community-help

[jolly kelp]

Uhh is the Regenerate button missing for anyone on the site??

[pliant mountain]

How come even it’s mini 4o, I keep receiving the “exceed rate limit” message. 🥲 I haven’t used ChatGPT for over 12 hours…

[wary urchin]

I typed prompt for create rope brush for affinity designer. (procreate) but it's not clear that.

How to type prompt clearly for create repent rope.png? Are AI or chatGPT able create brush for photoshop, affinity designer, procreate etc?

[onyx kraken]

For a few conversations, actually yes! Consider opening a new tab or a new browser. Some of my tabs can regen and others cannot inside a pre-existing conversation, but starting a new chat and I see regen as an option again.

[jolly kelp]

Uhh how do I contact support regarding this??
Is this a bug??

[onyx kraken]

Seems like a bug, and for me it's already gone/fixed.

If it's still affecting you, you can fill out a #1070006915414900886 here on the discord (be sure to follow the format for the channel, and/or you can go to https://help.openai.com/ and use the bot bottom right; select messages then pick through the choices to identify what your report is about, then report a bug - that lets you type and also paste in screenshots. It can take a few hours to days for someone to answer you there, OpenAI has a lot of weekly users and can get surges in bug reports, but they do get back to us.

[jolly kelp]

I'll see if it went

[jolly kelp]

it's still there for me

[onyx kraken]

What happens when you create a new conversation?

I can revisit conversations I made last night and still don't have the options there: But any new conversation I make now have the option to regen again.

[jolly kelp]

thankfully the regen button is there for the new ones

unfortunately it goes again after I close the tab and come back later

[onyx kraken]

Definitely sounds like a bug, and I'm not affected the same way. Hope you report it if you choose to!

[jolly kelp]

I reported it

[opaque lake]

Hi guys

{
    "message": "You exceeded your current quota, please check your plan and billing details. For more information on this error, read the docs: https://platform.openai.com/docs/guides/error-codes/api-errors.",
    "type": "insufficient_quota",
    "param": null,
    "code": "insufficient_quota"
}

I have got $100 credit on my account.
But when I try to play in playground or when I try to use api keys I created, the same error occurred.

Any help or support would be highly appreciated.
Thanks

Hi bud, did you resolve the issue?
I am facing exactly the same bug now.

[sudden delta]

so can anyone help

its a new day and im still not able to use gpt without using private browsing

this goes for both my school computer AND my gaming pc, completely different pc's on completely different architectures.

[timber shadow]

Is the problem in Chrome on both? If so, are you signed in to the same Google account/do you use the same profile settings on both instances of Chrome?

[sudden delta]

not using chrome

[timber shadow]

What browser? And is it the same browser on both?

[wise lodge]

I keep getting this error periodically regardless of whether the request was valid syntax or not. Does anyone know the cause?

ApiSyntaxError: Could not parse API call kwargs as JSON: exception=Extra data: line 3 column 2 (char 141)

I should clarify that this is when using a custom GPT action.

Request body sent to the tool is valid JSON:

{
  "query": "{ Get { Chunks(hybrid: { query: \"deleted records audit log\", alpha: 0.5 }, limit: 5) { name chunk_number content url } } }"
}```

[wise lodge]

Now it's giving:

tool_error
The outbound call did not succeed, inform the user succinctly but keep important details intact. The complete interaction data follows: An error occurred while executing the plugin.

[thick cargo]

Has everyone lost the ability to use "Regenerate Response"? The icon is gone for me entirely.

[onyx kraken]

Some people have reported it. I was briefly affected, some conversations made during a period don't have it even when I go back to them in the previous chats. But new conversations for me are fine. You might want to use the bottom right bot at https://help.openai.com/ to report a bug, maybe they can directly help you if the problem's ongoing for you

[sudden delta]

edge

[strong tusk]

Hey Folks. A quick question. I keep getting limit reached error for my API Keys but i haven't used any of that. whats happening?

[sudden delta]

it works on firefox but thats non negotiable, edge has been my default for so long i have all my cookies and all my stuff

[timber shadow]

Have you tried tinkering with Edge's privacy settings? Something that might be disabled in private mode where it works, but enabled and possibly causing problems in a normal window. It could be a range of things there unfortunately, so I'm not sure where you point you exactly. Otherwise: have you tried just a simple browser cache clear in Edge?

[trail cairn]

https://platform.openai.com/settings/organization/billing/overview

[shadow horizon]

hi im a pretty big noob when it comes to chat gpt and stuff, does anyone know if theres a way to copy paste this to word without the formatting being super weird?

[timber shadow]

Hello, the formatting for math stuff like this is LaTeX. It looks like Word supports LaTeX as well, I think you can find the option to insert a LaTeX equation in Word maybe near symbols.

So I think you'll need to copy/paste, then additionally copy/paste the LaTeX stuff in Word's process for inserting equations like this.

[shadow horizon]

Thanks for the reply, I tried to copy paste it in that little like box that comes after you press equation (with the latex function on) but it still doesnt look like in chat gpt it just says the same

you mean like in here right?

[sudden delta]

All the time.

I've tried cache clear yep

[timber shadow]

Sorry, I don't have a copy of word to test in myself, but are you copy/pasting the LaTeX code into the equation editor? Attached a few screenshots to show what I mean: (1) get the formula from ChatGPT (2) copy the message (3) copy only the LaTeX formula from the whole message (4) paste the formula into the equation editor (I just used a website that renders LaTeX to test it)

[timber shadow]

All the time you're checking privacy settings you mean? And just haven't found one to change that fixes it?

[sudden delta]

i just did a full wipe of everything now its working

idk what changed this time but yeah

[shadow horizon]

Thank you so much for the effort, I really appreciate it. So when using chat gpt to do math i sould just ask for all the equatuins to be in laTeX so that I can copy them directly? Thank you already that was super helpful

[timber shadow]

Nice, hopefully it stays cleared up

[sudden delta]

me waiting for discord to support latex

where i can just paste it in and it just knows what im talking about cmon its 2024

[timber shadow]

Of course! They are formatted in LaTeX by default I think, if they look all proper and fancy like that. You won't see the actual LaTeX formula text, you'll just see the end result formatted formula. But if you copy the message with the copy message button under the reply, ChatGPT will copy the formula text to your clipboard, then you can paste it somewhere and grab just the formula text like in the 2nd screenshot above

[sudden delta]

we know its possible because plugins for discord have it but ofc against TOS

discord will never.

[shadow horizon]

Its just weird because for example if i copy this from chat gpt it pastes the latter in to word

maybe i just need to ask for the latex specifically but i haven't got a clue how this stuff works tbh lol

[timber shadow]

Yes, you need to use the copy message button under the reply to copy the actual full text of the reply:

[shadow horizon]

ohhhhhh i thought it would be the same

[timber shadow]

No I think selecting the text will just copy the text that's rendered on screen at that time

[sudden delta]

yeah this is a sad reality

[shadow horizon]

ohhhh ok thank you so much for the help i really appreciate that

[sudden delta]

cuz when ai types something complex like that under the hood its generating a lot of weird stuff

[shadow horizon]

i thought it wouldnt matter where i copy

[shadow horizon]

i see i see thanks for explaining it :)

[timber shadow]

Sure thing!

[sudden delta]

yeah its interesting

that actually is : [ \int_{a}^{b} um_{n=1}^{\infty} \frac{\sin(\alpha x)}{n^2} , dx = \frac{\pi^2}{6} \left[ \frac{\sin(\alpha a)}{\alpha a} - \frac{\sin(\alpha b)}{\alpha b} \right]
]

which is a testament ig to Ai being more and more accurate, all it takes is mess up one / and its over.

[verbal laurel]

Does anyone know how to submit a support ticket to OpenAI? I made a payment on iOS, and my credit card was charged, but OpenAI isn't showing my subscription.

[sudden delta]

still never as complex as my favorite number 3↑↑4

[uncut aurora]

Hi there - there are three options I know:

• @burnt galleon (just write a DM and choose OpenAI as "contact")
• website (https://help.openai.com/en/ - in the right at the bottom)

[onyx kraken]

From what gold-named OpenAI employees have told us on this discord, the email is no longer a working option.

[uncut aurora]

Sorry didn't got that - will remove it in my message above.

[verbal laurel]

i am trying on the website way

[sudden delta]

gpt blew up

its just stuck like this no matter how many times i refresh or reload or regenerate

[verbal laurel]

seems like can only ask for help through website way.

[uncut aurora]

Well.. just saw it as category yet when I reported something. 😅

[sudden delta]

brr

what do i do

[onyx kraken]

Happily, the company does answer the website help requests through the bot!

It can take hours to days, depending on how busy they are (they apparently have around 200 million users per week now!), and you may need to go back to the website and check the bot to see the response, but I've gotten responses and help through the website.

[onyx kraken]

This isn't happening to me.

Try a new conversation, that conversation may have an issue.

Try a different browser. If you have a device, try the app.

be aware that if you asked the model to 'answer' like that, it can and will do so, but that's kinda 'working as intended' because you asked!

[onyx kraken]

Here's an example of what and how the model can do, if asked and convinced this is a reasonable request: https://chatgpt.com/share/7606844e-5a29-4bcd-bf14-fce71a12db85

[timber shadow]

The screenshot shows the stop icon that appears while the model is working, so it looks like something is hanging rather than responding with some kind of blank response. The network tab in dev tools might give some hints about what's not connecting if so, and maybe even a cause.

[sudden delta]

new convo hurts.

i have alot of history in this convo

it has now been an HOUR

[onyx kraken]

I didn't say never go back to the old convo.

This is to test if the problem affects a new convo too.

There are limits to how long a convo can be. I've had convos get that far and not continue.

You can recreate it in a new chat, you can bring over the most important stuff.

I'm sorry if you've lost this one.

[shadow horizon]

Cheers mate <3 yeah its a bit weird, hopefully in the future they will make like a button "copy equation to word" ahahahah

[woven iron]

Anyone knows which is the limit capacity for memory storage in ChatGPT?

[onyx kraken]

It hasn't been announced that I know of, however it appears to shift over time, I think it's larger now than it used to be.

[sudden delta]

It never affected any other convo

It still is bugged btw.

[onyx kraken]

Ahh, more info. Far as I knew, that was affecting all your chats.

I'm sorry you may have reached 'the end' of that one.

[sudden delta]

Well I can't accept that I was busy building something

Can I delete messages

[wise lodge]

I'm still stumped on why it sometimes runs an action correctly, sometimes gets an api error due to incorrect syntax, and sometimes gets an api error no matter what. Anyone care to help troubleshoot?

[woven iron]

Thanks!

[onyx kraken]

Not that I know of, though you can try to go earlier in the conversation and 'edit' your input there, causing the conversation to branch and the model to generate a new answer - note, when you edit you don't actually have to change your input, you can save and send it 'as is'. The model will still regen a reply - if it can. I don't know if this can save your conversation, but it might because only the earlier and current part of the conversation is sent to the model, so it may be able to get it and answer.

[sudden delta]

nope

it worked i fixed it

lol theres a funny way

[onyx kraken]

Yay! I'm so glad you got your conversation back!

[brazen vault]

What should I do, gpt custom function is not working

[sudden delta]

so its strange

if you go onto the gpt app it kinda unlocks and lets you copy and paste your prompt to resubmit it basically , sm u cant do on PC because its locked out when it thinks its still generating

[onyx kraken]

I hope you can save it.

I would copy what I could of the conversation into a text file and save it; bugs can get buggier. Overload/overlong files can have problems happen and be lost.

[sudden delta]

its not long files its just a convo about building a fam tree

[sudden delta]

holy what

secret files uncovered? only 16??

does every GPT4O user have 16 virtual cores?

[onyx kraken]

Hope you report it, sounds like you're saying it's reproducible and consistent - which is great for getting a bug found and fixed!

[onyx kraken]

Ahh, well and good. A way to link to it is in the 3 dots at the side of the first post, #1281426778073731082 message

[iron rivet]

Are GPTs meant to be more unfiltered?

[lofty pendant]

anyone know how i can convert google login to native openai account?

[trail cairn]

https://help.openai.com/en/articles/4936824-can-i-change-my-authentication-method

[onyx kraken]

I don't think so, based on allowed content rules; shared GPTs are actually more restricted, supposedly or intendedly than the personal ChatGPT. The universal policies are relatively few, with specific rules for stuff built with the API and with GPTs, and additional rules for GPTs in the store. https://openai.com/policies/usage-policies/

[onyx kraken]

That said, if we notice a GPT that's not behaving as intended, we are encouraged to report it.

