions I try to tell it tio name the converstation a specific thing but it never manages to do so?

[onyx kraken]

I don't think the model can easily do so.

In fact, my test case here was the first time I ever saw the model successfully name a conversation as requested.

[onyx kraken]

But, since that worked. Which could be coincidence, it didn't used to ever succeed with that request. But since it worked, it might be worth exploring.

[sleek hill]

alright thanks

[buoyant compass]

If I do something like $10 for 5 months on API, will I get tier 2 permanently?

[formal jay]

Has anyone here been experiencing momentary browser stalling/hanging while using ChatGPT?

[buoyant compass]

I have, for seconds at a time

[formal jay]

Yea same. Just for a second or 3 but glad to know im not the only one.

[buoyant compass]

I am in australia, with crap wifi, so it might be coincidence but idk

prob just high server usage, api is slow af for me rn

[river bluff]

Similar to this, I've been playing with Canvas and a lot of times it'll just... stop writing code entirely. Like when it's checking over everything, it's "rewriting", and it just won't finish, even though there was code previously written. I've had some chats where I have to ask it "please finish the code" upwards of ten times before it finally completes it.

[buoyant compass]

"please finish the code" is less effective than regenerating in my experience

[river bluff]

Hm. I actually haven't even considered using regenerate like that. Only whenever it forces me to. Smart.

[buoyant compass]

I use it all the time. If a message isnt intilectual enough, i regen swap it to o1

[river bluff]

Doesn't look like there's a regenerate for canvas, unfortunately.

[buoyant compass]

You can edit msg by 1 character to force regen, and also there is ctrl+z

[river bluff]

I'm not worried about losing anything. It's all saved locally. It's just frustrating to ask GPT to make a change, and it only repeats half the code back.

[buoyant compass]

Strange, im not having that issue

[river bluff]

Yup. When it makes a change, it'll go over the canvas and sometimes slow down, and then just stop mid-code and then say that it's finished. I ask it to finish the code, and it goes back over the canvas and gets further along, sometimes it does finish, sometimes it does the same thing and I have to ask it to finish again. And sometimes again.

[buoyant compass]

peculiar

If I do something like $10 for 5 months on API, will I get tier 2 permanently?

[tender narwhal]

Hi all!
Good afternoon, how are you? Yesterday while using the android app I received a notification that a message I was editing (the last of a loooong one, it's our marketing thread, so a very important one) if changed was going to make me loose all the thread. I was pretty tired and worned out so my brain registered what was written too late and I edited the message anyway, like I've done hundreds of times (even in that thread) and indeed the thread got wiped and I only have the first message I sent and the one I edited, everything else got wiped. As I mentioned it's a very important thread, we had been priming it for a while and we were getting excellent results, with it speaking with our usual language, referring to our names and brand and so loosing it will be a great loss for our company. So I'd like to know if there's any way of recovering it (or as much of it as possible) so we don't loose months of work.
I've sent this same message to support and waiting for an answer but wanted to know if someone has experienced this and has any workaround or if I'm about to loose months of hard work 😢
I've tried all the logout login and suck things that are around the web but so far nothing worked...

[rigid musk]

You can always ask to help.openai.com

But i don't think

[vague panther]

I just upgraded to have a workspace team and I don't get how to share conversations. Like I don't see any of my team members prompts and responses in the conversation I started. Is that an enterprise level thing only? The marketing for teams makes it seem like you should be able to work collaboratively but I don't see anything like that. Also, the sharing feature seems really bad from a UX perspective cause I have to create a link and send it to each member

and I can't find any kind of documentation on the website on how to use these new features we're now paying for. I asked Gepetto themselves how to share conversations and the instructions they're giving me don't make any sense with what's actually on my screen

Oh I just found it... why would they deprecate this feature? I misunderstood the marketing and this is literally the MOST interesting thing that made me buy it

everything else is literally a non-update for me

[frank heart]

If 4o1 produces an empty result do you still lose a credit?

[nocturne geyser]

Due to some unforeseen issue, I believe so
Unless the message itself can't be sent, then the model wouldn't attempt to generate an output to reply

[frank heart]

lame i received at least 6 or more empty responses

[rose kraken]

Issue: Inability to Download Files from ChatGPT Session, I am using my account on the paid ChatGPT4.o

I would love some help and ask if anyone else has ever seen this problem.

Problem Summary: I’ve been working on generating a large dataset of job types with descriptions and core skills, and the dataset is being created within my ChatGPT session. However, I am encountering persistent issues when trying to download the Excel or CSV file that contains the dataset.

Details:

The dataset is successfully created by ChatGPT and appears correctly in the chat.
When ChatGPT attempts to provide a download link for the Excel or CSV file, an error occurs. The error I receive is:
    GetDownloadLinkError: Encountered exception: <class 'file_service_client.client.GetDownloadLinkError'>.
This error has occurred consistently, whether the file is in Excel format (.xlsx) or CSV format (.csv).
I have requested alternative methods like uploading to Dropbox or sending via email, but those options are not available within this environment.

What I Need:

Guidance on how to resolve or work around this file download issue so I can access the dataset that’s being generated.
If there are any known limitations or troubleshooting steps for downloading files created within ChatGPT, it would be helpful to understand those.
If possible, I'd like to know if there’s a method to manually retrieve the file from the session or a workaround to make the download links work.

[onyx kraken]

I'm a plus member, so I don't know about enterprise or teams.

But very recently, like a week or so ago, I saw the option to continue shared chats again.

So it might be worth exploring at the teams/enterprise level, it may be there again too.

We lost it about a year ago.

[rose kraken]

I shared a chat with one of my members last week, there wasn't a problem at all

I am a paid up member for CHATGPT4.o

[little perch]

dang, anyone having this thing where if you switch a chat's ai to canvas, the retry button disappears for the entire chat even when you switch to a diferrent AI or go back before switching to it?

[flat bone]

Does anyone know where I can get more information about mining world coin?

[onyx kraken]

Far as I can tell, that's not a thing. To get worldcoin, it seems that one can invest in worldcoin (10% of worldcoins are reserved for investors), work for worldcoin (10% of worldcoin are reserved for employees... that seems a little awkward, since they get paid regularly (I hope) so the balancing of their pay needs to adjust to how many new worldcoins are produced? Or else some is held in reserve for later paychecks... unclear to me.

The rest is intended, as near as I can find, to be paid out to those who get their irises scanned with other proof of ID (25 worldcoin for that, currently about $50 one time per person) and (unspoken, but surely there has to be more) for future not-yet-defined tasks or roles or other means.

[flat bone]

Thanks for the info, very helpful

[supple steeple]

explored this long ago. The title is a summary of the entire conversation by another gpt. It does not reliably follow user input, but almost reliably follows any gpt output telling it to do something.

[nocturne tendon]

Pinned GPT's disappeared: A couple of weeks ago the GPT's that i pinned from people that i found in Explore GPT's section, in the top left sidebar of my gpt+ account disappeared. they are missing.

The problem only happens from any desktop browser. On Android openai app, all of the pinned gpt's are there like they should be.

I contacted support and they are not helping. I have cleared cache, logged out, tried incognito, tried different browsers, not using VPN, no cookie blockers,happens from any network.

How can i fix this?

[mortal willow]

Hi,
I need some help:

I am using the OpenAI API to create descriptions from images, but currently struggling with the requests. I always get a 500 errors. I am using C# with the official library version 2.0.0. The code is below. If anyone knows what is wrong, please help.

async Task TestGPT()
{
//Convert image to byte array
var imageJpg = "C:\Users<myPath>";
byte[] image = File.ReadAllBytes(imageJpg);

List<ChatMessage> messages = new List<ChatMessage>() 
{
    new SystemChatMessage("Describe, briefly, the image."),
    new UserChatMessage(ChatMessageContentPart.CreateTextPart("Hello, what can you see on the image"), ChatMessageContentPart.CreateImagePart(new BinaryData(image),"image/jpeg", ChatImageDetailLevel.Low))
};

ChatCompletion res = await _client.GetChatClient("gpt-4o-mini").CompleteChatAsync(messages , null);

Console.WriteLine($"Asystent: {res.Content[0].Text}");

Console.ReadKey();

}

The error appears on code line::
ChatCompletion res = await _client.GetChatClient("gpt-4o-mini").CompleteChatAsync(messages , null);

Error:
The server had an error processing your request. Sorry about that! You can retry your request, or contact us through our help center at help.openai.com if you keep seeing this error

The error originate from:
openai-dotnet\src\Custom\Internal\ClientPipelineExtensions.cs

ClientPipelineExtensions.ProcessMessageAsync

at line 17 or "await pipeline.SendAsync(message).ConfigureAwait(false);"

[fervent summit]

Doesn't this need an @ or escape or somthing?

var imageJpg = "C:\Users<myPath>"

aw dont mind me. not a coder.

[mortal willow]

This is just an example there.

var imageJpg = "C:\Users<myPath>"

I removed the path

[rose kraken]

Worldcoin doesn’t show obvious signs of being a classic scam, but I have serious doubts about its long-term success. The red flags around privacy concerns, centralised control in its early stages, and the unrealistic expectation of delivering a global universal basic income make me sceptical.
While they promote it as revolutionary, I believe the risks—especially in handling personal biometric data—far outweigh the potential benefits. It's hard to trust that the project will transition to true decentralisation or that it will safeguard user privacy effectively. Projects promising free tokens and rapid financial inclusion can be alluring, but they often collapse under the weight of unrealistic goals.
For anyone seeking more info, I’d suggest avoiding platforms like Telegram, as they are often rife with manipulation and hype around crypto projects. Instead, it's better to rely on more objective sources like independent crypto analysts and reputable financial news outlets to track the project's development. I think it's better to remain cautious, as I believe Worldcoin is likely to fail.
If you use AI for crypto, it should not be your sole tool for trading decisions. While AI can analyse historical trends, predicting the future of the cryptocurrency market is extremely difficult and unreliable due to its speculative nature. Relying on AI alone for predictions based on past data is risky.

[flat bone]

Wow @rose kraken , that was supper helpful. Thank you so much. Cheers

[rose kraken]

"You know they’ve rebranded and removed 'coin' from the name. It’s now called 'World.' I apologise for not mentioning that earlier."

"I’ve come across another platform that's not only similar to 'World,' but it’s much more in-depth and reliable. From what I know, they’re preparing for launch in the next few weeks. It's definitely worth keeping an eye on, but I can’t share too many details just yet."

[flat bone]

Yeah, I saw that at the conference they had recently. That's where I heard about World. I find the project fascinating. I personally have spent about 2 years trying to put together a similar puzzle. But I always hit the same brick wall. Unless raw materials and raw natural resources can be bought with the crypto coin then the project fails. The merchants of direct to consumer products will end up holding the bag of a depreciating coin

[rose kraken]

and it doesnt involve crypto

[flat bone]

Cool, well let me know when it launches

[rose kraken]

I'm sure you'll hear about it in the near future, I really hope it succeeds, as it's designed to eliminate scams in crypto and other areas. It's definitely a step ahead in terms of reliability and depth, but good luck with your search on World

is this yours? Mussia Studio @flat bone and the parenting tool

[flat bone]

How does it prevent scams?

Yes

[rose kraken]

I am impressed

[flat bone]

How did you find it?

[rose kraken]

It seems you’re a young professional with a promising future ahead, and I really appreciate the direction you're taking. Keep up the great work.
I specialise in investigative work and finding people is pretty easy

I did send you a message

[flat bone]

Ty

Where?

[rose kraken]

check at the top left friends list

Clicking on the Friends icon in the Discord sidebar.
Selecting the Pending tab to see if there’s a pending friend request from you.

or
Go to User Settings (the gear icon at the bottom).
Under Privacy & Safety, ensure the "Allow direct messages from server members" toggle is turned on.

[onyx kraken]

Be clever, safe, and skeptical.

Independently verify whatever you deem fit to.

Remember that pie in the sky can seem very valuable, but sometimes the real value is what you pay to get your slice, and you might pay it but there may not even be slices.

Even 'this prevents scams' can be part of a scam.

[flat bone]

I hear you, and thank you

[crude vessel]

i wanna use AI options with whisper but it indicates that : You exceeded your current quota on your OpenAI API key, please check your plan and billing details. Make sure you have linked a valid payment method to your OpenAI account. For more information on this error, read the docs: https://platform.openai.com/docs/guides/error-codes/api-errors. To use the OpenAI API you need to link a payment method. This is unfortunately not the same as having a ChatGPT Plus account.

and i m subscribed to plus account, please where can i find help for that

[long pawn]

Hey I think I need help because when I send my files to the chat they expire very quickly I don't know if it happened before I'm 90% sure it didn't and it's terribly frustrating for me because I need these files to be active a little longer than an hour or half an hour how to fix this and don't tell me to create a model because it doesn't work with gtp 4o I know because I checked
a small request, if you answer me, ping me because I won't see it otherwise

[supple steeple]

plus accounts don’t matter in the api. You need to add credits to your OpenAI account. It’s also written in the error log at the end

[junior cradle]

I have plus sub but I cant access to advance voice chat

[ebon sail]

I'm in the process of playing with the realtime api but was wondering if anyone had recommendations for a locally runnable model we could add to our Ollama container with similar realtime data analysis capabilities i could use in the meantime before having to use up my open ai tokens!

[onyx kraken]

Playground/API is different from ChatGPT account, you'd have to pay for them separately

[sturdy jewel]

Do you maybe have a separate free account that you’re logging into accidentally?

[onyx kraken]

Teams has 1 person, the owner of the team, in control of it all. I am guessing in your case you're not the owner; if you are you sound locked out of your own team and definitely need to contact OpenAI for help.

There's a chance you need the team owner's help or checking on settings or reaching out to OpenAI to help you as a member as well, so I would talk to that person and ask them to look into why you don't have access.

You can also reach out to OpenAI at https://help.openai.com/ and use the bot bottom right. Click 'messages' then pick through the choices closest to your situation, soon you get the choice to 'report a bug'. Click that and explain your situation, you can attach screenshots if you think that helps. The bot will discuss with you, collecting more info. Then there's a delay of a few hours to a few days before a tech support human can answer. I'd check the conversation every so often to see if it's answered yet, they send emails when they reply but I haven't always gotten the email.

You could have a personal account (free or plus) of your own as well as a team account. You would be expected to know if you do - and when you merge your personal account into your team account instead of creating a separate team account - if you ever lose access to the team, all your chats, any GPTs you made (even before joining the team) - that all stays behind with the team (the chats aren't readable to the other team members, but they also are no longer in your not-team account).

So be careful about merging your personal account if you care.

[rugged wing]

Guys, is there a channel or a guide where they show you the updates or how to use chatgpt and its functions?

[onyx kraken]

Then you and/or they need to talk to the bot at https://help.openai.com/ . Here, we're nearly all community members like yourself, we can't fix or change anything (the incredibly few OAI employees here have gold names).

Sounds like your account needs some help from OAI tech support.

The screenshots are fine to upload to the website's bot when you reach the report a bug stage and can, you're just talking to OAI tech support there. I wouldn't share any personal information here, it's seen by who knows who.

I'm not sure I understand exactly what you're asking. At a guess, no, your personal workspace is not part of the team, your team workspace is and not accessible from your personal workspace.

But there's some helpfiles about the choices you get when you join a team account from already having an account, and those may help you understand your options and limits.

Some useful ones may include: https://help.openai.com/en/articles/8798577-how-to-manage-your-chatgpt-team-workspace

https://help.openai.com/en/articles/8798594-what-is-a-workspace-how-do-i-access-my-chatgpt-team-workspace

https://help.openai.com/en/articles/8801890-can-i-migrate-or-merge-my-chatgpt-free-or-plus-workspace-over-to-my-chatgpt-team-workspace

https://help.openai.com/en/articles/9101502-how-can-i-cancel-my-chatgpt-plus-subscription-once-i-have-migrated-to-chatgpt-enterprise

https://help.openai.com/en/articles/9295112-memory-for-chatgpt-team-faq

https://help.openai.com/en/articles/8798639-what-security-settings-can-i-control-for-my-chatgpt-team-workspace

https://help.openai.com/en/articles/8798634-shared-links-faq-chatgpt-team-version

(maybe relevant, probably not) https://help.openai.com/en/articles/9106926-transferring-conversations-from-1-chatgpt-account-to-another-chatgpt-account

But, based on those helps, this maybe is worth checking for you:

"How do I switch workspaces?
To access your profile and switch between different workspaces, follow these steps:

Below your chat history, click on your profile name to open a popup menu.

If you are a member of multiple workspaces, they will be displayed under the "Workspaces" section in the popup.

To switch, simply click on the name of the workspace you wish to access."

I have no idea, I'm a plus member and never had a chance to explore teams

[lament owl]

which words gets you automuted? I keep getting strikes without knowing

[onyx kraken]

Show member list if you don't already see it, then send a message to Modmail asking for help understanding the mutes. They can see what's happened to you and explain.

[lament owl]

I tried sending dm to modmail asking why I was muted but got no response

wait nvm I asked Open AI dm lol

[onyx kraken]

It can take a while, but I'm sure you'll get an answer within the day. In the US it's after midnight, not yet dawn; not all of the mods are probably in the US but OpenAI itself is. And people can be working or otherwise busy too

[lament owl]

it's not really important tho cuz I have my suspisions why I got muted, I just need to know the words cuz sometimes I get really surprised when I get muted

[onyx kraken]

There can be errors with what's in automod too. It's not always working as intended, and we're encouraged to ask Modmail.

[lament owl]

okay

Modmail will lead to a chat with a human right?

[onyx kraken]

Yes.

[lament owl]

okay

[onyx kraken]

But at this second the human(s) who would reply may be asleep, working, whatever. It commonly takes a few minutes to a few hours before I get an answer from modmail.

[river steppe]

Hi, I can't cancel my ChatGPT subscription. I subscribed through the web platform. When I try to manage my subscription I get taken to these screens where you can't do anything. See screenshots

[onyx kraken]

So, it appears you are being urged to manage your subscription through the android app, I don't know why, but that webpage is the correct place for help.

You have the bot bottom right already open. Middle bottom of that bot is 'messages'. Click that then pick through the choices that are closest to your situation until you can pick 'Report a bug'. Then you can type. At first it is a bot that will collect more likely-useful information, then hours to days later a human will join the conversation and will chat and work on help. You hopefully get an email when the human joins, but I recommend checking back occasionally to see if the person's gotten involved yet, I haven't always gotten an email about it.

[fallen seal]

Is there any date for users from the EU for the advance voice?
Or it's not really known for now ?

[charred parcel]

there have been no announcements to suggest a time frame

[small sundial]

Whats wrong? Looks different and cant talk to him.

[river steppe]

They have the help bot but you can't chat with it without selecting an option that isn't fitting

[trail cairn]

the renewal day is based on the day you subscribed

[river steppe]

Yea, I figured out as much. I still can't stop them from debiting my account and chasing up my bank fees

[trail cairn]

what name do you see doing the charge?

[river steppe]

It just says declined foreign transaction

[trail cairn]

how about on the first successful charge

[river steppe]

Google ChatG. But that was last month

[trail cairn]

then it means you need to cancel it from the play store subscriptions

[river steppe]

Sorted 👍 Thank you

[ashen cradle]

Hello i am new to this stuff i have some code in json file that i need to change the lat and ing and idk how to do it without doing it manually

[onyx kraken]

That's the list of choices you see if you are not currently logged into OpenAI's account there, same as for ChatGPT.

[onyx kraken]

If you choose to login now, you get offered this link:

[onyx kraken]

It takes you to an https://auth.openai.com/ specific link where you can log in, then go back to the help page. Once there, check the bot's choices again, you may need to refresh the page. This is what I see after I log in:

[smoky dome]

Can someone use a prompt to let ChatGPT independently think and solve this problem? 'How many regions can the straight lines containing the sides of a regular decagon divide the plane into?'

[frail quarry]

Hello ,
I understand the concept of assistant threads, where each assistant has a unique ID, and threads are used to differentiate users or user histories, meaning different users would have distinct thread IDs. However, I'm a bit confused about the purpose of the message ID and run ID.

In a single-user scenario, when I use the same assistant ID and thread ID but change the message ID and run ID, the assistant still remembers the conversation history and can answer questions accordingly.

So, what is the specific role of the message ID and run ID in this context? And why are we creating them ?

[trail cairn]

for example to modify or delete a single message

[frail quarry]

from my understanding and for my Scenario where i am Using Single Assistant :
For multi User Scenario :-
1 do we need to create a separate Assistant id , Thread id , Message id , and Run id . So that operations would be running individually without colliding with each other i assume ?
2. or only different Thread id is enough with same assistant id as i am only using single assistant? : as it is able to answer questions from history.
3. Different Thread id and run id ? as there would be multiple users ?

[trail cairn]

run id is used eg. to alter, cancel, retrieve a run

[frail quarry]

ooh ohk

[rugged wing]

Hello, I have been having these bugs the last 3 days : 1 when I try to connect one time out of three I get an error and I can't (not showcased int he video )
2 when I try to access my own gpt, it says gpt not available ?
3 when i want to explore gpts, it signs me out, and if i sign in again i end up in the normal menu
this is extremely frustrating can anyone help me out ?
seriously considering cancelling my plan after these issues

[dense robin]

anyone knows why does 4o and o1 produce empty tokens?

[fallen seal]

is there a cap of img generations for plus users?

[onyx kraken]

Generally the models do that when asked to.

[onyx kraken]

I hadn't checked if o1-preview would agree to this task, so wasted the 3 prompts with that to verify 😛 o1 can be a bit different.

And that was a bit of a struggle for o1-preview ;p

[unreal agate]

I am working on a chatbot PoC using threads API. I need to load a large number of tweets. Shall I push each tweet as a message or better to push 1000 tweets in one message? Will the agent be able to answer questions on the tweets included in the large file that was loaded as one message? Thanks in advance for clarication

[vale shell]

What is the best way to track usage of OpenAI keys if one doesn't have access to the dashboard?