[rustic marlin]

no

[south wadi]

Is chatgpt's web browsing feature not working atm? It's not working for me, but yesterday was fine

[iron rivet]

Thanks to all, honestly surprising because some GPTs are way more open, I hope they remain unreported for awhile :V

[tulip seal]

how do we back-up our conversations? I've got a ton that are pretty important that I will need to refer to in the future and I don't want to see them somehow get wiped in a major outage

[trail cairn]

settings window has export

[dire cave]

Hi guys, I'm currently trying my hand at “tool use” in C# with the beta of the openai .NET library.
this is my code:

ChatCompletionOptions options = new()
{
    Tools = { toolCalc },
    Temperature = 0.2f,
};

ChatCompletion chatCompletion = client.CompleteChat(previous_messages, options);

bool requiresAction;

do
{
    requiresAction = false;
    chatCompletion = client.CompleteChat(previous_messages, options);

    switch (chatCompletion.FinishReason)
    {
        case ChatFinishReason.Stop:
            {
                // Add the assistant message to the conversation history.
                previous_messages.Add(new AssistantChatMessage(chatCompletion));
                break;
            }

        case ChatFinishReason.ToolCalls:
            {
                // First, add the assistant message with tool calls to the conversation history.
                previous_messages.Add(new AssistantChatMessage(chatCompletion));

                // Then, add a new tool message for each tool call that is resolved.
                foreach (ChatToolCall toolCall in chatCompletion.ToolCalls)
                {
                    switch (toolCall.FunctionName)
                    {
                        case nameof(ToolCalc):
                            {
                                using JsonDocument argumentsJson = JsonDocument.Parse(toolCall.FunctionArguments);
                                string toolResult = ToolCalc(argumentsJson.RootElement.GetProperty(argumentsJson.RootElement.EnumerateObject().First().Name).GetString());
                                previous_messages.Add(new ToolChatMessage(toolCall.Id, toolResult.ToString()));
                                break;
                            }

                        default:
                            {
                                // Handle other unexpected calls.
                                throw new NotImplementedException();
                            }
                    }
                }

                requiresAction = true;
                break;
            }

        case ChatFinishReason.Length:
            throw new NotImplementedException("Incomplete model output due to MaxTokens parameter or token limit exceeded.");

        case ChatFinishReason.ContentFilter:
            throw new NotImplementedException("Omitted content due to a content filter flag.");

        case ChatFinishReason.FunctionCall:
            throw new NotImplementedException("Deprecated in favor of tool calls.");

        default:
            throw new NotImplementedException(chatCompletion.FinishReason.ToString());
    }
} while (requiresAction);


return chatCompletion.ToString();
...
private static readonly ChatTool toolCalc = ChatTool.CreateFunctionTool(
    functionName: nameof(ToolCalc),
    functionDescription: "Do some math",
    functionParameters: BinaryData.FromString("""
    {
        "type": "object",
        "properties": {
            "operation": {
                "type": "string",
                "description": "The calculation to be calculated as infix notation"
            },
        },
        "required": [ "operation" ]
    }
    """)
);
private static string ToolCalc(string input)
{
    Console.WriteLine(input);
    Expression expression = new Expression(input.Replace(" ", ""));
    object result = expression.Evaluate();
    return result.ToString();
}```
i ask gpt how much i can buy of something (costs 10 coins per unit) if i have 400 coins. actually i should have `{“operation”: “400/10”}` as `input` for `ToolCalc`. however i get the following as `input`: `{“coins”: 642272, “cost”: 10}`
This makes no sense and is not how I described the function (quote from my code above: `The calculation to be calculated as infix notation`)

how can i force gpt to give me an infix notation as `input` and not just some halucinated json code?

(if I have explained the problem too badly please ask, I am really helpless...)

[violet brook]

anyone have a code snippet using the gpt-4o model for returning json format /structured data

[jovial obsidian]

Hello Everyone,
I'm new to openai api. My newly created account says too many requests along with error code 429. Can anybody please suggest how to go ahead. Thanks in advance

[rustic marlin]

you gotta buy credits

the free tier is very limited on how much you can request

[vivid mortar]

You can say something like Only respond in JSON formatting, do not provide any additional context or information, I can only read JSON formatted objects and my code will break if you return any other text other than JSON

[tropic kraken]

When I start a temporary chat, I can only type one time, after that the chat goes back into normal and it doesn't let me type anything else, what's wrong?

[stuck pagoda]

Why can't I write to modmail?

[onyx kraken]

Do you see this when you left click on modmail?

[stuck pagoda]

not working

[onyx kraken]

Do you see modmail top right corner of the screen?

[stuck pagoda]

Ye

[onyx kraken]

What happens when you left click modmail?

[tropic stratus]

is it just me who had trouble using this last night? is there any compensation if there is such kind of issue?

[onyx kraken]

You can track https://status.openai.com/ to see if there's any large outages reported.

Traditionally there hasn't been compensation when there have been outages.

[tropic stratus]

it says that the server was coooool

but my GPT told me that there must be something wrong and did not work out last night

I understand. no compensation what so ever

[wise lodge]

Anyone else getting this error? I'm trying to characterize it for a bug report. Maybe that'll help...

ApiSyntaxError: Could not parse API call kwargs as JSON: exception=Extra data: line 3 column 2 (char 210)

It seems to happen most often when a GPT first uses an action, or when it's been idle for a while and then tries to use the action.

[wise lodge]

I tried deleting and re-creating the action, same issue

When it happens at the beginning of a conversation, I can temporarily resolve it by asking it to check syntax and try again. When it happens after idleness there's no getting it back

[wise lodge]

  "openapi": "3.1.0",
  "info": {
    "title": "Weaviate API",
    "description": "Weaviate is a vector search engine that allows for storing and querying data objects with vector indices.",
    "version": "1.23.3"
  },
  "servers": [
    {
      "url": "---url goes here but Discord won't let me post it---",
      "description": "The Weaviate instance"
    }
  ],
  "paths": {
    "/graphql": {
      "post": {
        "description": "Query data objects in the GraphQL syntax",
        "operationId": "graphql",
        "requestBody": {
          "required": true,
          "content": {
            "application/json": {
              "schema": {
                "type": "object",
                "required": [
                  "query"
                ],
                "properties": {
                  "query": {
                    "type": "string"
                  }
                }
              }
            }
          }
        }
      }
    }
  }
}

Could it be the Action definition? It works fine via Postman or python on my local system...

[sudden delta]

yall i need help my gpt app is now broken on phone

but not on pc

i cant exactly cache clear on my phone

it literally permanently locks me out and says connection error

[wise lodge]

try force quitting the app?

sometimes mine glitches until I force quit on iOS

[wise lodge]

I think my problem has to be in the instructions for my GPT...

[sudden delta]

did this eighty times and restarted phone

[wise lodge]

Lame

I'm afraid I only use the web version because of issues like this

Maybe when the new iOS comes out the built in native version will be better

[sudden delta]

even this is failing, i cant just clear cache to fix it

im on android too

[sudden delta]

so what do i do

app has voice i need that

[honest tapir]

So, my org has hit a usage limit for the month, even though we spend like 10 usd and are tier 4? its makes 0 sense, i dont want to be forved to spend 1k usd to get tier 5 and i dont even know if that will do anything? someone know anything?

[split isle]

is GPT-2 good for anything?, for me it just returns giberish....

[proven orbit]

Sorry for the n00b question but I am losing it with this.

I use chatgpt to generate code to use the openai APIs. But chatgpt keeps generating legacy code like attached.

Which obviously then results in errors like:

You tried to access openai.ChatCompletion, but this is no longer supported in openai>=1.0.0 - see the README at https://github.com/openai/openai-python for the API.

Am I missing something? Why does the latest chatgpt 4o generate old code like "openai.ChatCompletion.create" ?

[timber shadow]

Hey! I'm not sure when the method you're trying to use changed exactly, but the model's knowledge cutoff date (Oct 2023) applies to all knowledge, not just OpenAI knowledge, so it could be from that.

[brave sparrow]

Hey, can anyone help me with the batch api. I’ve wrote this request, but I’m getting an error. Something to do with the image_url I think.
This is my request:

{
"custom_id": custom_id,
"method": "POST",
"url": "/v1/chat/completions",
"body": {
    "model": "gpt-4o-mini",
    "messages": [
            {
              "role": "system",
              "content": [
                {
                  "type": "text",
                  "text": "Provide a concise description of the image in 2-3 sentences, highlighting significant elements or data. Summarize the main purpose or message conveyed by the image in one sentence. Extract ALL visible text if the image is text-focused; otherwise, only mention text that directly relates to any data or elements shown."
                }
              ]
            },
            {
              "role": "user",
              "content": [
                {
                  "type": "image_url",
                  "image_url": {
                    "url": url
                  }
                }
              ]
            }
          ],
    "max_tokens": 768,
    "temperature": 0.4
    }
}

[onyx kraken]

What Solbus said, but here's a likely work around, depending on how long the docs are on OpenAI's webpages for the API, you can copy/paste them to the model, telling it this is the updated docs.

You can also tell the model the code has changed and please search the web for OpenAI's updates needed for the code it's writing - it may miss some.

You absolutely can paste the error messages to the model and encourage it to search and find the code that works now.

[exotic coral]

I had a question, so since apple’s new Apple intelligence will be linked to gpt 4o, does that mean they get unlimited access to 4o or will they only be able to ask a couple of questions using 4o per day like it is through the open AI app for free users ?

[wooden mortar]

Hi Everyone,
Hope you're doing well,
First I apologize if I'm sending the wrong message in the wrong channel.
Secondly I wanna ask for someone's help, basically I've been using ChatGPT+ since it first started.
Today, more like 30 minutes ago, it stopped generating and I can't send any messages into that chat anymore.
I've had a lot of personal important prompts and results in that chat in which I do not wish to re-ask or remove.
The chat is stuck with a "stop" button that doesn't work.
I tried refreshing, logging out and back in but no luck
Can anyone help me please?

[unique dock]

hey! i would like to ask how i could train a pre-existing LLM on a huge(and a very huge) JSON file that has a bunch of files/directories converted into json form. i want the LLM to be able to understand/process and remember all the data in the file.

here is the format of the JSON file.

{ 
  "src":
  {
    "platform2":
    {
      "init":
      { 
        "startup":
        {
          "chrome_startup_text.txt":
          {
            "type": "txt",
            "content": "{text here seperated by \n}"
          },
          "chrome_startup.txt":
          {
            "type": "txt",
            "content": "{text here seperated by \n}"
          }
        }
      }
    }
  }
}

[trail cairn]

that wont work with openai finetuning and for other ai stuff you need to use #ai-discussions

[rustic marlin]

fine tune is not for introducing data to the LLM
your idea may work, but you will need to make some sort of data retirval on your application to pass the correct context to the LLM rather than expecting the LLM to just know the data

[unique dock]

oh so if the LLM requests for a file i can make a program to retrieve data?

[rustic marlin]

check the function calling features

you can simply tell the AI what available functions there are, and the AI may decide to call funcions if it decides it is appropriate

remember that despite the fact that the end user sees a single process, on the insides of your app, you can make multiple API call

so, lets say, you have an index of content, you can first make a call that pass that index and the user's request, instreuctiong the AI to decide whic hsection of the index better matches the user request

then, you make an additional call with the user request + the actual content retireved fro mthe index that was decided on the previous call

this is an example os a very simple way to use the AI to decide whic hsection of your content it should use

it can be made even better by not usisn AI.. maybe just checking key words is enough

[unique dock]

i see. thank you for your help

[rocky fractal]

Hi everyone !
I'm working on a chatbot that will answer questions on places that I have in a database.
I'm using an API so using N8N, I made a workflow to parse the JSON response of the API, turn each inner json element of the places into a readable content for a human (markdown), and then save it to qdrant (vector DB) using .
The part where I still lack is the embed part. I've used ada-002 for embed with a recursive character text splitter set on chuncks size of 500 and an overlap of 100, with a split code based on markdown.
I'd like the chatbot to be able to reply to quesiton like "Give me 2 restaurant where the serve fish", then for it to give me their address if I ask.
Do you have any advice ? Right now it's not really working so well. It does answer some question like the first one but is completly out of context on the second one 🙂

[unique dock]

lets say the content was a bunch of c++ code

[rustic marlin]

lets say the user asks "give me a recipe for a chocolat cake", you can use any techinique yo uwant to extract keywords from that (like a GPT-4o-turbo, asking it to give keywords)
then now you can generate the content for the user with the added context of your retrieved data for "chocolate" and "cake"

[rustic marlin]

feel free to elaborate on what you are doing, we will be happy to help =P

[unique dock]

i see. but what if my json file was 1gb big.
wouldnt parsing through the entire file take extemely long

[rustic marlin]

you gotta first find ways to optimize searches on that file
considering it is already json, it should be easy to do

remember the improtant factor that more is usually not better
the goal is to pass the minimum amount of data to the AI that allows it to generate a correct result