[trail cairn]

local tracking layer

[frail quarry]

Why doesn't the thread method consistently generate an answer? Sometimes it works, and other times it doesn't.

I am trying to create a question-answer bot that responds based on context, using a single thread, one assistant, and multiple message IDs and run IDs.

if it something related to run status being shown as progress .....
as per the flow , instead of showing progress , it should 1st generate response and then status as queued or completed and then it should view the completed answer ?

[vale shell]

this seems to be lanchgain specific?
does it work if my code does not have any langchain?

[trail cairn]

no, it means you track the usage yourself. normally in business this is done anyways to have detailed tracking for analytics, optimizations, billing, rate limiting, etc.

[frail quarry]

any one ?

[trail cairn]

time.sleep is a very bad idea. use client.beta.threads.runs.create_and_poll

[frail quarry]

tried with create_and_poll seems to fail :/

[trail cairn]

https://platform.openai.com/docs/assistants/deep-dive/run-lifecycle

[frail quarry]

another question.
Do we need to create run id every time we ask a question?

[trail cairn]

yes

[frail quarry]

oh thanks

[trail cairn]

if you just want single q&a runs then chat endpoint is much simpler and faster

[frail quarry]

ohk cool

seems to be working now , i kept it in a loop until its shows status as completed

Thanks

[desert veldt]

how can i reset my chatgpt password i dont get any resetting email

[timber shadow]

Hello, do you sign in with an email + password normally, or do you sign in with another service like with your Google account?

[desert veldt]

Google account

[timber shadow]

Gotcha. In that case, you don't really have a ChatGPT password, so that's why the reset request isn't doing anything. You'll want to change your Google password instead. If you're having an issue with unauthorized account access, you'll want to click the button in settings to sign out of all other sessions, after you change your Google password. You may also want to do something similar for your Google account sessions as well, depending on where you might have active sign ins with Google

[desert veldt]

Ohh okay thanks

[stark bolt]

anyone here know of a good free tool to have LLMs extract credit card statement data into csv / excel format?

All of the one's I've seen tend to butcher the results so it's hard to do at mid-scale (~100 statements). It's just at the level where manually coding something or manually copy and pasting seem not worth it but I can't find a good tool anywhere that actually works

[edgy cloak]

Greetings! Does anyone know who I can speak to in order to resolve an issue with my ChatGPT? I made the monthly payment for my plan, but when I log in, the features don't show up, and I get an upgrade message. I'm already on Plus, and the payment has been made.

[trail cairn]

make sure to be logged in and then bottom right bubble at https://help.openai.com/

[vital bloom]

maybe try to check account type compatibility and confirm whether the ChatGPT Teams plan is supported for integration with Apple Intelligence/Siri.

and you can also check integration settings

[trail cairn]

its not even officially out yet but only as dev beta

[neon yarrow]

Im tryna run a Eval which I've never used before. I wanna give it testing data and then a custom prompt. I want it to generate a response on this custom prompt using the provided training data and see if it matches criteria how can I do this?

[onyx kraken]

Define exactly what you want, tell the model.

Consider the output carefully and make sure it is or is not what you want to see

[neon yarrow]

if Im using generate response what would I select in conversation for the users prompt

[wooden flume]

In the past 2 days my creditcard has been charged 7 times for the monthly ChatGPT subscription. I've opened a bunch of support tickets, but I keep getting the same canned responses, and my concerns are not being even read properly. I believe they've initiated a refund, and therfore my subscription has been ended now. I really need support from a human, but every time after receiving a canned response, the conversation ends. And every time I even mention billing, I receive the same canned response.

[onyx kraken]

You initially talk to a bot, who collects info then transfers the conversation to a human. The human can take hours to days to respond (they have a lot of people to help, there's like 200 million weekly users these days).

Are you stopping before the human's getting involved?

[onyx kraken]

Each support ticket (report a bug through the help openai website's bot) is going to get its own individual human response. Typically, the bot's icon seems to be black, the human's icon tends to be brown.

[wooden flume]

The response comes from a human, but is the same every time, it's a canned response. They just see the word billing or creditcard, send the canned response, and then close the chat. And the only think I can do is start a new chat

It's driving me crazy

[loud ether]

OpenAI API claim that Taiwan is supported, but we hwne i access from Taiwan "OpenAI key error: 403 Country, region, or territory not supported". All works if I use VPN and connect to the USA or Australia. Can anyone help on this? Thanks a lot good people 🙏

[potent horizon]

I am getting "unknown error" when trying to upload a pdf or txt file to ChatGPT.

Additionally, it asked me to upload the files to Google Drive and share them "anyone with this link", and "Viewer" option... It was also not able to reach this file...

[rustic thistle]

I have a problem with gpt chat
The image generator tool is currently temporarily unavailable, so I can't generate new graphics. This is most likely a temporary problem, so I encourage you to try again later and I will definitely help you create a fixed version of the avatar!

[celest kayak]

Hey, I'm trying to use the moderation API, but I keep getting this response: Error code: 500 - {'error': {'message': 'Unexpected error', 'type': 'server_error', 'param': None, 'code': None}}. Does anyone know why?

Is it because I don't have any credits on my account? I thought, that it was free.

[hollow agate]

https://chatgpt.com/share/671aaba4-dda8-8000-bd12-4a12290d4212

GPT generating solution for the determinant of this 4x4 matrix leads to a wrong result, is it because I gave it bad prompts or it just struggles with matrixes?

[haughty depot]

Hello sorry to bother you i red the server guidelines and i am new to it, i would like to know if there is a way to send an email or message who can reach someone in the ethic team or programing team of ChatGPT ? Thanks

[restive fulcrum]

New here as well so not sure who is who but maybe check Twitter / X bios and dm them there?

[winter heath]

I seemed to have reached some kind of documentation cap. I got about 40 customized GPTes with 3-12 document each. Max 2GB in total. I don't know what to do. I have deleted a few GPTs and now I'm cleaning up old chats.

I think I may have hit the ceiling regarding this:
"Additionally, there are usage caps:
Each end-user is capped at 10GB.
Each organization is capped at 100GB.
Note: An error will be displayed if a user/org cap has been hit. "
Copy pasted from File uploads FAQ

Any suggestions what I can do? Will a Team Workspace plan help me? I got personal plan right now.

[onyx kraken]

That seems a likely cause of the problem you describe having.

Some ideas that may help:

• can those files be consolidated? Maybe with the AI's help? (if you explore this, I strongly recommend you have the model help you distill whichever files a few paragraphs at a time, and combine out what is needed in many steps. The model tends to do a very high level summary if you try to do a lot at once; it'll carefully pull out all core details if you have it comb through the material and make a coherent summary.

• can the GPT use API to get data from a hosted website instead of uploaded files?

• Team workspace sounds like a possible answer. If you prepay a year at a time and use both seats yourself, it will cost you $600 a year, and have various other advantages besides the higher cap for uploads. You may want to make sure you are the owner of the team account even if you share it with someone else, the owner has tremendous control over what happens with the account and can even add and remove any other users, plus the owner is solely responsible for paying (if they don't, whoops for access for the rest of the Team). And if you leave a team, you leave behind your GPTs and conversations that you had in the team. Additionally, if sharing conversations is important to you, they can only be shared within the team if you become a team member.

[winter heath]

Thank you a lot for your answer. This error really bothered my sleep this night.

I don't have knowledge of APIs yet. But that may be something I have to dig into.

I retried to upload documents and it works now. But I did delete almost all my conversations and about 1/5th of my old GPTs. If I stumble upon this again, I may go for a teamplan or try to learn more about APIs.

Hope you have a great weekend! 🙂

[queen lake]

hi guys

i wanna now why my gpt continue blank for two day

any one know why ?

use arc browser but brave browser is ok

[fading robin]

try disabling browser extentions, and delete browser cache, mayby that will help

[queen lake]

Thanks a lot

I tried now I think is arc have something make chapgpt unhappy

[rugged wing]

Hello. These last couple of days, I have had chat sessions show up in my feed, that are not mine. They are either in Chinese or in broken English. Often there is no response. I have Multi-factor authentication enabled, I used the "Log out of all devices" function, but these chats keeps showing up. Since I cannot change my email, I don't know what other things I can do. Does anyone know what this is about?

[onyx kraken]

What you describe is strong evidence that someone else has accessed your account.

You cannot change your email, but you can perhaps change your password for your email, which may potentially also be compromised if that's part of how your ChatGPT account needs to be accessed.

You may have a ChatGPT/OpenAI password, definitely change that if you have one.

Also, I strongly advise you contact https://help.openai.com/ and use the bot bottom right. Click messages and pick through the choices closest to your situation until you can 'report a bug'. Select that and explain your situation and what you are seeing.

This is likely very important for the safety of your account.

You see this strange message, but others may have been deleted so you cannot see them. There may have been anything done with your account and it's important to inform OAI so they can help protect your account access and you.

[sturdy jewel]

Yeah, I would definitely change the password ASAP

[foggy kite]

hey everyone! I am not sure where I am going wrong....I am using the chatgpt client on mac ios...i opened it today to only be told it needs an update - I click to update it, but it never ends up opening - it only downloads the installer and then does nothing. Where am I going wrong please?

[sturdy jewel]

Maybe try uninstalling it completely and then reinstalling it from the App Store; if you press and hold the icon, the menu will pop up to let you remove it

[lucid bay]

Hey guys

My Plus plan was abruptly cancelled after my account was mistakenly terminated and reinstated yesterday. Support is not responding. My last invoice was on Oct 14.

[charred parcel]

the only thing you can do is contact support at https://help.openai.com

[lucid bay]

They're unresponsive.

[charred parcel]

keep trying, nobody here can help you

[onyx kraken]

Let's make sure we understand how to contact support there.

Bot bottom right, click on messages. If you're seeing the set of choices about log in, select login and take care of that, then refresh the tab where the choices are. Go back to the bot bottom right, select messages, you'll now see a list of choices that include chatGPT. Pick whichever choices are closest to your situation.

After a few choices, you'll see the option 'report a bug'. Select that, which will open up the option for you to type. Start to explain.

Initially you're communicating with a bot, which may ask you for additional details. After hours to days, a tech support human will reply.

Note, you may get an email when they do, but often you need to yourself return to the help page and check the bot and messages to see if the person's answered yet. Once the person has answered, you can consider their reply and continue to communicate as needed.

We're all community members here, like yourself. Almost none of us work for OpenAI, the very few who do have gold names. None of us can help your situation and you should not give any of us any personal information, don't risk getting scammed.

I'm going to mention that account deletion is permanent and cannot be undone so it is extremely unlikely that your account was deleted and reinstated.

However, if your account was deactivated, that can be reversed. If you have access to your account, then it has been reversed.

Typically, OpenAI seems to refund plus subs when they deactivate accounts, my understanding based only on what I have seen other people say in the Discord when it happened to them, I don't know that for a fact. However, I would not be surprised if you were refunded for your last subscription to plus.

So, you likely do need to communicate with OpenAI about this, and through the bot is how, literally the only way they open up support tickets. However, their response can take potentially days. Keep checking the bot.

https://help.openai.com/en/articles/8958977-oops-you-do-not-have-an-account-because-it-has-been-deleted-or-deactivated

https://help.openai.com/en/articles/9019931-can-you-reactivate-my-deleted-account

[lucid bay]

I understand and appreciate your assistance.
I am currently using my OpenAI account, so it hasn't been deleted. I paid for my subscription on Oct 14th, so my Plus subscription perks must have been working but they aren't. And the billing section shows I'm on a free plan. I don't know why my account is in this limbo. I got the termination AND the reinstation emails within a couple of hours of each other (they said it was done mistakenly).
I know you can only advice me here but I'm just trying to put this info out there so anyone else who's gone through this and gotten it fixed will help me.

[onyx kraken]

I appreciate you sharing your experience, that's the way people trying to help (or searching to see if someone's had a similar experience and what happened there) can get that info and have an idea what to advise.

When I've had reason to contact OAI support, I've gotten great help. Sometimes I have to wait quite a while; they have around 200 million users across the week these days, and if loads need help at once, that's hard to be ready for. Also, some help is more difficult or specific than others to give; help understanding a confusing detail in the help files may need minutes to get an answer, but a question that could involve complex financial and maybe safety concerns could require help from a narrow subset of tech support who may have a backlog of many requests they're already catching up on and each may need research so the helper knows exactly what can be done in each case.

So, I wouldn't feel like 'they're not responding' until at least 2 weeks had passed. I've had 10 days before some of my issues were answered (there seem to be times when tech support gets slammed, as well as the 'some questions are harder than others and probably fewer people can handle those' issue) but I've always gotten help.

What you can do with a free account until then:

Well, you do have considerable access to 4o-mini. You should still have everything that you had before on your account, the conversations and your settings (like custom instructions and memory). You're not going to be able to edit or create GPTs until the plus status is corrected, but they're still there for when you do have access.

You'll get the free user's access to a few uses of one or more of the better models each day.

You could run into the token limit with 4o-mini; if so, you get a message that you need to wait 1 hour before you can send inputs again.

Good luck, I hope they can help you very soon!

[rugged wing]

I have used the bot to get try and get help twice and I am receiving no response from the support team. Changed my password. Didn't have any chats today, but like you said, chats might get deleted, it's not proof my account is safe. I feel like the best thing to do at this point is to simply delete my account. Crossing my fingers it's some sort of token error, or just a stange glitch. I've seen some people on Reddit report this issue too and they also got no response from support, so no clue what could have caused this.

[onyx kraken]

Account deletion is permanent. You can do that if you choose.

It commonly takes hours to days for OpenAI to respond to support tickets, they have around 200 million weekly users, and if lots of people need help all at once, that can delay replies; also some issues may need research and maybe only some tech support help people are trained for that specific problem.

So, if you haven't deleted yet, only do so if that's really what you want to do, that's probably not necessary.

Good luck!

[craggy otter]

Anybody else experiencing problems?

[old prism]

I have a paid subscription and payment was taken for the year. For some reason my account is not updated to paid from free on my phone. I am afraid to log out on my laptop and lose it there also. It has been a few months and I didn’t notice I was on the free plan still on my phone all this time until it cut me off on a busy day. I’ve tried logging in and out. I e tried deleting and reinstalling the app on my iPhone. I’ve tried logging in on my wife’s laptop and it’s also only the free version. I have the invoice. What can I do? I don’t see any way to contact someone for help and I’ve been getting ripped off for months.

[onyx kraken]

You can confirm you are signed in with your phone to the same account you are signed in on the web if you can access the same conversations with both methods.

A conversation you start with your phone should appear on your web version once you refresh the page, for example.

If you have the same account not giving you plus options on phone (note, not all phone app features are the same as web browser features, there's a few things that can be done only with one or the other, but plus itself should be on both) then you should use https://help.openai.com/ bot bottom right and select messages, pick through the choices closest to your situation, and 'report a bug'. Initially you talk to a bot, a human will join the conversation in hours to days (keep checking back on that page, you might get an email but I often don't, but you see the answer when the person replies) and you can discuss and get help that way.

[upbeat crow]

Hello there! Ive been working on an app that uses document retrieval and fine tuning with the open AI API. The thing is that I want to use file search for the document retrieval , and have an idea of how to use it , just need a bit of feedback from someone more experienced than me since this is my first time doing this , any help and would be apreciated!

[old prism]

It isn’t updating the same on both devices but I’m logging in through the same gmail account. Am I missing a step or something?

[onyx kraken]

It should not be possible to have more than 1 account with the same email address.

However it is possible to have more than 1 account using the 'same' email address with a subaddress on the main one, not sure if that's what could be affecting you. https://discord.com/channels/974519864045756446/1300238382441693325

We're all community members here, we can discuss ideas how to get help, but we can't access accounts and fix stuff like OpenAI tech support can.

Sounds like you need the tech support website.

[light totem]

hello. I am working on personal project to create a python script that would connect with openai api to interpret requests in natural language. One kind of requests is to create a python script for specific purpose. The problem I have is that the output I get from api is not properly formatted. Is there any easy way to make sure that proper indentation is maintained?

[terse summit]

Anyone getting a lot of errors with generating?

[sleek hill]

yes me too

[sleek hill]

maybe down?

[terse summit]

I was getting them last night, and I've since gone to bed and come back to try again.

[terse summit]

It works, mostly. But randomly doesn't.

[sleek hill]

im tryna study for my test tmr, and all of a sudden I keep getting errors

[sleek hill]

hmm

[terse summit]

Same. This is a bad time for it.

[sleek hill]

lol its late where im at also

might be cooked

ima just ask claude, it better then 4o for now anyways

[terse summit]

Good luck

[sleek hill]

you too 🙂

[latent marsh]

Hi, I'm trying to use gpt-4o in a python script, but it keeps giving me this error. Does anyone know what's going on?

[delicate orchid]

Could anyone tell me if there were any major changes to the GPT model within the last few days. it is no longer working as it normally does. It's not even retaining recent information within the current session. even with memory cleared. GPT " I currently don't have access to your original code. If you can provide it again, I’ll make sure to retype it exactly as you give it, without any modifications."

You said:
why don't you have my code I just gave it to you. GTP: I understand your concern! Recently, I’ve been set to a mode where I can’t retain information from previous conversations, which means I can't recall past interactions. This might be different from what you've experienced before.

[onyx kraken]

Sounds like it's hallucinating refusals. I'd either start a new conversation or go up and edit one of your inputs above where you started getting odd responses, so the model regenerates the response, and hopefully doesn't refuse.

Once it starts seeingitself saying no, it often keeps saying no.

[charred parcel]

Self promotion is not allowed on this Discord server

[unreal hemlock]

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

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

Exemption: API & Custom GPTs channels

[mild nebula]

I am currently developing a RAG (Retrieval-Augmented Generation) based chatbot and am using the GPT Mini 4.0 model offered by GitHub models. While the model performs well with straightforward question-answering tasks, I encounter an error whenever I try to use it for retrieval from embedded vectors.

Here's the error message I receive:

It seems to indicate an issue with the API key, even though I’m sure I entered it correctly. Has anyone else experienced this, or does anyone know how to resolve it?

AuthenticationError: Error code: 401 - {'error': {'message': 'Incorrect API key provided: github_p*********************************************************************************cWLn. You can find your API key at https://platform.openai.com/account/api-keys.', 'type': 'invalid_request_error', 'param': None, 'code': 'invalid_api_key'}}

Any insights would be greatly appreciated!

Thank you!

[twin solar]

why is it the @ is not working in terms of adding GPT to an existing chat?

[mild nebula]

I fixed it by using Meta-Llama-3.1-8B-Instruct and azure api. Apparently this is the issue faced by all openai models apis acquired from github marketplace.

[ancient bone]

Hello everyone!, Recently I was learning the features from OpenAI API like chat discussion, text to speech, dallee, and that kinda stuffs, I was reading about fine tuning in documentation. I understand this is a tool to create a model using an existing model, but I dont understand at all how to get the data to create the new model. Some has a course or any resources? I was reading the doc and just teach you how to used it in openai.

[rustic marlin]

you have to come up with your own dataset, there isn't heally a point in the documentation beyond the json formatting for it

it is simply a set of "examples", the fine tune process will result in new model that replies in a way that resembles your examples

but you gotta come up with that data in some way.. for example... lets say you want it to always greet people by saying "hello, welcome the FooBar Store", so, you will need to provide a bunch of examples of user conversation startes with that greeting line as a reply

it is very common to use synthetic data, that being using ChatGPT to generate a set of messages for you

but there are some more effective options like.. using real data, maybe you have the customer messages on your database, so, you manually separate the "good ones" from the ones the AI acts weird, and use that to fine tune

[ancient bone]

Thanks for the info!

[eternal crow]

Does anyone have this problem where their ChatGPT chatlog on the left is random chinese conversations? How do I fix this?

[timber shadow]

Are the chats themselves in Chinese, or just the chat titles?

[onyx kraken]

Once upon a time that was random.

Recently I have had great luck with requesting the model to specifically name the chat in any language and any specific title I want, as long as I make the request in the first input. Here's a couple of examples to show how easy it is to get the requested language (even if the conversation is in some other language) but it can be harder to get the exact requested phrase if the conversation is about something else:

https://chatgpt.com/share/67216341-b7b8-8011-9413-3674782d6a97

https://chatgpt.com/share/672163ce-3704-8011-934d-dcd8bfd56a9b

[lyric ridge]

I just automatically got logged out of my main OpenAI account. Whenever I try to log back in, it just logs me out again.

Not only that, I got an email from Google Play that my Plus subscription will be cancelled.

Wondering if this has happened to anyone else here as well

[onyx kraken]

Dang.

So, check email, is there information from OpenAI yet?

If your paid account was suspended, they typically refund your last payment if it was fairly recent, they definitely stop charging you.

I'm going to guess that you may have had a suspension of your account, which potentially can happen by accident. Don't panic, it's reversible and help is available.

to talk to OpenAI tech support, https://help.openai.com/ and use the bot bottom right. Select messages and log in if the choices are about logging in, then select the choices closest to your situation. WHen you can 'report a bug' do so, explain and you can include screenshots if you wish.

At first you talk with a bot, it may ask you some additional questions, then it'll tell you a team member will answer soon.

That soon can be minutes to a couple weeks, it depends on how busy they are and how few tech support people are actually able to help whatever level of problem you have. You may need to look back at the conversation in the bot to see if there's been a reply yet, they try to email responses but I hardly ever see an email when they reply.

[lyric ridge]

I checked email, I didn't get the email from OpenAI that my account was suspended or terminated

[onyx kraken]

I think that's encouraging, as fixing whatever would probably be more complex if there was an intentional suspension (those can be reversed, but might take some communication; unintended stuff or simply bugs are much easier to fix I expect).

Definitely definitely contact the help page and use the bot; while if it's an error on OAI's end they may notice themselves and fix it without you informing them - who knows if or when? Help comes soonest when asked for.