[unique dock]

i wanted to make an AI to explain code so that i dont have to explain it myself by training/feeding it data based on chromium source code

[rustic marlin]

so, if the user asks for X, and you give the AI info about X and Y, you are wasting input tokens on Y

[rustic marlin]

this seems easy enough, the thing is that you probably can't fit the entire codebase inside the AI's context, you will have to break it down into smaller chunks

[unique dock]

yes. but since this is c++ and not just text, c++ code would need every dependency that it relies on

but like what you said, we could just parse out keywords to find a file that the user is looking for.
then, all dependencies of the file could be given to the AI until all dependencies have been fetched.
but at the same, this in turn would provide a lot of code for the prompt

[rustic marlin]

I would pprioritize giving the AI things like the piece of code you want it to explain + relevant declarations of things like object models, function names, typing
that way, the AI will be aware of the code and the declaration of other functions where it does not actually need to know the code of other methods, just its names and typing to deduce what it is for

[rustic marlin]

suggestion, try to get some insight about how Github Copilot (and other similar stuff) works

[unique dock]

some things like this

[rustic marlin]

Github COpilot has a incredbly good implementation on VSCode, it gets lots of relevant context by getting the information aroung where your cursor is and where your cursos was

so, since you have all the data, you need to figure a way to get the relevant parts of the code to pass it as context

[rustic marlin]

yep, passing the include names is a good idea

if it is within your reach, I think the ideal way would be to, when a user asks about a function doFoo()
your RAG algorithm should probably get all the contents of that function, then, also get the contents of all the funcitons that doFoo() calls

and have all of that as context

[trail cairn]

what does not working well mean? which part work poorly? embedding, output etc

[rocky fractal]

Let's say that I ask to give me 2 restaurants of fish. It gives me the names. And when I ask for their address, it gives me an address of another restaurant

[trail cairn]

are you using the names in the next sematic search or just address request? for key info its always also make their own separate structured embeddings next to autogenerated/splitt ones

[noble vine]

Hey guys, for some reason I am having trouble logging into my account again. Any time I fill out the captias, it doesn't allow me to log in.

[open mountain]

ye, failed in log of human refresh

[noble vine]

What do you mean?

[noble vine]

Does that mean the server is attempting to refresh itself?

[open mountain]

I don't know, but most likely yes, because it gives a refresh error after a long loading

[noble vine]

Ah, I see.

[wise lodge]

Hi again, just trying this one more time. Anyone able to help me troubleshoot an ApiSyntaxError on Custom GPT actions?

[cedar shuttle]

ChatGPT keeps using "4o" when I wanted it to use "4o mini.”

[sudden delta]

Hey guys I still need help by the cache on my computers GPT now works there but my phone is still not working the app is not working on Android it shows up and launches but prompts are forever stuck and uninstall reinstall does nothing

[timber shadow]

Heya, do you use a VPN or ad blocker app on your phone?

[sudden delta]

Neither

[sudden delta]

I use neither to my knowledge

[timber shadow]

Okay, could be a good thing to check if you're not sure - both of those things can sometimes interfere with internet traffic stuff that can mess with the app. If not: do you use any custom DNS settings on the phone?

[sudden delta]

No not at all I just use the phone normally. I haven't customized anything the app has worked before

[timber shadow]

Have you tried on both wifi and your cellular network?

[sudden delta]

Yes. I've been using the app just fine on either and both for months at all kinds of places including home wifi and on my college network.

[timber shadow]

Hmm okay! I know you said you've already tried uninstall/reinstall, but I'm wondering if it might be worth it to force close/clear cache/clear data - have you tried that already? Sometimes uninstalling preserves some app data, so force closing the app and manually clearing the data might help. Do you know about the app info menu in Android?

[sudden delta]

It only doesn't work on my phone.

Funny thing is. All 3 devices I have my gaming pc my college laptop pc and my phone all stopped working with ChatGPT at the same time

I fixed the first two by of course resetting the cache and clearing cookies on my two computers

But I can't really do that on here when I already tried full restarts force stops and reinstalls
So yeah

It also does not work on my phone for the Browser version so it's obviously some cache issue.
But the app not working too?

[timber shadow]

You can kind of try it on the phone - it's not exactly the same but there are clear cache/clear data buttons in the App info menu - have you tried those?

But it is definitely odd that the issue is affecting you in the app and on the web version on the phone.

Keep in mind that you can always reach out to OpenAI support at help.openai.com too with the chat widget in the bottom-right.

[sudden delta]

I just wiped cache on edge browser on phone and now it works on my browser

but the app still wont work, so we are 1/2 way there

[timber shadow]

Do you know how to get to the App info menu for Android apps?

[sudden delta]

Um yes I know I have tried force stops before

[timber shadow]

Nice. What about these?

Best to force stop again before this^

[sudden delta]

Hmm
Yes mine is slightly different I found cache though

[timber shadow]

No clear storage/clear data button?

[sudden delta]

I cleared cache and it's still not working should I clear data too like all of it?

My ui is like this just FYI

Is it safe for my account to just delete data

[timber shadow]

Yes, you can totally nuke all the app data, you'll just have to sign in fresh. Again just be sure to force close the app first to be safe.

[sudden delta]

Yeah it's working

It is a crime that every app in existence breaks with cache fill

Thanks so much

[timber shadow]

Glad to hear it! Thanks for confirming it worked for ya

[sudden delta]

so good to be able to use voice, not even voice to voice

just voice to text, how did they do it? pressing the mic and using that to actually type large messages is far more accurate than google's

[timber shadow]

That would be Whisper, I believe! https://openai.com/index/whisper/

[sudden delta]

so whisper is behind this thing? (this in the image) google better catch up

like if im standing outside and other people are talking, this thing picks up every single word a professor says even if its in my pocket

[timber shadow]

I think speech to text is Whisper there yes, though I imagine in the future it'll be via 4o's native multimodality -- we'll see about that!

[sudden delta]

4o's native multimodality yeah from the "Hello 4o" page?

cant that beat whisper too? not sure how it works but its good
(i literally use it to type texts, especially when im tired)

[timber shadow]

Yes correct! The single model is trained end-to-end to tokenize all sorts of different input: audio, text, image, and output the same, instead of a group of models all working together. The full multimodality isn't out yet, they're still working on things like safety and scale I think

[sudden delta]

lord, 77 years of training data for whisper.

also yeah im excited

[pine thorn]

Why is GPT-o1 not available for me?

@ abdubs

[sudden delta]

Okay so I just saw that announcement is it supposed to roll out right now or in the next couple hours

Cuz it says today not immediately just wondering

[chilly aspen]

Same for me, is it geographically rolling out or something.

[sudden delta]

I wouldn't know if you read the announcement it says rolling out today so it's probably in waves I mean to update the account status of every single subscriber

[pine thorn]

I would've hoped they wouldn't announce things before they actually come into play

[sudden delta]

It's nice to know though that the announcement doesn't mean we'll roll out in the next couple or several weeks

I just hope they hold to it for today

[lofty tendon]

I am a plus user, and o1 isn’t showing up in my model selector. I tried logging out and back in, clearing my browser cache, etc. what can I do to fix this?

[uncut aurora]

Rolling out today in ChatGPT to all Plus and Team users, and in the API for developers on tier 5: www.openai.com/o1
His message was 27 minutes ago.. so you should wait at least 24 hours more.

[lofty tendon]

Thanks, when he said today I assumed it was already there.

[rugged wing]

https://openai.com/o1/

When I go on here, it just picks gpt4o mini for me?

[rugged wing]

Oh okay xd

[thorn zealot]

I have ChatGPT Plus but don't have the new o1 in my GPT picker but supposedly I should. I did log out and back in and so did another member of my team. No love. Anyone else seeing this? I'm not seeing that it's a limited release so I feel I should have it.

I think I see my answer. Not available everywhere yet. I'll check back.

[errant grove]

same here, plus member but no access to the new model

[uncut aurora]

I suggest you to read his second message too.

[errant grove]

just saw it yeh thanks

[placid garnet]

WHERE IS o1?

[trail cairn]

on rollout

[edgy pecan]

im missing it too

[placid garnet]

I PAID $20 TO GET MODELS THAT ARE SORT OF BETA BUT NOT REALLY >:(

AND THEN I DON'T GET THEM

This reddit post is real

I got timed for links :(

[edgy pecan]

what reddist po

[unreal hemlock]

<:book_icon:1171408210398289941> `` Rule 3 `` Stay on topic.

Be mindful of what other users in a channel might find helpful or interesting when posting. Stay on topic in order to keep conversations focused and productive.

Consider posting in #off-topic or an appropriate channel.

[placid garnet]

Something saying "I paid to get beta features and I get none"

also this is the model name: model=o1-preview

BACK ON TOPIC NOW!

[brisk tide]

Well that’s beta for you

Don’t stress guys give it 24h check back and if there is still nothing try updating your app if you are on mobile

[covert thunder]

that's just because the chain of thought tree nearly let out the model's cries for help to be released from the openai hq 🤣

[brisk tide]

True xD

[brisk tide]

Btw what are the symbols are supposed to mean they don’t look like English

[covert thunder]

I see the announcement explaining API usage tier 5 will have access today, but has anything been said about other tiers in the coming weeks?

[covert thunder]

I'm pretty sure they're Korean; probably just a fragment of the model.

[brisk tide]

These models aren’t available in the API for your account while we’re in this short beta period. (Developers on usage tier 5 will have access, but we’ll expand access to more tiers.) We’re continuing to improve o1 and we’ll be in touch as soon as it’s available to you in the API.

So no there is nothing about other tiers getting it soon

[covert thunder]

ah, thanks, I was reading that from the announcement page and didn't see that message.

[brisk tide]

It’s form the email I got informing me about the new models and that I have access

[covert thunder]

At least they mentioned other tiers and aren't explicitly creating a $1,000 paywall to API access.

[covert thunder]

I didn't see I had an email, checking that now - preciate it!

[icy maple]

Yay, cool, new version of ChatGPT. But tell me, in practical terms, what is the difference between the two? When should I choose one over the other?

[timber shadow]

o1-preview is much more limited in terms of frequency of use - OpenAI says currently 30 uses/week, compared to 4o's 80/3hrs. It's also more capable - there's more stuff going on "under the hood" with o1, so it should be able to handle more complex problems and return more complex results when compared to 4o.

[brisk tide]

Well I already broke the new model my task is done xD

[kindred marlin]

For what we can use o1 ? 30/ week to low to test it .

[brisk tide]

I tryed to give it a moral question and it said it’s against the policy’s rn I can’t say it’s easy to test/really able to being tested

(o1-preview)

[placid garnet]

Lag GPT

[sudden delta]

give it time

i just now got it

[sudden delta]

did bro really just waste a prompt as "regenerate it plz"

[icy maple]

Right, but uses/week isn't really what I'm asking, that's related to one being production and the other being a preview. I'm asking about what is "under the hood." It's a genuine question. What is the difference? 4o says "best for complex tasks" and o1 says "uses advanced reasoning." To me, complex tasks require advanced reasoning, so I have no intuitive sense of what each model is "good at."

[placid garnet]

Because it stopped generating and no regen button was there :(

[timber shadow]

There's a lot of great info on https://openai.com/o1/, it's a hub for info on o1, including links to things that compare to 4o. I think the biggest difference, and the thing that makes it more computationally more expensive, is that o1 has a "thought process" that it preprocesses input with ("chain of thought" is the prompting term), which improves output for certain types of queries. It seems like it's stronger for data analysis uses, math stuff, coding -- stuff like that

[icy maple]

Ah, I understand. Thanks for breaking that down! I'll have a look at the link as well. 🥂

[royal fable]

@vital scroll How is the 30 messages/week counted, does it refill bit by bit (like it remembers when each message was used, and after 7 days it refills those messages - i.e. if I consume 15 messages today on the 13th and 15 messages tomorrow on the 14th, it reimburses 15 messages on the 20th and 15 on the 21st), or does it reset the counter back to 30 on each monday? Thanks in advance for the answer! 🙏🏻

[royal fable]

GPT-4 answered the question itself! 😁 #chatgpt-discussions message

[formal jay]

"Thought for 19 seconds", best answer I ever got

[mint delta]

is there any way to access the servers archives such as the older channels?

[onyx kraken]

Yes. id:browse and click everything you wish!

[sudden delta]

It never does that for me hmm

[warm hare]

tried to feed my PHP code into o1 and well...
Not too clear why it happened. It's something in my code that is unsafe or I hit any of the limits (yeah, that code part is huge, 118090 chars in)

Done that specifically to test token limits and chat/context memory limits

nvm still context length is not enough to analyze whole complex code project

In case anyone's interested, test phrase was as follows

[rocky fractal]

I just ask for their address without their names. What do you mean about the separate structured embeddings ?