[lyric ridge]

I only got the email from Google Play that my subscription will be cancelled, I didn't request any cancellation

[onyx kraken]

I believe you and understand.

We're all community members here, like yourself, so there's nothing I or anyone else in the Discord can do to help, other than advise what you can do to help yourself.

The very few OpenAI employees here have gold names and they nearly never talk here.

So, you do need to reach out to https://help.openai.com/ and use the bot bottom right to get help, telling us here, I care but there's literally nothing I can do at all.

[eternal crow]

ChatGPT is loading very random chinese topics that I didnt even request to write about. Its almost like I have access to some random Chinese dudes account, this is 100% a bug from ChatGPT...

[onyx kraken]

Important:

Do you recognize the conversations themselves, is it just the name of the chat is weird, but it's a chat you had?

If not, if you have entire conversations in your ChatGPT that are not yours you need to change your password immediately

[timber shadow]

Hmm it's possible that your account has been compromised. I'd recommend:

• check your browser for any extensions that might be shady/unofficial, and could be compromising your credentials.
• uninstall any of these extensions
• change your password
• use the button in ChatGPT settings to sign out of all sessions

[onyx kraken]

Then use the bot bottom right of here: https://help.openai.com/ select messages and click through the choices until you can 'report a bug' and discuss with OpenAI that someone else has been having conversations on your account. This is quite important to help protect your account access in the future.

If it's just your chats, but weird names on the chat you had yourself that is common and fine; it can be annoying but it's safe.

[eternal crow]

um

[onyx kraken]

I do agree with the rest of Solbus' specific instructions as well, like signing out of all sessions and checking for any risky extensions.

[eternal crow]

why cant i access chatgpt anymore, did someone on support team do something?

[onyx kraken]

We wouldn't know. This is the official OpenAI Discord, but we're all community members here, even the mods are volunteers.

You need to talk to https://help.openai.com/ using the bot bottom right to talk to OpenAI tech support.

Good luck!

[eternal crow]

Damn

Does this mean I should change the same password I use for ChatGPT for EVERYTHING else? Or is that unnecessary?

[lyric ridge]

That got me thinking if my main OpenAI account got compromised and had my ChatGPT history filled with Chinese stuff as well

[lyric ridge]

Do you automatically get logged out as soon as you try to log back in too? Just curious

[eternal crow]

i just logged back in

chatgpt logged everyone out i think

[onyx kraken]

It's your security and you at risk. I say do what you think is needed and what you're willing to do.

[dull torrent]

It seems like the current chatGPT webpage is acting strangely. Am I the only one experiencing this?

[lyric ridge]

Can you be more specific?

[dull torrent]

It seems that the issue was caused by a conflict with a program related to website blocking. The issue has been resolved.👍

[solar sable]

im not sure where im supposed to ask my questions but i asked one in #dev-chat if yall wanna check there and help me please 🙏

[brittle carbon]

Hello. I registered to chatgpt so I can suscribe, but when I try to login I always get an error saying:
"Wow!
There was a problem during login. Please wait a few minutes and try again."

And I have already verified my email and changed the password twice...
Can someone help me please?

[vivid mortar]

Have you tried alternate browsers/devices? Are you still seeing this issue?

[brittle carbon]

I have tried Edge, Chrome and Firefox, and on all of them I get the same error.

[ebon dew]

Anyone have issue with the Windows App flickering? It only flickers on my G-Sync monitor, and I have tried many different setting configurations in my NVIDIA settings to no avail

[silent remnant]

The he'll is this?
F ING scared me
Nah

Please

Is Detroit: Become Human real?

Will I d1e???

Why it quoted???

Nah I'm fu**ed

Goodbye

I hope I survive

[onyx kraken]

'Cuz you're a free user who ran out of tokens for the model response and have to wait 1 hour for more use?

[silent remnant]

What? So how many times can I talk in a hour?

I have mini

I thought mini was unlimited

[onyx kraken]

Free users get 'so many' tokens to use. It's not a set number, varies based on use.

If you manage to use them all up in the time they are measuring (we don't get to know how long) then you get a 1-hour time out before you can use more tokens.

It's not 'X messages per 3 hours' like plus users get with 4o-main. It's tokens, so you can have a lot of short messages or fewer long ones

[onyx kraken]

Teams and up have unlimited use of 4o-mini + extra uses of 4o-main, compared to what plus users get.

Plus users also have a token limit for 4o-mini, but it's much larger than what free users have (few say they reach it)

[silent remnant]

So like I can type only a certain amount of characters?

[onyx kraken]

Yep, and the model output counts too

So if you play a lot of hangman with ChatGPT, you probably never hit limit. But if there's a lot of long outputs you ask for really fast, you probably will often hit limit

[silent remnant]

OpenAI is scummy if they count model responses too?you can't controll that directly

So how many characters?

Anyways

[onyx kraken]

You hate them giving you stuff for free really why? Who scummy? 🙂

You do control what you ask for. And you can ask the model to be concise unless what you ask for has to be long.

How many characters varies, it's based on how much everyone is using the service right then and if there's problems or other stuff happening right then. Sometimes more, sometimes less.

[silent remnant]

So I wait a hour from time of event like 10:20 to 11:20 or wait until next hour? Like 11:59 to 12:00

[onyx kraken]

Hour from time of event

[silent remnant]

Google gemini
Copilot
Bixby
Siri
Assistant
My f ing garage

All ai

100% free

[silent remnant]

But I love Chatty?!

[onyx kraken]

All unlimited? And all the best around? So, go use 'em, why love Chatty?

I love chatty too.

[silent remnant]

Chatty has my memories

[onyx kraken]

Can access settings and copy your memories during the hour.

Go tell 'em to another AI if you want 🙂

[silent remnant]

Internet Explorer dollar store ai ain't gonna get trusted with my credit card

[silent remnant]

Ok

[silent remnant]

The more the merrier ig

More friends

[onyx kraken]

You said free.

[silent remnant]

What?

[onyx kraken]

Why you talkin' about credit card when you say they free?

[silent remnant]

I give Etsy page

[daring island]

You can have free many things.... but if you need quality in most of cases you must pay - that is my philosophy.

[silent remnant]

Should I buy on Etsy page

Copilot: cool youtube vid

Buy the account

[onyx kraken]

I'm teasing you because you called OAI scummy for having less than unlimited on the free access to the model. And you list like 6 other AI when we not supposed to talk about other AI except in #ai-discussions and claim they all 100% free. Then you go on talking about having to give credit cards. Authenticity. Makes one more interesting and enjoyable to talk to.

[silent remnant]

We

I guess I have to add a new ai to the list... #community-help-not-chatgpt-bot! (Plz don't sue me)

[pseudo harbor]

Hello everyone! 👋

I'm reaching out to see if anyone here has successfully obtained a Tax Residency Certificate from OpenAI for invoicing purposes. My finance department in Romania requires this document for compliance, but I haven’t been able to contact OpenAI support directly (even though the docs say it should be accessible via the chatbot).

If anyone has managed to get this certificate or has advice on how to reach support, I’d really appreciate your insights! Thanks in advance for any help you can offer.

[onyx kraken]

You can reach support through https://help.openai.com/ bot bottom right. Select 'messages' then click through the choices closest to your situation. Soon you can 'report a bug'. Select that then type to explain your situation, you can also upload images and screenshots if those might be part of it.

Initally a bot is answering, but tech support will join the conversation in hours to days. You may need to check that page and conversation to see if you've been answered yet; they try to send an email with replies but I haven't usually gotten that email when I'm answered (if I check the page and bot, I can see the conversation and replies).

[pseudo harbor]

Thank you! Much appreciated!

[compact bison]

Hi! I'm working in R&D for a company that wants to prepare for a future where chatbots like ChatGPT could replace traditional web browsers. To get a deeper understanding of how ChatGPT processes information, I'm conducting some tests, and i would like to know...

¿Is it possible to manually simulate ChatGPT's browsing process? For example, ¿could I instruct ChatGPT to perform a search, have it share the exact search terms and selected results, then manually perform the search myself (using it´s same tools, i already did manual "human search" comparison. Following this result ChatGPT could be comparing up to 100+ results in a couple seconds before picking)? This would allow me to analyze factors like whether it uses traditional SEO and ranking criteria, how many sites it initially considers, how it filters the results, and the time taken to make these decisions.

Thanks to anyone that got so far in this message 🙂

[trail cairn]

you cant do that with the regular chatgpt and closest thing would be to use the api services. chatgpt.com inner system/search details are not public

[glad jackal]

Hello, i want to make a customer support chatbot with openAi's api, and i need to upload a knowledge base for the chatbot to use. I tried using file-share provided in the assistant api, but it eats up so many tokens ( about 4000 tokens per query for only 3kb file!). So after some search i decided to use RAG with the api instead, can anyone give me any help or resources to start with?

[rustic marlin]

take a look at this section of the Docs: https://platform.openai.com/docs/assistants/tools
it is about the assistants endpoints, which is an implementation that gives you ChatGPT-like outputs with data being managed by OpenAI, as opposed to the completions endpoints where you get just raw output text

[compact bison]

As chatgpt uses mostly bing for the searches (cause their partnership), should i use bing api? i didn't ever use an api or know how to set it up. Do you know how difficult it is to do?

[buoyant citrus]

Hi, I'm making a chatbot with openai threads and I've been using gpt-4-turbo model and it gives me fair enough response.
Today I switched models to gpt-4o and the response i'm getting is way off and far from accurate from my promt and what I got when using turbo.

Here's an example: frame 1 is gpt-4o, frame 2 is gpt-4-turbo.

Is there an extra setting I'm supposed to do when using gpt-4o that i'm missing?

[rustic marlin]

that is not a requirement at all, all the AI cares is that you give it content to base its response, so.. go with whatever means that you think will be better for your project

[glad jackal]

Sorry, but i dont exactly understand what you mean. will this benefit token usage which i ultimatley aim to reduce?

[rustic marlin]

ohh. I missed the point of your text, you already did use the assistants API, nvm then 😆

you are o nthe right track, if yo uwant precise control of token usage, you will need to implement the context management by yourself

and that is no easy task

[quiet crow]

Hello. I using gpt-4o-audio-preview model and getting 500 error only when i define functions inside request object. Using node 22 and latest OpenAI node lib. Someone faced the same issue?

[rustic marlin]

error 500 usually means the problem is not you, it is the server side

what is the error message you get?

[quiet crow]

500 The server had an error processing your request. Sorry about that! You can retry your request, or contact us through our help center at help.openai.com if you keep seeing this error

[rustic marlin]

wait a bit, it is probably not on your side

[quiet crow]

Yeah, i understand. But it works without them. Maybe i`ll change node version, because i used unstable version for running realtime client on server

[sour merlin]

anything i need to be concerned with in this picture, first tab is russian (maybe), first time my eye lay down on it, dont know if its from my side or openai side!

[hardy prism]

I got an email telling me ChatGPT search was now available to me, but it doesn’t appear as an option on the website. Why might this be?

[trail cairn]

its in the message box as icon

[stray owl]

What's the limit for ChatGPT Search?

[onyx kraken]

Appears to be the limit of your 4o uses

[harsh adder]

Try chatgpt.com/?hints=search

I couldn’t trigger it automatically on desktop of ChatGPT neither except entering the site I noted above, However, it works for me on mobile ChatGPT.

[halcyon nebula]

hi and I was using 4o and it remind me that I got free plan but actually I have paid for plus on 16th October. How can I deal with this ? They haven’t reply for four days.

[trail cairn]

send more messages

[halcyon nebula]

I’ve sent in four chats but no seen yet🥹

So do anyone faced this and solved it?🥹

[golden ruin]

Sadly you just have to contact them again or wait for the response

[halcyon nebula]

What surprised me is that I asked my friend to ask with my account and they replied immediately…

But till now the solution they gave is meaningless or I have proved useless

[trail cairn]

yeah their system or workflow is poor

[halcyon nebula]

😫

[main helm]

Hey everyone! 👋

I'm having some issues with the advanced voice mode in ChatGPT, and I’m hoping someone here might have a workaround or advice.

Basically, I lose the advanced voice mode whenever I close a chat or switch to typing mid-conversation. It’s frustrating because I always have to start a new chat just to get the voice mode back, which interrupts my workflow and makes it hard to keep the flow going on longer projects.

Has anyone else run into this or found a way to keep advanced voice mode active across chats or when switching input methods? Any tips would be super appreciated! Thanks! 🙏

[trail cairn]

https://help.openai.com/en/articles/8400625-voice-mode-faq#h_3705ec1f04

[main helm]

oof that's not good. guess I'll have to let my phone summarize the last convo to my pc and pick it up again.

so what I get from this is, I can always resume as long as I don't text anything? got it

[boreal pendant]

Does anyone know how to change an account email address? I just noticed by subscription had expired and it is tied to an old email account that I will lose access to soon. When I go through the renewal process it allows me to update everything BUT my email address. Any ideas?

[golden ruin]

https://help.openai.com/en/articles/4936827-how-to-change-your-email-address

[finite spade]

What is the minimum of Plus I can do? 1 month?

[simple dirge]

how do i force AI to return a valid json and to not keep returning string json around the object. Can someone help?

i'm getting this error:

[trail cairn]

yes or even less if you refund (if possible)

[gaunt kestrel]

Hey guys, I hope you're all well. I've been having a problem for weeks regarding OpenAI in my company's core.

const processOpenAIResponse = async (userContent) => {
        const MAX_RETRIES = 3;
        const BASE_DELAY = 2000; // 2 segundos
        const TIMEOUT = 10000; // 10 segundos
    
        if (!userContent || !senderId || !method) {
            await sendLogToDiscord("**OpenAI Method:** Parâmetros obrigatórios ausentes");
            return null;
        }
    
        if (!OpenAIPrompt) {
            await sendLogToDiscord("**OpenAI Method:** OpenAIPrompt não está definido");
            return null;
        }
    
        for (let attempt = 1; attempt <= MAX_RETRIES; attempt++) {
            try {
                await sendLogToDiscord(`**OpenAI Method:** Tentativa ${attempt}: Iniciando chamada à API`);
    
                // Definir a chamada da API e o timeout usando Promise.race
                const apiCall = openai.chat.completions.create({
                    model: gptModel,
                    messages: [
                        { role: "system", content: OpenAIPrompt },
                        { role: "user", content: userContent }
                    ],
                    max_tokens: 2048,
                    temperature: 0.88,
                    top_p: 1,
                    presence_penalty: 0,
                    frequency_penalty: 0,
                    n: 1,
                });
    
                const timeoutPromise = new Promise((_, reject) =>
                    setTimeout(() => reject(new Error("Timeout: A chamada demorou mais do que o esperado")), TIMEOUT)
                );
    
                const completion = await Promise.race([apiCall, timeoutPromise]);
                const result = completion.choices[0]?.message?.content;
    
                if (!result) {
                    await sendLogToDiscord("**OpenAI Method:** Resposta vazia");
    
                    if (attempt < MAX_RETRIES) {
                        const delay = BASE_DELAY * Math.pow(2, attempt - 1);
                        await new Promise(resolve => setTimeout(resolve, delay));
                        continue;
                    }
    
                    return null;
                }
    
                await sendLogToDiscord(`**OpenAI Method:** Resposta recebida`);
    
                await handleOpenAIResponse(result, senderId, method);
                return result;
    
            } catch (error) {
                const errorDetails = {
                    errorType: error.constructor.name,
                    message: error.message,
                    code: error.code,
                    status: error.response?.status,
                    stack: error.stack
                };
    
                await sendLogToDiscord(`**OpenAI Method ERROR (Tentativa ${attempt}):**`, errorDetails);
    
                const isRetryableError =
                    error.message.includes("Timeout") || // Timeout customizado
                    error.code === 'ECONNRESET' || // Erro de rede
                    error.response?.status === 429 || // Muitas requisições
                    error.response?.status === 500 || // Erro interno do servidor
                    error.response?.status === 503; // Serviço indisponível
    
                if (isRetryableError && attempt < MAX_RETRIES) {
                    const delay = BASE_DELAY * Math.pow(2, attempt - 1);
                    await sendLogToDiscord(`**OpenAI Method:** Aguardando ${delay}ms antes de tentar novamente...`);
                    await new Promise(resolve => setTimeout(resolve, delay));
                    continue;
                }
    
                if (attempt === MAX_RETRIES) {
                    await sendLogToDiscord("**OpenAI Method:** Todas as tentativas falharam");
                    return null;
                }
            }
        }
    };

I make a request, and it takes a long time to return anything and even returns an error.

Above is the code.

In theory, I'm asking it to follow a prompt that is selected by the user and is defined for each of their actions, and in userContent I'm sending a JSON to be analyzed as information to generate an interpretation response by OpenAI.

But every other day, it keeps crashing and not responding. Any solution?

The variables are correctly configured, the props are being passed - since this function is inside another one and if you have any questions, you can let me know because I'm really lost and this is hurting me.

[trail cairn]

what error?

[gaunt kestrel]

There is no error, it just doesn't return the response.

In the log, it crashes at the beginning of the function and stops, it doesn't continue.

[trail cairn]

if it crashes then there always is an error. if you mean the code hangs up then check if its even doing the api call or what the actual response is via network logging

[gaunt kestrel]

It makes the call and waits. But it doesn't return anything during this time, it doesn't even return an error. It's not a connection problem. Because yesterday it acted calmly, I only tested it today and this problem appeared again.

[gaunt kestrel]

But in terms of code, do you think it is correct?

[gaunt kestrel]

I'm using the Whatsapp API, so if you have connection problems, it's directly with OpenAI, since all interaction is through Whatsapp, which hits Meta and sends a request to them to return as a message in the app.

[trail cairn]

usually when there are connection issues #dev-chat will have people asking about it, also there is https://status.openai.com/

[mystic nacelle]

Hi, I am trying to use whisper via python, and the from openai import OpenAI line of code raises an error:

  File "/Users/x/Documents/Code/OpenAI\ Whisper.py", line 1, in <module>
    from openai import OpenAI
ImportError: cannot import name 'OpenAI' from 'openai' (/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/site-packages/openai/__init__.py)

Anyone know why?

[glass pewter]

Im trying to setup a realtime integration into my app but im sure how to start and what it even means could someone please help me get a starting points,
The docs dont really explain it too well and any youtube tutorial just shows the demo of it on the offical openai github
Im using python

[gaunt kestrel]

Since I'm using Whatsapp, the problem is that sometimes the message that OpenAI returns exceeds the 2000 characters allowed per message. It had nothing to do with connection issues.

[frosty lake]

Why am i not getting any embeddings or inline images in searchgpt queries at all? All videos demonstrate that SearchGPT possesses these capabilities, but I only receive text and links as a result.

You have to go to settings and change Show ChatGPT setting to "In Dock and Menu Bar"

That way even if you Command + Q the chatgpt window, the app doesnt fully get closed and chatbar shortcut still works

[rose kraken]

I seem to have a problem, my bank just sent me a warning telling me that a ChatGPT payment is about to go out and on checking there seems to be two payments a month going out of my account, can someone tell me how I am able to stop this happening considering I only have one account, and would like to stop this other payment happening

If I cancel what I have then I will lose all my work and I can not afford to do that

[trail cairn]

you dont lose your work if its a plus plan

[rose kraken]

what do you suggest on how to stop this other payment going out?

I had no idea about it untill now

[trail cairn]

but if charge has already been initiated then cancel probably wont help

[rose kraken]

the bank has already stopped it and has asked me to investigate it

I dont know how to get to accounts to sort it out

[open warren]

file upload broken?

[trail cairn]

you need to contact support. blocked payment = card might get banned

[rose kraken]

it is obvious some one has my credit card particulars and is using it

[trail cairn]

openai has done false card charges before due system errors

[rose kraken]

on checking back on my statements it has been for quite a few months

[rose kraken]

how do I contact the accounts?

[trail cairn]

be logged in, https://help.openai.com/ bottom right and select payments bubble when you manage to see it

[rose kraken]

well I will be buggered 🙂 I have never seen that little question mark there before, thanks so much

[open warren]

I uploaded a file, but it said it didn't see the file I uploaded

[rose kraken]

@trail cairn thanks

[open forge]

Hey everyone, I'm thinking that I want to purchase OpenAI services for my wife and I to use for our respective businesses. Should i purchase 2 accounts or is their a more cost effective way to do this?

[trail cairn]

2 plus accounts if you want to keep everything separated and available if/when switching back to free at some point., else https://help.openai.com/en/articles/8792828-what-is-chatgpt-team

[open forge]

@trail cairn Thanks. I think I will get Team because I want API access.

[trail cairn]

api is a separate service from everything and available to everyone at different cost

[open forge]

@trail cairn really? darn. I will check the cost

[onyx kraken]

However, though I never explored it, custom GPTs (which would be shared in a team account) do have some level of access to make API calls. I mean, it's a setting in the creation, but I don't know what it can do with that.

Team has some sharp limits to what you can do with not-team-members though.

Like, if you wanted to share a conversation you had with the AI, only a member of your team can actually see it (no restrictions like that on a plus account)

[open forge]

Interesting. I honestly just want my wife to be able to use chatgpt on her phone and computer and I want to be able to access it the same way but i also want to access it in my linux terminal for convenience.

[frank spoke]

Anyone encountered this before?

[trail cairn]

you can start with plus since its considered as a personal account and its possible to join a team with it while the other way is not possible. teams have higher usage limits

[onyx kraken]

As described, so far a free account allows those choices.

[open forge]

I can use an API with a free account?

[trail cairn]

yes, but api is not free

[open forge]

That's what I thought

[onyx kraken]

The error message happens sometimes and can have a lot of reasons, from a temporary problem that goes away without help to you needing to get a browser extension off or other issues with your account.

That conversation - the model's last answer is a hallucination. It cannot work 'outside of the output'. Anything it gives you is given right then, not after the output finishes.