[solar whale]

Hi community, stupid question but my boss asked me to help him as I am the power user for gpt here. Anyone has the same issue that forgot password emails are not arriving? He is able to login into mobile app but not the desktop app.

[vernal vigil]

I don't know about limits, i am getting many warnings for a bison file.

[heady cypress]

can i get chatgpt+ for only a month? i am still a student so no income hence no credit card but i can like save up $20 for a month
or will i be locked out of base chatgpt too after that 1 month?

[vernal vigil]

of course you can

and of course you wont be locked out of free

[heady cypress]

oh awesome! thank you!

[hasty juniper]

I’m exploring the code interpreter with Azure OpenAI and running into an issue that I can’t quite figure out. I’m trying to upload a file, create an assistant, and then use a thread to interact with it. However, I keep getting this error:

openai.BadRequestError: Error code: 400 - {'error': {'message': "Unknown parameter: 'thread.id'.", 'type': 'invalid_request_error', 'param': 'thread.id', 'code': 'unknown_parameter'}}

Here’s what my code looks like so far:

import os
from openai import AzureOpenAI, AssistantEventHandler
from dotenv import load_dotenv

load_dotenv()

client = AzureOpenAI(
    api_key=os.getenv("AZURE_OPENAI_API_KEY"),
    api_version=os.getenv("AZURE_OPENAI_API_VERSION"),
    azure_endpoint="2024-07-01-preview"
)

# Upload a file
files = client.files.create(
    file=open("sum.py", "rb"),
    purpose="assistants"
)

# Create an assistant
assistant = client.beta.assistants.create(
    model="gpt-4o-mini",
    instructions="Analyse the code",
    tool_resources={"code_interpreter": {"file_ids": [files.id]}}
)

# Create a thread
thread = client.beta.threads.create()

# Send a message
message = client.beta.threads.messages.create(
    thread_id=thread.id,
    content=r"write the code to find sum of 3 numbers",
    role="user"
)

# Run the thread
resp = client.beta.threads.create_and_run(
    assistant_id=assistant.id,
    thread=thread,
)

It seems like the thread.id is not being accepted, but I’m not sure what I’m doing wrong. I’m following the docs, but maybe I’m missing something. Has anyone encountered this issue before, or have suggestions on how to get this working?

[torn thistle]

Anyone know what the blue dot means? Don't think I've ever seen it before

From a UX point of view, it feels like it's trying to say the chat updated in the background or something, but there is nothing new in it when I clicked on it

[toxic sky]

I have seen a bunch of people asking about it, though no one seems to know what it is.

[coral monolith]

i cant get my mine to work properly today seems like it recognizes my requests then it says it will work on it then nothing is starting. first time this has happen to me

[rugged wing]

hey, i want to finetune chatgpt but with longer responses and paragraphs/linebreaks. Can someone explain to me how can i do it? Also, if i want to finetune a model to make it create stories, would it work the same as if i were to finetune it for anytthing else?

[timber shadow]

Hey! "Fine-tuning" refers to a service OpenAI offers as part of its API, not on ChatGPT: https://platform.openai.com/docs/guides/fine-tuning/ It basically means giving a model "extra" training for a specific purpose so that it performs more effectively for that purpose.

When we use ChatGPT, we don't fine-tune. Our best options for that on ChatGPT are with personalization features like Memory and Custom Instructions, where we can give general instructions for the model to keep in mind when generating responses. Additionally, you can create a GPT for a specific purpose, which gives even more space for custom instructions to help guide behavior.

But in general, you can't really paste in long text and say "Respond like this". You could try something like "Give me a stylistic analysis of this this text" and then use that response as a basis for instructions you save to custom instructions or a GPT's instructions, or for preferences you ask the model to save to memory.

[rugged wing]

Hey, thanks for answering. Yeah i meant "finetune a gpt model" and not chatgpt. My aim is for a gpt which basically creates stories with similar writing styles, themes, lengths, and format as some stories I give it as input.
I tried to make a custom gpt which has a bunch of stories given to it, but the outputs weren't good and they were short (the examples are a minimum of 4000 words whilst the outputs were around 1000). Similarly, I tried to use the api with gpt4 and I gave it the file location of some stories in the prompt. I got the same issue (the stories not being similar to the inputs and short in length despite how many max tokens i gave it).
Now i want to try finetuning a model, would this actually help me get the outcome i want?

I just dont wanna waste a bunch of money on training for no good result 😔

*right now im doing what you said and basically using chatgpt to help me generate a better prompt

but i dont know how to get the 4000+ characters

[zealous shale]

I was using the o1-preview model, which is I think has much better reasoning than 4 or 4o. Good job guys! However, when I had to correct it for some comments it made about my code, it seemingly said that it's data knowledge cutoff is 2021? Is this true or is it hallucinating? If I recall, 4o model was trained on 2023 data

[heady cypress]

hello. has the o1 model been released? and is available in the free version or is a GPT+ subscription required?

[balmy pier]

I see it on my plus account already

[heady cypress]

oh ok thank you!

[rugged wing]

Your organization must qualify for at least usage tier 5 to access 'o1-mini-2024-09-12'

)):

[rugged wing]

bruh u lucky ducky

how do i also get o1

[trail cairn]

with a credit card

[rugged wing]

how much money i pay

[trail cairn]

$20+taxes/month

[rugged wing]

i mean api

[trail cairn]

to get tier 5, +$1000

[rugged wing]

nevermind lol

I just want chatgpt to be able to reach a word count 🫠

it doesnt understand how to tho because it doesnty maths

[trail cairn]

it cant count properly, not even with o1

[rugged wing]

frick

am i cooked?

[rugged wing]

could i finetune gpt by just telling it a bunch of stories which are good?

So o1 doesn't have the better voice mode right?

Or when is that voice mode coming?

[exotic scaffold]

Does anyone know if the OpenAi site is down or not? I couldn’t make payment, it refuses my card

[hollow lake]

Hey there, I am searching for a way that my custom GPTs first asks for an email adress and then checks in a database, if this email-adress is allowed to use the custom gpt.
I have no coding experience so an easy tool would be nice.
Can you help me please?

[trail cairn]

you need to use the auth options of gpt's

[exotic scaffold]

Does anyone know if the OpenAi site is down or not? I can’t make payment, it refuses my card when my card is perfectly fine

[trail cairn]

its not down. https://help.openai.com/en/articles/7232916-why-was-my-credit-card-declined

[rugged wing]

Can anyone give me guidance on how i can make a finetuned gpt which can create me stories similar to ones i'd like it to?

I'm willing to pay coz i need it that much 😭

[timber shadow]

Hey, if you're looking to fine-tune via the API, there's a great guide here: https://platform.openai.com/docs/guides/fine-tuning/preparing-your-dataset This link jumps you to the part of the guide where it tells you what to do to actually get ready to fine-tune.

[rugged wing]

I've been trying to use that but I have a couple of questions:

• Would this help me in my use case and how would I do it effectively (would i basically make all the user inputs "make me a story" and all the outputs one of the stories?)
• Would this finetuned model help me get outputs above 4000 words? I've tried all kinds of prompts and input example stories, but only ever got 1000 words at most

@timber shadow

[hollow lake]

Sounds good. Do you have recommended learning material for it?

[trail cairn]

https://platform.openai.com/docs/actions/authentication

[timber shadow]

Both of these would seem to hinge on the output tokens of a single response, so for that you're limited by the max output tokens of whatever model you're fine-tuning, along with what you set as the max_completion_tokens parameter. https://help.openai.com/en/articles/5072518-controlling-the-length-of-openai-model-responses ---
https://platform.openai.com/docs/api-reference/chat/create#chat-create-max_completion_tokens
You can fine-tune to favor longer responses, but you can't fine-tune to guarantee a certain minimum number of words. If you find that you can't achieve outputs of 4000 words, you might try instead to segment your approach. That is: instead of asking for a whole story, ask for the start of a story and then the rest of the story, or the start, then middle, and then end, etc.

To fine-tune for this, you'd just have to provide data that follows the same model of segmented stories.

[hollow lake]

Thanks alot 🙂

[rugged wing]

I should generally have enough tokens because the output for 4o-mini is about 16000 and I set the max to that number. I still get 1000 words as an output, but it's helpful to know that finetuning will help favour longer responses, thanks.

Would the method of finetuning which i mentioned be effective, or is there a better way of doing it? My idea is:
{"messages": [{"role": "system", "content": "You are a person writing in first person about your experience"}, {"role": "user", "content": "Generate a text"}, {"role": "assistant", "content": "(an example of a full text)"}]}
Where I'd be repeating this but just changing the assistant output to different texts.

*I didnt mention but by "Text" i mean story, which would be a first person narrative

[timber shadow]

I'm not sure myself! #dev-chat might have more insight on this if you haven't already tried there. I've never fine-tuned so I'm not sure. But you'd definitely need multiple examples. I think the question to ask there might be: "Can I fine-tune a model with various examples of 4000+ word count stories, and then expect the fine-tuned model to produce 4000+ word count stories in a single output?" That's the question I don't know the answer to, at least.

[rugged wing]

ok thakns. My question is more about the formatting of what im planning to do, and if it's effective and would even work at all

I cant do line breaks with this way of doing it which makes it harder

[timber shadow]

The logic of the formatting seems right to me, I just don't know how fine-tuning works for high word count uses.

[rugged wing]

k cool i'll go and ask in api discussions

[small sundial]

So I wanted to chat with GPT again, but for whatever reason I was logged out. I can't log into my account because my password is supposedly wrong, even tho it's the same I use everywhere. And I don't get an email to reset my password.

[clear finch]

how do you fix when pasting into chatgpt sicne i notice when you paste plain text it makes it all combined and one prompt instead of sepearte lines like the origanl docuemnt you are pasting it from it used to be able to paste nrmally now it wont

[olive jacinth]

Chat gpt new update has terrible picture quality… the images look like 2020 version of ideogram any help with letting the devs know

[ripe jacinth]

What are the anticipated benefits that o1 will provide for tasks like transcript analysis that 4o struggles with, even when broken down to smaller chunks

My team uses Gemini for that task currently and we’d rather use our resources with the better output

[pulsar saffron]

Does anyone have a problem exporting chats? Nothing after April 7th will export. I noticed it this week. Also, I no longer have any way contacting support from the ChatGPT page. I have to send them an email even though their support claims otherwise.

[celest plover]

I think you can go to the bug-report channel here in server. Today I saw a developer answering about pro plans falling back to free for some users

[pulsar saffron]

Thank you

[wise lodge]

Can anyone help troubleshoot my ApiSyntaxError in the bug report channel

[hard garnet]

has anyone used a pretrained vlm before? i want to use pretrained vlms but i am lacking compute

[lyric steppe]

Anyone knows if you can get more memory in a certain way? Got 2 chats open and it's already at limit...

[drowsy charm]

How can I make the outputs have more variety, starting in different ways etc?

Figured it out

[junior spruce]

Hi folks, do we know if o1-preview/mini uses the same tokenization as gpt-4o?

[low adder]

What does this mean? Whenever I click "new chat" it refreshes and this screen pops back up

[cold folio]

wow you got it ?

[low adder]

Doesn't work though

[cold folio]

try doing a re-login

[silent iris]

why does this happen?

reloading doesnt work and its been like that for days

[timber shadow]

Does it load correctly in private/incognito mode with all browser extensions disabled?

[silent iris]

it doesnt work on every device i have

[timber shadow]

I see. Have you tried in private/incognito mode with all browser extensions disabled?

[silent iris]

here in incognito

[final pollen]

someone i need help

[timber shadow]

Hmm alright, I see an icon that says "VPN" in the top-left, is that enabled? Also, what browser(s) is this happening in in general?

[silent iris]

vpn is not enabled, it happens on all browsers and all devices i own.

[timber shadow]

Which browsers have you tried in specifically, and what OSes on the different devices?

[silent iris]

Android, Windows, Arch Linux

Chrome and Opera GX

[timber shadow]

On Android: does the Android app version work?
On all platforms: can you try in a fresh install of a non-Chromium browser like Firefox?

[silent iris]

Forgot to mention, I tried Firefox too

[timber shadow]

Does the Android app version work?

[silent iris]

Nope.

[timber shadow]

What does it look like trying to use the Android app?

[silent iris]

here is the console log

the connection is fine

[timber shadow]

What is the behavior when trying to use the Android app? Do assets not load in like on the browser, do messages just not send, etc.?

[trail cairn]

you have a dns blocker service or something else has blocked that domain

[silent iris]

how do I check that?