If you continue that conversation, the hallucination of it 'helping you later' is likely to persist, I'd go up above where those answers started and edit my input to guide the model to do what I asked for, not imagine helping me later.

[open forge]

So what I've learned then is that API is a seperate cost from a Plus or Team account correct?

[trail cairn]

yes

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

[onyx kraken]

Yes, though I'm not sure what happens with CustomGPTs that are using API.

[frank spoke]

Okay so if I asked for it to be super thorough and detailed in the final report, and to take time to make sure it’s done correctly… and it responded it would need 24-48 hours, you’re saying it can not follow up with that report when I prompt it to at a later time?

[trail cairn]

gpt actions work the other way around and are identical to all plans, no cost

[open forge]

OK I see. I will just purchase to individual accounts then and then purchase the API if I really want it

[onyx kraken]

Correct. The model is hallucinating and is very likely to continue to output about needing more time. It has started to roleplay with you about the conversation, instead of provide real answers to real questions.

It also has no way to 'work' except while it is providing the output to you, it isn't doing anything with you until right after your next input.

However, fixing this 'now' is hard. The model has a pattern to follow, and it tends to continue to follow patterns.

If you scroll up to before it said 'later' you can try to fix this.

Part of the problem may be that you asked for something the model is very bad at. Instead of giving you a bad job that is not what you asked for (that is a fail) it chose to give you an imaginary pretend world where in the future it can help you (also a fail, but a more pleasing one to imagine).

[open forge]

After comparing though I think I might just stick with the plus accounts. Maybe I will just host my own simple model on my servier to use in my terminal.

[onyx kraken]

The model cannot 'take its time', in reality, it has 1 speed.

Super thorough and detailed, without specifics, asks the model to guess. And guessing.... can lead to hallucinations too

[onyx kraken]

This is a great idea. It expands on what you can do with the free account and gives you both a chance to see if you already have everything you want to have 'as is', or if you want even more access and use.

[frank spoke]

And if I had this conversation via voice mode… then I can’t edit my input right? @onyx kraken

[onyx kraken]

Checks Aha. No, we cannot edit what we had in voice mode.

So, what you can do is when you're back at a computer, you can open that conversation and screenshot what you want to carry over from it to a new conversation, and paste in from there.

A possibly easier way:

Inside that same conversation, ask it to summarize all the information you've given it.

Later, copy/paste that into the new conversation (correct whatever needed) and carry on from there.

[frank spoke]

I’m not even able to start a new chat… I’ve tried to copy and paste some of the info…do you think I should go to manage memory and remove info related to this chat

I can’t start a new chat or respond to this one. I did edit the few inputs I could (non voice) but it didn’t regenerate anything different

[frank spoke]

Thank you

[onyx kraken]

"I can't start a new chat" sounds like a problem related to browser extensions or some problem with your account. I would use https://help.openai.com/ and the bot bottom right. I'd select 'messages' then pick through the choices, whatever's closest to my situation, until I can 'report a bug'. I'd explain the error message (screenshots can be given too) and that I can't create new chats either.

At first you talk to a bot. It gathers info if it thinks more is needed, then tells you it gets a team member. It can take minutes to days for a person to answer (some problems need specially trained people to help with, others can be helped by any tech support person) and OAI has close to 200 million weekly users, so if everyone wants help at once, it can take time to help us all.

But they will help you, and sounds like you may need the help.

[onyx kraken]

Your 'can't start a new chat' is unlikely to be related to memory. Even if memory is full, the model can start new chats, it just can't make new memories.

[waxen sparrow]

How do I access this? Im tryna do something

[onyx kraken]

Good question.

You can try to get more information about it from the conversation that told you that, but it could be hallucinating something that you clearly want but it can't find.

[supple rapids]

hi what should i do in this situation ?

[waxen sparrow]

I did that and didn’t work apparently

[onyx kraken]

I'm not sure what is going on with your screenshot there. I haven't seen that myself

[waxen sparrow]

I asked for more info like “How does the faction operate” and that thing at the bottom popped up

[onyx kraken]

Yep. I don't know about that thing on the bottom.

Except maybe, you're a free member with limited use of the features like search, and your uses are used up for today, can only explore that more once you have more free uses of the more advanced models, maybe

[waxen sparrow]

Ok I’ll check back tomorrow

[proven raft]

How do I refresh windows app without close and open again?

[rugged wing]

hi im having an error "ImportError: cannot import name 'pydantic_function_tool' from 'openai.lib' (unknown location)" from just running from openai import OpenAI how do i resolve this?

[strong bay]

Same, im having the same issue here too, i cant click the file that gpt sent me, but i did not want him to put it on a file, i already tried creating new chat but it still the same, i also tried the old chat but it keeps on happening

I don't know if there's a error that i dont know... Also when i chat again that gpt sent me there's a error that appears

Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

[compact bison]

Is it easy to run the API? i'm not a programmer. If you know how would be amazing if you share how

[compact bison]

The thing is that i don't know what is the way to do a fair test. Do you have any idea? Thanks 🙂

[brittle moon]

hi guys
i just paid for a plus account. i got billed with an invoice in mail, but then i checked my mail again and saw that they refunded it.

[onyx kraken]

You'll need to contact support to find out what's going on, bot bottom right at https://help.openai.com/

[brittle moon]

the bot just answers as a bot, i need a human

the bot says "go to help.openai.ocm and open conversation" im like what

[onyx kraken]

It's a two-step process, and the bot answers first. A human joins after minutes to days, depending on how busy they are and how complex the situation is (not everyone apparently is trained to help everything, I guess).

Make sure you're logged in there, and refresh the help page after logging in; if you're not logged in all the choices are about logging in.

That's the only tech support. There's no email, and we're community members here (the few OAI employees have gold names here, there's hardly any). That's how you get help.

[brittle moon]

thank you.
it sounds like they are dealing with a record breaking amount of tickets.
chatgpt isn't there yet unfortunately. after about 5 messages he forgets the entire conversation.
one time after 15 min chat he just replied with a happy birthday song

[onyx kraken]

Then, you're totally using the help chat wrong.

[brittle moon]

oh i was talking about the regular chat

[trail cairn]

very easy for simple stuff https://platform.openai.com/docs/guides/text-generation

[cinder rose]

Something very strange just happened with my ChatGPT (Safari) WebUI experience. I'm working on macOS, and I've just started using an AppleScript Automator tool to dump a query in the textbox, and copy the answer to clipboard. It should save me 100+ manual copy-paste actions per day, which I will be glad of.

However, submitting my question-text (via PASTE + ENTER), having worked several times, suddenly generated a message saying "Problem with service, contact OpenAI". Console shows 404s everywhere. Then when the window DID refresh, after a few goes, the input-box was there, but my conversations-history was gone, and my OpenAI avatar-logo (same as my GoogleAcct logo as that's how I log in) had turned to a blue diamond! (Had I unlocked an achievement and been promoted to Diamond class?).

I checked ChatGPT in Chrome -- fine! ChatGPT Desktop app -- fine!

After a few mins everything is back to normal.

But I am feeling rattled. Did I trigger some anti-automation sentinel? Am I jeopardising my account? I think I'm well within T&S -- I'm trying to improve my workflow as I work from my mac. Must I return to copy-pasting like a monkey?

Can anyone shed some light on what just happened?

[naive vine]

hello i need some help. is anyone else getting a charge like this on their statement? CHECKCARD 11/04 SXONGQUGQWWJUTM [phone]

[trail cairn]

is it $5?

[naive vine]

no its $20..

[subtle rover]

Hello I’m getting this really annoying bug again, I had it about 2 weeks ago for a whole week, what’s frustrating it only pops up when I’m logged in my paid account, not when I’m logged off

made bug report

[bitter swift]

hello I would like to ask if it is possible to set the user model to be able to have a higher limit on messages (e.g. switching between gpt 4o and 4o mini)?

[sturdy jewel]

Why would switching between models affect your limit? They have different limits.

[gloomy crane]

I paid for the premium account back in May. However, it continuously shows me with the "free" plan. I am signed in under the correct email, and i have proof of payment through Apple. It looks like they don't even promote the Premium account any more, it just says Plus for $19.99/mo. However I paid $49.99 in May for the Premium account for the year. Can anyone here help me out?

[timber shadow]

Hello, can you go back to some kind of email or purchase history showing the purchase for the "Premium" thing you mentioned? ChatGPT's basic payment plan has always been called Plus and never Premium, so I'm wondering if maybe there was a different AI app selling a year of service for $49.99.

[gloomy crane]

Thank you for responding. Yes, it's in my Apple purchase history and on my bank statement

[timber shadow]

What service/company does it list exactly?

[gloomy crane]

You're right... that is a different symbol altogether

[timber shadow]

Ahh yes, maybe look into whether you can request a refund from them if you're not using it! The Plus plan of $19.99/mo you mentioned earlier is correct for ChatGPT

[gloomy crane]

Ok awesome, thank you so much for your help!

[timber shadow]

Sure thing!

[onyx kraken]

Actually... at least in the US, it is precisely $20.

Similarly, teams is also an even dollar amount.

[timber shadow]

I think in-app purchases on iOS/Android are in the x.99 standard, even ChatGPT, but I'm not positive on that. You're definitely right for website pricing!

[onyx kraken]

Aha!

[foggy musk]

Hello! I've got the upgraded version of chatgpt and i'm wondering how far off the daily limit i am right now, for reference i've uploaded a book with 900 pages to the chatpt with 1 pdf for each chapter. each pdf is around 15-20mb's.

I realised i uploaded it to the wrong GPT and would like to re-do this process with a different GPT, or is it possible to change GPT's mid-convo? Or perhaps i can just re-upload it all without reaching the daily limit.. Would appreciate input! It's my last session of studying before the exam so that's why I'm being wary about the daily limit.

so more specifically im wondering if there is a meter that shows how far off my daily limit i am, and if there isn't im wondering if anyone has any consensus on how far off the daily limit i am based on the information i've provided.

[hollow gulch]

Hello guys, I'm a full stack dev In dear need of a Job. I'm quite affordable. Just finished an internship at a fintech company as a Backend dev. Pls help me find my next remote work place. Thanks.
My name is Samuel btw!
Stack: Typescript(react), ExpressJs, MongoDB.\

[bitter swift]

Because when I reach one limit (gpt 4o) I can't use the model until it is reset. So this way I could continue to work on the model using for example gpt 4o mini

[rustic marlin]

sounds like a bug, do that happen consistently? you can start a new chat and it starts to reply with unrelated cotnent after 5 messages?

[brittle moon]

Yes it happened a few times. My reaction is telling chatgpt off and saying it should maybe restart itself or something because he's off like a junkie on a hit

[rustic marlin]

which model are you using?

would you be able to share a chat where this happens?