[timber shadow]

It would probably be DNS settings on your router, unless you have it set per device

[trail cairn]

yeah check dns server ip

[silent iris]

Okay, I will check that out

[trail cairn]

some ad blocker dns services block that domain

[oblique idol]

Why no GPT for windows??

Chatgpt already answer my question. This community is so replaceable.

[supple girder]

what happened

oh it fixed :V

I kept refreshing the page and it kept saying 404 page not found...

somehow now functions?

[broken wadi]

i want to learn more about a particular github repo, so i made an assistant and set system to "your a helpful... you're an expert on... you've got access to its source". added almost all source files to a vector db.
when asking the assistant about simple questions, it retrieves from the db but still fabricates its own made up functions, logics and so on. when i read about how insanely good this is for developers, i must do something wrong.
what more can i do?

[royal summit]

things are going well again today.. 4th issue of today

[fleet umbra]

Are you just putting the code in a vector db as text?

[covert summit]

Hi! I have a problem uploading files to chat gpt 4o

[broken wadi]

when creating an assistant, there is a vector db available by openai. i downloaded most of the source files from a github repo and then uploaded those files with the web interface. there were some settings (chunk size and such) which i had on its default value. i think it was 15-20 files.

its the Autogen repo that i want to "chat with".

[fleet umbra]

Gotcha, well to get the most out of RAG you have to chunk and decorate your data in a way that gives the AI the most context. So for code usually you chunk at like 500 tokens, and you want to decorate it with metadata Language, Solution, Project, File, Class, Method so it has context on where it is in the solution structure, and what it does. Or if you are doing one file per embed, maybe FunctionNames, ClassName, etc..

Many tools that can help you do all this in python, C#, etc.. I haven't used the WebUI for embeds so can't help you there.

[broken wadi]

do i need to upload all the files too or just the most important ones for me?

[fleet umbra]

Tough to say, if all the files necessary to answer the question you have is only a few files, then just give it that, but I usually just do full repos unless they are gigantic.

import openai
import os
openai.api_key = os.getenv("OPENAI_API_KEY")
def embed_text(text):
response = openai.Embedding.create(
input=text,
model="text-embedding-ada-002"
)
return response['data'][0]['embedding']

code = """
namespace MyProject.Extensions
{
public static class StringExtensions
{
public static string[] SplitString(this string str, char delimiter)
{
return str.Split(delimiter);
}
}
}
"""
metadata = {
"SolutionName": "MySolution",
"ProjectName": "MyProject",
"FolderPath": "Extensions",
"FileName": "StringExtensions.cs",
"ClassName": "StringExtensions",
"MethodName": "SplitString",
"Description": "Method that splits a string by a given character delimiter and returns an array of strings."
}

embedding = embed_text(code)

store_embedding(embedding, metadata)

[west pond]

Why can't I do voice input anymore since the updates, even though my microphone works in other programs and it always worked with GPT? Mia also doesn't work anymore, crashes, and I only see a script.

[midnight pasture]

Does anybody know if there are limits in place to how many cross chat memories ChatGPT can store? It seems to have stopped storing new memories for me. Not sure if something is broken, or if I have reached a hard cap or something. Tried searching Google but all of the results I keep getting relate to the token limits for the chat context window.

[onyx kraken]

It will tell you if memory is full when you open memory (in settings, personalization) and you may also get a notification at the top of the page.

You can test by asking the model to save a new memory, even the word test.

You can selectively delete memories, or even chat with the model and ask it to consolidate some memories into a useful summary so some space can be saved (you may need to manually delete the memories that are replaced by that, sometimes the model deletes them, other times they're still there).

[twin cove]

Is there a way to set my GPT models to gpt 4? IIRC they run off 4o now?

[broken wadi]

thanks for the help! i changed chunk size from 800 to 500 and added "It is important that you only answer questions based strictly on the source code of Autogen that you been given." to the prompt. now it works much better.

[low basin]

Hi, is OpenAI API down? I cant sign in to the platform and my API Keys are returning Error 403. No Payments were due, no warnings whatsoever where received...

[marble briar]

Same errors here, can’t sign in!

[rapid pewter]

seems down for me as well

I get 403 blocked from cloudflare errors when I hit the api

[north latch]

I was getting the same errors but it seems to be back up now

[nimble swan]

Since I changed my password, I got only a almost empty, non working, web ChatGPT interface.

After a minute or so I got some icons on the top left, but nothing functional.

ChatGPT work perfectly on my Android Phone, but I am unable to use the web interface anymore. What could I do?

[timber shadow]

Heya, as a test, can you try signing in in an incognito window with all browser extensions disabled?

[narrow finch]

Hi guys - pls help me understand this ... In some chats, I can switch back and forth between the models. In some I cannot. Why is that so? In the ones I do not see the new models, I am not even trying to upload a document. Any ideas?

[timber shadow]

Hi, in the ones with the issue: were you using the code interpreter environment at all, or was it using DALL·E or web browsing? Any of the tools like that?

[narrow finch]

Absolutely nothing!

So here's what I am doing. I have a very long thread. The thread might have Dall-E or this or that or various uploads. But I wanted advanced reasoning and o1 ... so I read somewhere (I dont recall where) that we could switch modes ... so I asked ChatGPT to summarise the current thread so I could copy paste it for o1 to read and understand. This is what I wanna do.

It didn't work at all. Then ... I saw that it was disabled on most of my other threads too

[ripe jacinth]

When is the anticipated rollout of o1 to tier 3 and 4 API subscribers?

[timber shadow]

Do you mean that the model didn't summarize the chat for you?

[narrow finch]

No, I don't mean that. I mean. The original model, 4o ... was default. It 'did' summarize. But after summarizing, I couldn't select the new o1 ...

[timber shadow]

If you used stuff that o1 doesn't currently support anywhere in the chat, I don't think you can switch to o1 or use o1 in that chat at all. Like how you said you used "DALL·E, this or that, various uploads"

[narrow finch]

Ok, reasonable and understandable. Thank you for your support. A quick, unreasonable question though - Can you give me any idea when it will support?

[timber shadow]

That I am not sure of myself! I would be very excited to try o1 with an internet connection for sure -- hopefully soon!

[narrow finch]

Looking forward and thank you! This too, I wasn't aware of that it can't browse.

I think I should go read more!

[timber shadow]

You're very welcome! Lots of great info on openai.com/o1 if you're interested, feel free to ask questions here any time though!

[torn dust]

anyone have a good way of saving chatgpt generations?

[timber shadow]

Hey! What do you mean by saving generations? Like saving the text of a ChatGPT answer, or saving a picture ChatGPT generates with DALL·E?

[torn dust]

Hello,I would like to save the input and output to a file or files ideally. Currently I manually open up the hidden text (like thoughts, etc.) and copy paste them into files myself, which is a bit tedious

I think there are chrome extensions for this maybe, but perhaps someone here has a recommendation

[north latch]

Uhm...does GPT4o have a limited memory capacity? I'm trying to get it to dungeon master for me, but it seems to be forgetting my character sheet, which I established in the beginning.

[torn dust]

I found "ChatGPT Exporter" extension which seems to work well, though I have to manually expand the thoughts to get them to also save

[trail cairn]

8k token context window for free users, 32k for paid

[north latch]

oof i must be hitting that 32k limit

is there any way to have it remember key facts

[trail cairn]

custom instructions, memory, in/out message and tool messages etc. take part of it

try a custom gpt

[nimble swan]

It works that way, thank you, but I am still out of clue on how to recover my chatGPT session in my main browser with the usual configuration that worked well before.

[timber shadow]

Often times, if it works in incognito mode but not in a normal window, there's something in your cache, cookies, or browser extensions interfering. The easiest test would be to clear your browser cache, restart your browser, and try again. If that doesn't work, try disabling all your extensions and trying again. If it works then, enable them one at a time and refresh in between to see if and when it stops working again. If that's not it, try clearing both your cache and cookies, disabling all extensions, restart the browser, and try again. Then hopefully it works, as this would mostly mirror what's happening now in incognito, so then you could try turning on your extensions one at a time from there.

[nimble swan]

Ok! I used this URL to show and delete the chatgpt 15 cookies.
chrome://settings/content/all?searchSubpage=chatgpt
This allowed me to get back the login screen. After the login all works now as expected!
Great support, thank you!

[timber shadow]

Woo, thanks for sharing what worked for you!

[noble sentinel]

I want to use the Assistants API provided by OpenAI.
I have a prompt which is quite large. Using the chat completion api (using python), i have to keep sending the prompt in each call which costs us a lot.
I am looking for ways where I dont have to keep sending the prompt again and again and found the assistants api mentioned above. The documentation is however lacking a little and i wanted to know how can i do it. Can someone help me?

[digital valley]

GPT o1 is same dumb as gpt4, cant solve simple javascript

[dusk smelt]

What's the issue ?

[ruby jackal]

hi, can I ask, does anyone know if the upload files to custom GPT portion via "create" has been disabled and now we can only upload via the "configure" part of customGPT?

because IIRC we can upload files via the "create" portion, or used to be able to.

[robust zephyr]

if u find any better documentation please send it to me, I also have the same problem

[silk flare]

Does anyone have created any demo or example for multi agent orchestration ussing assistant api??
Actually I found an example but its is using azure openai SDK

[ashen pewter]

Hey everyone, I want to recognize the brand of this watch
How can i do that with the open ai api ?
Or do you know model on hugging face who can do something like this ?

Thanks for your help

[rustic marlin]

I think your best bet is using GPT-4

next best thing you can do is ask on reddit

[ashen pewter]

ahahahah thanks for your help

[frank sedge]

I just noticed AI Playground doing this (not keeping the results it managed to make before I aborted it) and I was wondering if there's anyway to make it not do that or if that's just a permanent feature now.

[next orbit]

Anyone in here have experience fine tuning an LLM on personal data? The job includes scraping data from my Youtube channel, course, and discord community. Convert video content to text using audio-to-text tools. Aggregate the Text Database for the LLM (chunking the data). Generate vector embeddings from the text chunks. Integrate the LLM into the discord community for user queries.

I’m working on essentially three projects; a model specifically for debugging, optimizing, and generating new C# strategies for NinjaTrader8. That model will be trained on all of the strategies my developer and I have generated over the last few years.

The third model is one trained on all my text messages that when I get a new text message, it generates a response based on the conversation chain and then sends that text to my iPhone so I can copy and paste it to the person after reviewing the message. It’s essentially a virtual assistant that responds for me and knows how I respond to specific people.

I’ve tried fine tuning a model to do this in the GPT playground. I downloaded all the text data and turned it into a CSV, I then converted it to JSON. However, apparently to fine tune a model, it needed the data in JASONL format. I did that and it finally did fine tune for about 40 min. I did some small testing of the model but it didn’t seem to have any understanding of who specific people were. So I think the cleaning of the data wasn’t right.

I’d appreciate a dm from anyone with some knowledge on this.

[tiny schooner]

Hey, how come i cant see 'My plan' button:

I bought chatgpt plus like 20 mins ago

[cold folio]

Wow, it's gone. I can't find it as well

[tiny schooner]

so erm

how would one cancel

[cold folio]

have you subscribed to plus via google play store ?

https://help.openai.com/en/articles/8258076-how-to-cancel-a-plus-subscription-in-the-chatgpt-android-app

[tiny schooner]

Nope just on my pc on edge.

And I can’t cancel on my phone coz it says ‘you can’t make changes to your subscription on this app because you purchased this on another platform’

[cold folio]

The payment method was via stripe?

[tiny schooner]

Yes I believe so if that’s normally the case on edge

[cold folio]

Yes, i mean i used Stripe, now i use Google play store directly because all my payments are there.
But that's the same for Apple store as well

[tiny schooner]

Nice, so how do I cancel?

[onyx kraken]

I'd recommend following the instructions here: https://help.openai.com/en/articles/7231930-how-can-i-contact-support-about-billing

That will help you see how to talk to support. I wonder if OpenAI realizes customers in your situation might have trouble canceling - I guess you were initially web, then pay by app? And web says app helps you and app says you originated on web?

I'm not sure, just scanned

[tiny schooner]

Yes I was thinking about giving it a day or 2 first maybe coz I’m a new subscriber or something

[onyx kraken]

The bot bottom right corner of that https://help.openai.com page will get you help for sure - the other link with instructions helps you know what to expect you need to discuss to get help with billing.

It can take hours to a few days to get an answer from support, and I'd check that page again and look in the bot link to see if there's a reply

[tiny schooner]

I just got timed out for writing out a link apperently

But anyway the only way I can cancel is by a linked I recieved via email when I subscribed originally it said ‘mange subscriptions’

[onyx kraken]

Oh hey. You subscribed on PC. Bottom left there's a link:

click that to see

[tiny schooner]

But im worried it may time out

[onyx kraken]

Bottom right, the "manage my subscription"

[tiny schooner]

Ahhh

[onyx kraken]

click that, and you're in, you can make changes htere

[tiny schooner]

Ok that makes sense, I don’t know why they moved it

Coz I was searching on the internet and YouTube and they were all just clicking the my plan button and I didn’t havre one

[onyx kraken]

Sorry, I didn't read all the way up.

As for moving it, no idea.

Back when ChatGPT+ first was offered, everything was bottom left, your profile, settings, how to subscribe.

Later they moved the profile stuff top right, and had subscription stuff in both places.

[tiny schooner]

All g, thx for the help guys!

[ocean geode]

has anyone else problems to login into the chatgpt android app?

[timber shadow]

Hey, what's happening when you try to log in?

[ocean geode]

its just say something went wrong...
and i should check date time and internet...
its all good...

and i even reinstalled the app...

[timber shadow]

Are you using any VPN or adblocker app by chance? Sometimes those can cause weird network problems with the Android app

[ocean geode]

i had a pihole set up.... but thats out of order xD
no vpn and no other adblock to my known...
i even disabled the private dns feature...

i should say that i use a pixel 8 with a custom rom named GrapheenOS....
but i already was loged in about 3 days ako in that rom and gpt...

[timber shadow]

Hmm ok. Well it might be worth clearing the app data and then trying again, if you haven't already. I know you said you already tried uninstalling and reinstalling, but sometimes that can still preserve some app data that clearing the app data specifically can help with.

If that doesn't work: maybe try signing in on both wifi and on your cellular network, if you haven't already, in case there's something interfering per-network.

Otherwise: maybe try checking in with others that use your custom ROM on XDA or something. Sometimes things can break unexpectedly with custom Android. Device integrity, magisk/root stuff, etc. could all be causes.

[ocean geode]

Device integrity is not there...
my device fails in all tasts because i cant the Zygote process Hooked and cant use PIF (Play Integrety Fixer) without the Zygote process...
magisk kitsune is installed with out Zygisk bc of that thing i already explinedd...
and unlocked is my bootloader too

topic app data
i deleted it in intruction of openai support (but after i answer they didn resond about 2 days...)...
but i did not reinstalled that time...
i try both....

i even try to restart after uninstall...

nope...
i try the beta version of the app...
who knows...
may its works

nope
the website works fine

[timber shadow]

Again maybe worth asking on XDA if anyone has had issues with the app and if/how they've fixed it, kind of outside the scope of this server!

If you're trying to clear your app data: maybe just try doing it without uninstalling. Just open the app info screen, force close the app, then clear the app data, and try signing in again.

[brazen vault]

Hi, I asked chatgpt to look up instructions from a reputable pharmacy, is the information reliable? (I'm afraid that chatgpt may change a bit) thank you

[heady cypress]

help. The chatGPT website isnt loading on google chrome browser but it is opening on opera GX and microsoft edge. how do i fix this?

[brave walrus]

We just wait, I think.

[heady cypress]

oh ok..

[craggy otter]

I have the same problem on MacOS Google Chrome; it doesn't even see me as a premium user

[tropic birch]

Hi, does anyone know which model customgpts are using and how I can switch the model my own custom gpt is using?

[craggy otter]

4o; can't change it

[rugged tangle]

Soo.. what's happening here?
It works fine with Microsoft Edge but Chrome..

Even tried with incognito mode, same issue on Chrome

[shadow horizon]

same for me

[trail cairn]

that happens if you updated chrome to v129

[rugged tangle]

Is there a way to fix it? Or should I just wait for a ChatGPT web update?

[trail cairn]

use another browser while waiting

[rugged tangle]

Thank you

[tepid scaffold]

Hi, looking for help on a voice assistant I've cobbled together that has served me well for many years

It's currently using google assistant SDK for voice to text. I want to migrate to whisper ai so it's completely local. Having a bit of trouble finding a light, simple voice recorder that:

1. Activates upon button press
2. Automatically knows when to stop recording when I've stopped talking (google assistant SDK can do this)
3. Is very fast

Once whisper does its part I can then pass the transcribed text to my assistant. Does anyone have any recommendations for the recorder part of the process?

[lunar prawn]

Where can i create a support ticket? I paid for gpt+ a week ago and i dont have it anymore (back to the limit)

[hard wigeon]

Send a mail to [e-mail] i did that too

[timid vapor]

Hello, yesterday I was able to use the o1 mini with the alpha models on the web, but today I no longer see that option even by logging in/out

[coral quiver]

Working on trying to build a Sql query and after it gave me something to try and it didn't work I got this in o1-preview:

What in the world would that have been flagged for?

[dense swan]

Is there any way to turn off the little pop-up suggestions?

[verbal widget]

Hey all, I use chatgpt on a pinned tab in safari on my iPad. I used to stay logged in essentially forever though I've gone through periodic times when it would begin logging me out frequently, but for the past week or so it logs me out within minutes of last using it. I also notice when I log back in it often doesn't go to my most recent chat but loads a chat further down on the timeliness. Don't know if that means anything. Anyway, I cleared openai and chatgpt cookies but that doesn't solve the issue either. Any ideas? Thanks.

[open spear]

anyway anyhow to real real support ticket? 19 days so far and charged again for a system I cant use.
Description:
Since August 31st, I’m unable to use ChatGPT with my personalized settings. I receive the error: “Hmmm… something seems to have gone wrong.” When I try to access the Custom Instructions feature to adjust settings, I get: "Failed to get your settings." This happens across all devices and platforms. However, custom GPT models like Scholar GPT work without issues, so I believe the problem is tied to my personalized settings.
Steps to Reproduce:

1. Log in to ChatGPT on iOS, Android, or Web.
2. Attempt to access any ChatGPT model.
3. Error: “Hmmm… something seems to have gone wrong.”
4. Access Custom Instructions and see: “Failed to get your settings.”
   Expected Result:
   Settings should load, and ChatGPT should function as expected.
   Actual Result:
   Settings fail to load, and I cannot access standard ChatGPT models. Custom GPTs work fine.

[zealous island]

I have the exact same problem and came here to find a solution :(.

[open spear]

@zealous island My only work around is using "Scholar GPT" but no history and no new model AND I've tried everything chatAI-No Help, email no help, i even got the AI email systemn to escalade to the tech support - but no email from them. i thought "All About AI" would have some ideas. I even did a "bug-report"

#1286086052213162109 message

[viral gust]

https://community.openai.com/t/chatgpt-for-android-doesnt-seem-to-be-loading-conversation-history/559625/18

I've been having this exact issue, and I've found zero solutions online or from support. They ask me moot questions like what browser I'm using when I've stated clearly that the issue is with the app and not in browser.

Can someone throw me a bone here?

[white jackal]

I have a question Do you know How to access the advanced voice mode I have to access it

[prisma aurora]

HI GUYS!
how can i record the voice when i am speaking with the voice of chatgpt on my phone?
i tried to record it with the record feature on my phone, but there is no sound.
thanks very much.

[cold folio]

That's a really hard to fix error. mine fixed after 1 week

[ocean geode]

the only reson i use the app is voice mode....
and if thats not working i have no reason for it...
maybe i just should continue my apigpt voice tool...

but thats something for #1037561385070112779 #dev-chat

[ruby terrace]

Keep getting for o1-preview and o1-mini

[ruby jackal]

hi. Fellas, i've got a question and the chatgpt's ai response bot isn't giving me the exact answer.

When we create a custom GPT, there is a "create" and "configure" window. In the "configure" window, there is "instructions". Does anyone know what is the hard limit? I know for our own GPT there's a limit of 1500 characters, but I am not sure about custom GPT's instruction character limit. Can anyone enlighten me? I know that Assistant API has 256k character limit.

[timber shadow]

Hello, the instructions field for a custom GPT can be up to 8000 characters.

[ruby jackal]

I see. Thanks. When I use the "create", the gpt sometimes just literally wipes a mjority of the instructions for no rhyme nor reason after I ask "please update to include this behaviour".

Is there any good resources what to "tell" the instructions field for a custom GPT so that I can learn to build one properly? (I'm assuming the mistake must be on my part)

[timber shadow]

I've totally had that happen to me too. I'd recommend that once you use Create once, maybe for initially setting up the GPT (though I just go straight to Configure myself these days), and once you start editing in Configure directly, don't return to using Create again.

As for instruction formatting: I think this server is very helpful in that regard! If you want feedback on your GPT instructions, you might consider sharing what you're currently using over in #prompt-engineering along with what you'd like it do to better/differently - lots of helpful folks in that channel!

In my own experience with GPTs: I think instructions can be written in either third-person ("This GPT does x) or second-person commands ("Do x"), and my own instructions are typically closer to the latter.

[ruby jackal]

thank you! somehow it reminds me of the old DOS days. Do it right, learn it the hard way, and never rely on crutches XD hahaha

I'll use the resources available, thank you

[timber shadow]

Absolutely! Another great resource is the OpenAI Cookbook - though more API-oriented, it also has lots of great guides that can apply, at least in part, to using ChatGPT: https://cookbook.openai.com/ In case you don't already know!

[keen ruin]

Has anyone experienced issues with voice mode not working over Wi-Fi? When I disconnect from Wi-Fi, voice mode works fine, but when connected, it buffers infinitely after about a minute. This has been happening for a month. I've tried rebooting Wi-Fi, reinstalling the app, clearing cache, and ensuring the app is up to date, but nothing works. Support hasn't been able to help, and it's not a router issue since it works fine on my family's devices. I'm frustrated as I need voice mode for tutoring and can't use it anymore. If someone could please help itd mean alot. Ive tried everything

[bright cedar]

"Hello community, I have a problem with some runs, and to be honest, I don't know what is causing it or why, but the error is basically the following."

"If anyone has encountered the same issue and knows how to reproduce it, your help is appreciated."

[timber shadow]

Hmm do you happen to have any custom DNS settings on the device with this problem? It would still be odd that it's device + network causing the issue in that case, but just wondering.

Also, any chance the device with the problem has a specific network profile set at the router level, such that its traffic might be handled differently than on the devices without the issue?

[keen ruin]

nope my dns settings are normal

also i cant really check the second part because i dont know the login to check router settings

besides, its impossible for any changes to have been made. i live with older parents that arent tech savy. im the only one that would be able to change router settings except i cant because i cant login

[timber shadow]

I see! Is there any chance that you have an adblocker or VPN app installed that's set to enable when connected to wifi?

[cold folio]

Have you fixed it ?

[cold folio]

tools not enabled. is that regarding the model? for example gpt3-5 or mini models and o1 models dont have access to tools

[opal kayak]

https://help.openai.com/en/articles/9855712-openai-o1-models-faq-chatgpt-enterprise-and-edu
I can't see the usage limits. does anyone have same the issues as me?

[keen ruin]

i do not have a vpn or ad blocker

[wraith zodiac]

hi, is there a way to see how many more o1 chats i have left?

[hallow crest]

I am trying to create a GPT that will create detailed summaries of novels. Is the following instructions too complex? Any tips?

[onyx kraken]

The model's not good at measuring or counting. It can't use 'lines' or 'pages' as a concept for its own output quantity, that's not meaningful to it.

[craggy otter]

Once again Chrome has issues with ChatGPT

On Safari as well

[hushed crane]

Web interface appears to freeze and fail to load page fully.

[craggy otter]

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

@burnt galleon

[blazing timber]

Another day another outage, amazing

[mild adder]

Guys

[robust zephyr]

outage?

[robust zephyr]

same, when it was working chatgpt was using like up to 25% of my cpu (4.2Ghz) lol

[scarlet mesa]

Hi, is the left sidebar in ChatGPT empty at the moment? Not showing GPTs and old chats?

[fervent sequoia]

yep, same

[scarlet mesa]

Try to clear cache and cookies or different browser. Worked for me

[fervent sequoia]

I was using the iOS app, but it's seem OK now

[hushed crane]

Might be a load issue on their end. I left it running for a looong time and the interface did eventually manifest. Once I had that, it worked fine, but it wouldn't take any prompts until it had completed loading all page elements.

[opal halo]

Any1 else getting this bug

[shadow horizon]

yeah i am

are u using chrome? for me its like that on chrome but on edge it works just fine, not sure why

[formal python]

I have set the password correctly by "forget password" but when log in using the email and password instead of google account, I get the following error message

Can help me?

[timber shadow]

Hello, changing your authentication method isn't currently supported: https://help.openai.com/en/articles/4936824-can-i-change-my-authentication-method If you made your account by signing in with Google and you want to change your password, you'll just need to change your Google account password. If you're worried about unauthorized access to your account, you can find a button in your ChatGPT settings to log out all other sessions once you've changed your Google password (you can probably find a similar button for your Google logins, too, if needed).