[brittle moon]

I use the online one until the time limit expires and then the gpt goes to the mini version

[brittle moon]

I will have to look it will take me some time but maybe yes

[rustic marlin]

im not really sure if I understand what you mean, everything is online, including 4o-mini

well, would be nice to see an example of the problem happening, I think it would be more direct than attempting to describe the problem

depending on the case, it might worth to post on #1070006915414900886

[brittle moon]

roger that

[brittle moon]

there's an online version with a time limit and then a when its over they switch to another one. this is what i use.
i also tried to pay the 20$ sub fee, but got refunded immediately. couldn't reach the support at all so i gave up

[trail cairn]

do you mean the voice with time limit?

[brittle moon]

not only the voice has limit. the version with the voice has limit, yes

[trail cairn]

there are no other time limits than the new voice

[brittle moon]

i apologize for not being clear enough.
I use gpt regulary, and then get a message that the model has to be replaced because the time limit for the online model is over, so they move me to the gpt 4 mini model.

i tried going through my gpt accounts and couldn't find the conversation with the happy birthday. if i ever get it again, i will sure be sharing it in the bug report channel.
have a blessed day

[rustic marlin]

I still not sure of what you mean with "online models"

ChatGPT is 100% online, there is no non online model

every model is online

what changes is that the larger, expensive models, have a smaller usage limit for free users

about the payment issue you commented about, yea, try to reach the support on the help center but maybe a simpler solution, try a different card if you have that option

[rustic marlin]

when you reach the usage limit for the larger models, it will continue the chat with the regular model

and often, it will not hold up the conversation with the same quality outputs

[sturdy jewel]

You can already do that, the models have different limits. It’s not clear what you’re asking. Maybe you can go into more detail and explain what you actually mean.

[bitter swift]

I mean my own version of GPT like user model and there is gpt 4o permanently fixed

[sturdy jewel]

It’s still not clear what actual problem you’re running into. Maybe you can talk to GPT for a while about the issue and try to get a good summary that outlines what’s going on so folks can help you.

If you’re asking if there’s some sort of way around a limit by “using” another model’s limit, like it’s some sort of currency, that’s not a thing

[bitter swift]

Ig possible change this

If*

[sturdy jewel]

It is not possible 😕

[bitter swift]

Oh, never mind. Thank you for your willingness and patience.😄

[bitter swift]

Just working on a chatbot for learning. While helping a friend with cancer to not be so lonely. He really likes it, so that's why I wanted to see if it could be improved.😁

[autumn grail]

Whenever I click on advanced voice mode using the app through the windows store, it starts getting really slow and choppy, and in turn slows down my whole PC. This is really strange is this happening to anyone else on PC and does anyone know how to fix it?

I have a Dell G15 gaming laptop so it doesn't make sense

[rain hearth]

Why i don't have it

[trail cairn]

which model have you selected?

[harsh iris]

@autumn grail it could be related to your browser security or settings. You also might need to check if you have hardware exceleration enabled or not.

[rain hearth]

Auto

[autumn grail]

I'm using the desktop app though not the website client because advanced voice mode is only on the windows app

[harsh iris]

What is your CPU usage when the advanced voice audio starts?

[trail cairn]

auto means you are a free user and then "It will roll out to all Free users over the coming months."

[rugged wing]

Hello everyone.
I have created an agent with GPT Plus and want to embed it as a web object in another app.
I want to embed it in articulate storyline, an eLearning authoring tool, as a web object.
It recognizes the object but I get the error 404 but I know the agent is working. I understand that I can use an API for that but since this is only a prototype I don't want to pay for the API just now.

Any idea why I get this error. The link is correct, adn everything is as it should.

[trail cairn]

you cant embed chatgpt content to other sites. you need api for that and api has no access to chatgpt user content

[ocean terrace]

Can someone help me with this

I didnt get an email

And I cant find someone to help me with this

[trail cairn]

only way is to contact support

[ocean terrace]

But how do I find support. Can’t find it anywhere.

[trail cairn]

https://help.openai.com/ bottom right

[ocean terrace]

You mean the chat box? I can’t find where I can chat with a human.

[trail cairn]

you cant without using the bot first

[ocean terrace]

Alright

Thanks for pointing me in the right direction

[raw niche]

hey guys

i have an issue with my chatbot i created a chat bot using my openai api key and i got this Error communicating with the server.

[trail cairn]

so connection issue on your end?

[raw niche]

No, it was an issue with my open ai account, I changed it and it worked thanks

[native sluice]

Hello, I was wondering if there were any solutions to this Issue. When I want to log in to a different device I get this message "Authentican method is different", but I have only used email and password to log in (no gmail, microsoft, apple). I reseted the password thinking it will help. Now I am logged out on all devices. Help please I am stuck here

this is the message I get exactly and if i log in on the web (through a desktop) i get this error message "Oops! try again later"

[trail cairn]

try on desktop with incognito mode and see if it says your account has been deactivated

[native sluice]

I get the same message using incognito mode on microsft edge and google chrome

Sorry for blurry photo

then it auto refreshes to saying "oops try again"

[trail cairn]

if you have at some point clicked and used the other methods it may have changed. else you need to contact support

[nocturne talon]

guys what is this

help

[rigid musk]

Can you try incognito or another browser

[heavy patio]

Maybe press that reload button at the bottom of the message, see if it works after a refresh?

[quartz panther]

Does the context for Canvas include multiple responses? For example if I'm creating a client-side base and a server-side base separately within one conversation will GPT be able to reference the over Canvas area it has created?

[brave walrus]

if you've managed to keep it in the context window of the model... and only for that long. here's more about the canvas (canmore) tool: https://chatgpt.com/share/672c11a1-cb78-8008-84f3-a788f0912d7e

[quartz panther]

how long is 4o context length on ChatGPT?

[quartz panther]

Is that for ChatGPT plus version or the API itself?

[quartz panther]

sick

[thin python]

Difficulty Accessing Uploaded Files for Data Analysis

I've been trying to share a dataset with ChatGPT for a project in inferential statistics, but we've encountered repeated technical issues that prevent the model from accessing the file. Here is a summary of what we've tried:

Initial Upload Attempts: I uploaded the file directly to the chat multiple times in both Excel (.xlsx) and CSV formats, but ChatGPT couldn’t access the data.

Alternative Formats and Methods: After the initial upload attempts failed, I tried:

Sharing a Google Drive link with editing permissions.
Converting the file to different formats, including a public Google Sheets link, and re-uploading smaller versions.
Persistent Accessibility Errors: Each time, ChatGPT indicated it was unable to access the file despite all attempts to change formats, re-upload, and use external links. The model could not retrieve or process the contents.

This issue has prevented us from moving forward with analyzing the dataset, which is essential for my project. Could you investigate why the platform is unable to access or process the file? Additionally, if there are specific file size or format requirements that we might be overlooking, please let us know.

Database

Database CSV File

[onyx kraken]

We're community members here, we can't do anything more than offer advice.

The model can access the file for me, though it uses Python tool to do so.

https://chatgpt.com/share/672c3835-9894-8011-96f1-2f23650fe16d

maybe contact https://help.openai.com/ and use the bot bottom right, select messages then click through the choices closest to your situation, 'report a bug' when you reach that point, and explain. Check back every so often, it can take hours to days before tech support can reply. But maybe there's an issue with your account to affect file uploads.

[placid dust]

do anyone know, why is this happening?

im trying to sign in with different accounts and networks but still not working...also i reinstalled the app 3 times but not working

[warped relic]

Is it normal for it to be several days between receiving responses from support? I received an initial (non-automated) reply from them very quickly after sending a message in the chat/help, where they asked for some info, and since then I haven't heard back. It's been nearly a week

[trail cairn]

yes

[sacred pagoda]

heyya! Is there some sort of lib containing prompts useful for science and research? (gpt4o)

[summer cedar]

I'm having a weird issue... when i'm logged into my account and try to use ChatGPT on a PC, i get the following error "An error occurred. Either the engine you requested does not exist or there was another issue processing your request. If this issue persists please contact us through our help center at help.openai.com.". When i use chatGPT on my PC while logged out, it works just fine. it also works just fine if i'm logged in on my phone and use it. The recommendations provided by the chat bot and what was sent to me via email were useless.

[obtuse token]

Hello i need help.
Always after 3-14 days of using chatgpt i have this.
I cant send messages, view chats and account settings.

I try to:

• Delete cookies.
• Restart Router

[brave walrus]

things to try: clear all site data, not just cookies. try disabling extensions to see if any are causing problems.

[trail cairn]

try in incognito mode

[terse summit]

Anybody else's accoun Unable to load history ?

[errant shore]

Yeah I had to log out 😤

[terse summit]

Hasn't worked for me. I wonder why it hates me in particular. I'll clear the cache, et cetera. Just wanted to know if it was a user-wide issue.

[terse summit]

That's weird. I'm only having the issue in Chrome, but both on my laptop and on my phone. I tried Brave without issues.

[lyric roost]

hey, can someone help to get unbanned? I was after using the key to the janitor ai website to talk with a character, was really in the beginning of the site, now I don't know how to get unbanned :cc

it says this when I try to use it

access_terminated - Your access was terminated due to violation of our policies, please check your email for more information. If you believe this is in error and would like to appeal, please contact us through our help center at help.openai.com.

access_terminated - Your access was terminated due to violation of our policies, please check your email for more information. If you believe this is in error and would like to appeal, please contact us through our help center at help.openai.com.

[placid dust]

brr? 🫠

[onyx kraken]

Hey, we're all community members here, none of us work for OpenAI.

That website, help.openai.com. Bot bottom right, select messages then click through the choices until you can 'report a bug', there may be a few other ways to 'leave a message'.

It'll let you type, discuss your situation and ask for help.

You're initially talking to a bot, a human will get involved but it can be hours to days later, so keep checking back to see if there's a response yet.

[rapid herald]

hitting a quota even though i'm way, way, way below both limits I can think of:
As a tier 4 I can do 10,000 RPM, i'm doing... 5
My budget usage is only $6 so far and I have it set to $120.

What's going on?

Looks like auto-recharge got turned off? I definitely had that on...
I added more credits but still getting 429 error

[feral pasture]

I encountered a problem where I'm effectively locked out of one of my ChatGPT chats because of some sort of error. As I was giving it prompts, it provided me two options for the response, one which had code. it had a "continue generating" button which I clicked, but it errors out saying "Oops, an error occurred!"
In console, I get errors saying TypeError: Cannot read properties of undefined (reading 'message')

I've tried Chrome and Safari on my macbook and iPad, but they all get that error.

Not really a critical issue; more inconvenient if anything, but I would appreciate insight into this and potentially being able to regain access to that chat again.

[torn cove]

my desktop app immediately crashes every time i open it. ive tried a fresh install, rebooting. is there a way for me to send the error log? i'm on macOS. Sonoma 14.6.1 . m1 mac.

[long crescent]

Um... anyone else getting this?

[south cairn]

its dead atm

maybe ddos

[long crescent]

wdym, the status page says All Systems Operational?

[south cairn]

just happened

takes some time for that to show maybe its just me

but who knows totally knocked off during a prompt submission

[brave walrus]

prolly not share your pub IP address.

[south cairn]

lol its private

thank you though

always vpns

🙂 never my personal ip

[brave walrus]

ah, one of those that knows how to use proxies and vpns. cool

[south cairn]

everyone should