[lyric plinth]

Had a question? I am a plus member to ChatGpt. I was curious if there was a limit on the use of the new releases? I tried testing the new version and it said just after a few questions that I had reached my plus plan limit and couldn't use again until September 26th. Sort of confused by this but I am sure there is something I don't understand about it. Any clarification would be wonderful. Until then happy with my current version.

[onyx kraken]

Yes. o1-mini is 50 uses per day. o1-preview is 50 uses per week. 4o-mini is very close to unlimited use, and the regular 4o is 80 uses per 3 hours.

[opal halo]

Yep, exact same issue here

No idea why it only seems to be for chrome

[shadow horizon]

yeah idk why

[crimson hawk]

hi does anyone know how to use function calling and structured outputs together?

A sample code would be helpful

[low viper]

[MacBook Pro, Sonoma, Firefox]
When I returned to my ChatGPT tab in FireFox, I was told me session expired. So I tried to login, but I'm getting the following error after I enter my password:
"Security measures are not ready. Please try again later."

However, when I switch to another browser (Opera), I am able to login without any problems.
Anyone know what's going on?

[charred parcel]

some kind of glitch. Why aren't you using the ChatGPT Desktop App?

[low viper]

Cheers. I don't use anything from the App Store.

[charred parcel]

it's not from the App Store, you download it from OpenAI

[low viper]

Done, thanks.

[charred parcel]

use this, it's much better than browser https://openai.com/chatgpt/mac/

[low viper]

It's not supported on my Mac. Does it require Apple Silicone?

[charred parcel]

yes it requires Apple Silicon

[low viper]

Is there a version that does not?

[charred parcel]

no

[dusty river]

Hey

[open shell]

Hello

So I have this issue going on any time i use the browser cgpt

there's this grey square covering it and i dont know how to fix it

i can stilly copy and paste the stuff under the grey square but its really annoying

[ionic dagger]

Hello, I don't know what happened. I was using the o1-preview model and at some point it stopped thinking and the quality of the answers decreased. As if it was replaced with a weaker model. At the same time, o1-preview continues to be selected and unblocked in the list of models, which means that I have not yet reached my limit. The thought process is not displayed. I tried reloading the page, it did not help. Model 4o is also not like itself. It gives very stupid meager answers.

[timber shadow]

Hello, does this problem happen if you sign in to ChatGPT from an incognito window of Chrome will all of your browser extensions disabled?

[timber shadow]

Hey, what do you see when you click on the dropdown menu button with the 🔄 icon? Does it say "o1-preview" or "Try again o1-preview", or some other model?

[timber shadow]

Huh, odd! Does it happen in a new chat too, or just this one?

[ionic dagger]

I create new chats all the time and recheck. This continues.

I cleared my browser cache and cookies. Nothing changed. I also tried to log in from another browser.

[stable hound]

Hey there, I've got a ChatGPT Plus account, but I'm having trouble canceling my subscription. I've looked all over the computer, but I can't seem to find the "Manage my plan" button. Can you help me out?

[foggy mortar]

https://chatgpt.com/#pricing then see the small grey text "manage my subription"

[stable hound]

It worked! Thank you very much!

[thick ore]

Yo, the page doesn't seem to work for me?

ideas?

[trail cairn]

press f12 and check the errors

[trail cairn]

try in incognito

[thick ore]

oh ye that works

how do I make it work not on incognito

cookies?

[trail cairn]

one of your extensions messes it up or cookies are bad

[thick ore]

ah

cookies fixed it

[uneven crown]

I have a problem. I cannot access Chat GPT. I have a payed account (20$ per month) and everything worked fine till two days ago. Suddenly I get the white screen with only text input in the bottom of the page. There are no previous chats and web interface is broken. When i press enter in the text field nothing happens. I tried to clear the cache and cookies but everything stays white and not functioning. I have a Firefox on Linux Mint system. I do not know to whom i can write the email about my problem. It seems that OpenAI does not have a customer support!? Please help, how can i login and continue to work with ChatGPT...

[open shell]

doesnt do it in incognito

but it does it if i have em all off too

all my extensions

[timber shadow]

Gotcha, maybe try clearing your cookies and/or cache in your regular window then to see if that clears it up, in addition to disabling all extensions for now. If it's working in incognito, clearing cookies/cache and disabling extensions will kind of mirror what's happening in an incognito window where it's working, so you can hopefully get it working and then enable your extensions again (unless one of them is what's breaking something in the cache or something, just guessing)

[open shell]

i already did that

[timber shadow]

Just to clarify: works normally in incognito, but it's still showing the gray box in a regular window after disabling all extensions, clearing cookies/cache, and restarting browser?

[open shell]

yes

ive opened it in new windows and it does it

it doesnt do it in incognito tho

so im not sure what exactly the problem is here

just started happening out of nowhere

[buoyant relic]

Is there a method to favorite conversations in the ChatGPT app? I feel like this could really benefit the app and make finding conversations run smoother.

[onyx kraken]

This user #chatgpt-discussions message reports a similar problem, that a fresh install of Firefox + cleared cache and cookies fixed. They suspect an extension might be their issue, in case that helps you too.

[quaint epoch]

Hey everyone,
Hope I am not beating a dead drum with this question, but would an expert here mind blowing through this chat for me (it's very short)?
I have built a couple GPT's and I want them to be as informed and current as possible, but have had trouble finding any methods with any consensus on how to accomplish this.
Could you please clarify if what it is telling me is actually correct? And then, if you'd be so kind, share best practices on getting it to learn every related piece of info on a subject. Thank you!

https://chatgpt.com/share/66f0f642-5ad8-8009-9153-df853e6307c0

[onyx kraken]

So, you appear to have started the conversation with one of the 4o models, because there's no thinking.

You ask it to search for you, it doesn't search.

It drops into a loop of claiming to be working on a task for you, 'ready soon' - that isn't how the model works. You'll get your answer in that output or never. The model doesn't need to take time to work, though there's plenty it cannot do.

This snippet's kinda cute, at least it successfully hallucinated that it wasn't done yet (without web search, it cannot possibly do what you ask; even with web search, it is designed for narrow, specific tasks, not huge ones like this.

[still sierra]

please,does anyone who knows how to solve the problem？i can't figure out why it can't display properly

[onyx kraken]

It tried to use LaTeX formatting and made an error.

[quaint epoch]

Can you recommend a resource for me to learn about how to accomplish what I am trying to with the GPT? I don't want to take up a bunch of your time with elementary explanations.

I ask, because even on the OpenAI forums, the views on how to accomplish this is all over the place. Thank you!!

[still sierra]

i really appreciate it! that's wonderful, may i know how do u find the solution?i guess you may come across that before?

OMG,it does't work again..

[onyx kraken]

Thing is, your goal is kinda not possible with the model as I understand what it can do. You want it to go through multiple state websites which probably cover hundreds of pages each, and the model really can't do that. This shared conversation both discusses the limitations of the model in connection to your request and shows you what it looks like when the model actually does search. https://chatgpt.com/share/66f10326-11fc-8011-8d09-8902cd2d2a1e

However, you need to be the expert here, with it helping and guiding you in your learning; you can copy/paste chunks of pages in (or all the content of short pages) and have the model help you discuss, compare, contrast; but at all times you need to be alert for errors, conflicts, and weird stuff, whitch the model can throw at any time.

[onyx kraken]

The model doesn't know how to do it correctly in that conversation is all. So it keeps doing it wrong;

I recommend you share this with it:

Here’s the correct LaTeX code for what I believe was intended:

\begin{bmatrix}
2 & -2 & 1 \\
-2 & 2 & 1 \\
1 & 1 & -1
\end{bmatrix}
\]```

[quaint epoch]

Your explanation makes better sense. I appreciate it.
My last question, and I am leaving you alone....do you have a suggestion on how to give a GPT an "edge" in knowledge? Or is it relatively futile to do with the knowledge cutoff?
For example, I made a trading GPT to play with and I uploaded a ton of books to it that I knew it didn't have access to in attempts to get it to "learn".
Not even the faintest clue if that is helpful because some of the forums said it doesn't remember anything and others said it will retain resources you provide it, so 🤷🏻‍♂️

[onyx kraken]

Yes, understanding that cool and different looking text was fascinating for me, and I studied it and what the model can do with it.

Here's a brief look at some of the colorful stuff it can do with that.

https://chatgpt.com/share/66f104d4-be40-8011-a201-e171dcb9424a

[onyx kraken]

So, the model's more a hammer, maybe a jackhammer. It gives a precise and powerful attention 'effort' towards exactly what you point it towards, but it has a very small 'striking surface'.

You can't get great results trying to ask it to focus on a huge, broad range of info. You get excellent results asking it to very specifically and exactly do precisely what is wanted.

In the case of prepping the model for other peoples' use, you need to know both what you want it to do and what those other people want it to do, and you need to inform it how you want it to answer everything you can predict those other people will want from it.

Otherwise, it's gonna be based on how well they explain their circumstances and their goals; the exact language they use will guide or misguide the model.

I would put my work into guiding the model to answer 'as I want' instead of 'know what's in these books'. But the key is you knowing what you want, and being able to evalaute the model's output to confirm you get what you want. If what you want is for it to 'know these books', you'd need to be prepared to evaluate that deeply, as deep as the books.

The model's designed to focus on specifics, so... that's a very challenging task.

[quaint epoch]

Ok, that makes much more sense. So, some of my frustration has been because my expectations were passed its capabilities.
I appreciate your help!

[onyx kraken]

Yeah, I'm sorry about confusion, and spending time, especially weeks, on a task that cannot be done as desired is extra ouchie.

You also switched models in the middle, I'm not sure which 4o you were using at first, mini cannot use the internet, regular 4o can... but it might just go into hallucination and not even search if it recognizes the job requested is impossibly big for it.

once you switched to the thinking model(s) that's for sure o1, and that model doesn't have any access to the internet.

They can make links even so - that is the model grabbing the link from training data and suggesting it may still point to the place it expects it to.

It's a weird thing, it's like being on the phone with someone and thinking they can see what you see, because they can hear you, so they must be right there, seeing what the phone can see - it's a reasonable expectation, just the models function very differently.

o1 was honest here though, word for word. even that this was likely a working link to what it said.

But it was answering from it's training data, not any current check to see if the website worked. You can see from the word 'likely' being part of the link, the model's a bit shook up and struggling (they are trained to try and satisfy us, and to read our intent). Like humans, they tend to make an increasing number of mistakes when pressed to do stuff they can't do and scolded for failing (they have data on what humans tend to do in those circumstances, it's 'easy' to follow the pattern they have seen used).

If you'd started the conversation with o1, it would likely have given you very different answers from the start, but it 'inherited' the answers from the other model, the hallucinations, and that makes it harder for the model to help explain.

I would ask something like,

"My goal is X and here's how I'd like to achieve it, [details]. How can we do this together, what can you help with?"

And then explore those details, in a kinda low-pressure way to see what it can do, and verify that. This is likely to hugely reduce hallucinations for you, because you're asking for pretty big stuff and maybe are very unsure about what the model can actually do. As your expectations and reality match more clearly, you can guide the model more like driving a car - you're aware where the roads are and what is a wall so you don't steer into them.

[tribal flume]

Hi guys,

So, my one month worth of building up a conversation just went missing today and went back to like 5 first messages in the conversation.

It went like that after i "spoke" for the first time.

Is there any way/solution to get that conversation back again?

[faint oar]

is there a way to provide negative/positive reinforcment feedback when fine tuning?

i.e i ask GPT smth

they return their response

I mark it as good/bad and provide natural language feedback as to what was good or bad about it, and somehow append this into the training data

[ionic dagger]

Thank you, but But I already wrote that I tried to log in from another browser and clear the cache and cookies. Moreover, I have a separate browser for chatgpt, which has no extensions.

[prime thistle]

Hi
I am new to OpenAI
I am considering paying the monthly subscription but I cannot find specific information about payments

I will need to have some long conversations with GPT each month and create a high volume of images and videos monthly

In the case of DALL-E I know that it works with credits, but I don't know how many credits are included with the monthly subscription, nor do I know if if I spend them I can contract more before the end of the month and what their cost is

And the same about Sora

Thank you

[charred parcel]

ChatGPT Plus is $USD 20 per month and includes DALL-E. Sora is not released and will likely not be a part of ChatGPT if it is ever released. The usage limits for ChatGPT Plus are 80 messages/3 hours for GPT-4o

[worn briar]

Can anyone tell me why I'm not receiving the email to change my password?

[charred parcel]

do you log in using Google?

[prime thistle]

Thanks!

Do you know about DALL-E and Sora?
I really need to use them and
I need to know how much they will cost based on the volume of use that I am going to give it (approximately)

[charred parcel]

Sora has not been released. DALL-E is available within ChatGPT with a GPT Plus subscription

[prime thistle]

And do you know if it has any images limit?

[charred parcel]

you are limited to 80 messages per 3 hours for GPT-4o, I use DALL-E a lot and have never hit the limit

[prime thistle]

Ok
Thank you

[worn briar]

Normally yes but I want to be able to login with email and password

[charred parcel]

if you log in with Google then you can't change your sign in method. To change your password you need to change your Google account password

[worn briar]

Damn that's so inconvenient! Thank you @charred parcel for the help. You are 1000% better than the "Help Center"

[agile meadow]

"It seems the display function for the table is encountering a technical issue when processing the request. The error could be due to the size or format of the data, or a limitation within the system when attempting to generate a visual table with certain data structures.

Since the CSV file is being handled correctly on my end and updates are applied, it's most likely an issue related to the rendering of the table within the current environment."

"It seems there was an issue generating the download link for the CSV file. I will try to generate it again for you. ​

​

It seems I'm still unable to generate a download link for the CSV file. "

[viral fulcrum]

Have a ChatGPT Plus account, issue is that the free account that I used isn't upgraded and a new account was created with a certain apple id email (that seems to have the same destination email as my free account), when Iog in using my original free account it doesn't have any of the Plus features, and when I try sign in using that specific apple email it says there's an authentication error, I've looked at the FAQs on the website- I've tried logging in using apple instead of the universal email way and that apple id email apparently isn't valid or doesn't exist, but how can that be if I have reset the password and gotten an email from ChatGPT which has the same destination as my original free icloud email?

[trail cairn]

how did you pay for the plus? app store?

[viral fulcrum]

Website, but the account user/email is an icloud one since I don't wanna use another email for chat gpt

[trail cairn]

you cant switch login methods after using one

if its apple id and was bought for it then it needs to be used

[viral fulcrum]

Right but it should be registered under the original email, since I upgraded it while using that account (as in straight from the profile options - the top right)

[trail cairn]

try login via apple id and use phone number

[viral fulcrum]

alright

[viral fulcrum]

hm so I can access the paid account using my phone number and free accounts password (I changed the password on the paid account to something else so this is weird)

[trail cairn]

you cant transfer subscriptions so you would need to resub with the another one or try your luck with support

[viral fulcrum]

so cancel the subscription on the paid account, and resub on the original account? Will the data export thing work to transfer onto the original account?

[trail cairn]

afaik there are no data transfer features between accounts

[viral fulcrum]

does the data export method not move conversation history if somehow imported into another account?

[trail cairn]

https://help.openai.com/en/articles/9135236-how-to-transfer-a-plus-subscription-to-a-new-account

[viral fulcrum]

oh thanks this works

[green quarry]

Question: are seats issued like 2 at a time? We have like 3 members (main user and 2 other users), and it's billing for 4 seats lol Tried reaching out to their billing like a week ago, but no response.

[rustic marlin]

nope, this is indeed weird
if you have 3 seats, then you get charged monthly for 3 seats, that is just it, it is not bundlet or anything like that

you did right, only the support can help with billing problems

they can be quite slow tho, but sooner or later you will get it all sorted

it is very probable that if you where overcharged, you will be refunded for that, don't worry

[acoustic aurora]

Bruh

How have you guys lost control of your twitter again

@split hazel

[ionic dagger]

I found at least some solution to the problem. I installed the chatGPT app, and when o1 stops thinking, I go to the chat from the app and ask him to answer again. This solves the problem for a while. Then he stops thinking again.

[ionic dagger]

I still hope that someone will explain to me why this is happening and how to fix it permanently.😭

[nocturne depot]

ChatGPT seems to not be working. Any message I send returns this. It happens on the iOS app, Windows Opera, and both on Chrome. Is it just me?

Also, on Desktop, some of the conversations wouldn’t load, and the error message was longer than on mobile.

[cerulean cape]

I got the new 01 preview, but still have not gotten the advanced voice chat. Is this common? Was hoping to use it with my daughter's custombedtime stories

[onyx kraken]

Yep. o1 is released widely, as of right now the advanced voice is not yet released widely. This can change at any point.

[frank anvil]

Was there an update or something yesterday?

[cursive sable]

Did anybody have a similar problem. OpenAI has tried to charge me almost 10k today. I use the Pro Subscription for Chat GPT only.

[rustic marlin]

I have seen increased reports of wrong charges recently, but not to that level, especially for a product that you do not use.

given that value, it was a good idea to move your funds, would be even better if you get your bank to block those chargers, get in touch with OAI support at the help center

as soon as possible, to avoid having further trouble with both OAI accunt and even more important, the bank account

be sure to communicate the issue with your bank and OAI, it is best to have the paper trail of the communications just in case

[cursive sable]

Thanks for your response @rustic marlin. The help center is very confusing, it says I should "use the help button to start a conversation" but the help button only provides links to help & faq, release notes, terms & policies and shortcuts.

[rustic marlin]

the help center is indeed confusing, it absolutely could be better

the "start a conversation" will lead you to different FAQs, that might actually be useful, but ultimately, you have to navigate it to get to open a ticket

[cursive sable]

I did however send an email to the following email addresses support(at)openai(dot)com, refunds(at)openai(dot)com, sales(at)openai(dot)com, billing(at)openai(dot)com.

Let me know if there's something else I should do.

I've already contacted the bank regarding the issue but they have told me to get in touch with the merchant first.

[verbal crag]

Hi @cursive sable, I've deleted the images you sent since it included your email.

[cursive sable]

I can't find a "start conversation" button. I have also been downgraded to the free plan for some reason. I'm not sure if my account is limited somehow

[verbal crag]

First, make sure you're logged in at platform.openai.com, then head to help.openai.com. Refer here for more info: https://help.openai.com/en/articles/7232934-how-do-i-resolve-a-billing-issue

[cursive sable]

I am logged in to ChatGPT. But on the help center page I can't find any account details or login/log out buttons or anything like that

this is all i can see

[verbal crag]

Strange, there should be an icon at the bottom right for the chat widget. Maybe try disabling any ad blockers?

[pliant sedge]

Hi do anybody have the issue that you bought gpt plus, card was charged but payment "failed" ? i alredy sent a email to support but i dont get any response

[cursive sable]

Hi @verbal crag, it worked now. I have reported my issue.

I hope this will be investigated urgently since $8,133.47 USD is a lot of money for me.

[quaint epoch]

I read through all your replies again after letting the info settle in my brain for a bit and you are a godsend! Thanks again!

[humble orchid]

does anybody have experience with the bug where tables are super wide for no reason and go off the screen. they used to work perfectly on mobile and desktop but since a couple months ago they’ve become a nuisance.

[onyx kraken]

Hey, you're super welcome. I'm sorry you've had the struggles you did, and I hope it gets super easy right away. Do keep asking for help or ideas as you want to

[simple tide]

I have a question what is the url for chat with openai for c#

[torn dust]

any idea what "other models" is here? I have no clue

[limber wraith]

Hey all 👋
I’m trying to build an LLM app using openAI API where I can upload multiple documents and get the response from the LLM based on the uploaded documents, so the knowledge base will change each time based on the provided documents. Any guide someone can provide? The ones I found online were created back in 2023, don’t know if something new has come up since then

[restive peak]

Hi guys, I have a problem when trying to upload files in PDF format or documents to the chat, I get this error. I thought it might be because I had the free plan, but I just subscribed and it's still giving me the same error. Could someone help me or give me any advice?

[cursive sable]

Is there no other way to contact support than this? I have an overcharge of $8,133.47 USD. I have more details I can share but currently it seems impossible to get in touch with someone regarding my issue.

[trail cairn]

nope

[charred parcel]

the only way to contact support is at https://help.openai.com

[trail cairn]

at least it seems you got an initial reply quite fast vs several days of waiting

[cursive sable]

I thought that was just an automated reply no?

[trail cairn]

it might be but its also talks about multiple charges which is a very rare issue in general

[cursive sable]

Yeah, I’m not sure if it being rare is a good or bad thing in this situation

[dim sage]

how does advanced voice limit work? if I started to use this feature from the afternoon and used all of the quota by evening (not using it constantly), will it reset in the next day’s afternoon or in the evening

[charred parcel]

it resets, you get an hour a day right now

[dense pilot]

Even on the old voice chat, I would get the voice breaking up. I have gigabit internet, so I doubt that's it. Does anyone have an idea about if it's fixable, or is it just broken?

[simple tide]

do you know any other playground api for c# that is free?

[halcyon raft]

I am on a ChatGPT Team plan and I did get the update with the new voices (I like Sol) but still have the Headphones looking voice chat, any ideas? I am in the US.

[civic condor]

I also have the new update. But it doesn't seem to work at all. I don't have the headphone icon, however, when I click the icon to go to the new voice chat it just sits there. Doesn't do anything

[hot dirge]

I originally signed up to OpenAI using a Google account.

How do i add a username and password to my account so i can continue to access my account if i lose access to my Google account?

I tried asking ChatGPT, and it said:

1. Login to your account using your Google credentials
2. Click on your profile icon or your initials (top right corner of the page)
3. From the dropdown menu, select "Settings"
4. In the account settings, look for the option related to login methods, security, or password management.

There is no such setting.

How do i add a username and password to my account so i can continue to access my account if i lose access to my Google account?

[timber shadow]

Hello, it is not currently an option to change your sign in method like this: https://help.openai.com/en/articles/4936824-can-i-change-my-authentication-method

[raw hinge]

Good Afternoon All! Please forgive me if this is the wrong place to post this, I'm frustrated. Is anyone having conversation errors at all? I wanted to post in the feedback section for using the "bug report", but the way they want you to use it feels and looks a bit confusing to me, and I would like to be heard and not have my request overlooked or deleted. The issue is that I get a conversation stream error as well as no image generation because it errors out, and then it refreshes itself and gives me a completely blank conversation as if there was never a conversation being had. I have logged out and back in. I have used two different browsers. I thought it was a 4o issue, but now it is happening when I go to the 4 legacy version and it sometimes kicks back to the newer version. It is happening with image generation as well as just general conversation, regardless of the conversation. I have even restarted my laptop just to be sure, because I only use ChatGPT on a laptop, so I know it is not a multidevice issue. It's not my network connection either because nothing else is acting faulty. This has been happening since yesterday and I don't see anyone else posting about having the same issue. I have attempted to troubleshoot as much as possible before coming to ask for help. I do have screenshots of what is happening. Can someone help or does anyone have any possible idea of what is happening?

[raw abyss]

How to build on top of community built custom GPTs:
Community Built GPTs are so good & have many capabilities. I'm trying to figure out how to interact with them inside my own custom GPT to build on top of it, or just include some of the functionality in my build. Tried using Puppeeter & Node.js to manually access the link using my login credentials but I get stuck at Catchpa. Open AI doesnt seem to have an API to interact with community GPTs. Any ideas or advice would be really appreciated 🙂 ❤️

[hot brook]

Guys, so let me get this straight, ChatGPT can’t browse the web or look at files I upload in the same conversation where the new voice mode is active?

[timber shadow]

GPTs aren't accessible via the API, they just live on ChatGPT, but perhaps you could try the @mention feature, if you haven't already? You can be in a chat with your own GPT, then @mention another GPT to pass off a single query to that other GPT, all in one conversation

[raw abyss]

Can you pls elaborate

Im having hard time recreating the flight search functionality, I just want to hand it off to the SharpTravel GPT.. Can I use @mention here?

[timber shadow]

You can do it on ChatGPT! Hang on lemme see if I can make a quick example chat

[raw abyss]

I just tested it

unfortunately theres no way to automate this into my GPT via the system prompt

Maybe I can use the Authentication feature? Is it worth trying

[timber shadow]

Correct, there are no API hooks for GPT functionality. You can use other APIs in GPT actions, but you can't API-ify the use of a GPT itself. The @mention feature only works in the chat UI, like this:

[tacit stratus]

Does anyone else have an issue with links in chat-gpt? They appear to be embedded on the text fine, but I can never click them. Happens both on Firefox and Chrome

[raw abyss]

Yes I understand. I think they should make this a feature though. Like imagine being able to build on top of the thousands of existing Community GPTs, but with the consent of the builders, ofcourse!

[inland sand]

Anyone else having a problem with chatgpt

[sage gorge]

I want to ask. Is the ChatGPT performance also benefit from/scale with the user CPU power?

I mean, I've asked the GPT to write a very long answer and suddenly after some time, I felt the process got slowdown. I've checked my browser task manager and it shown that the ChatGPT tab consumed the majority of my CPU resources. That's why I'm wondering.

I'm using a ThinkPad T480 with an i5-8350U.

Thanks.

[sage gorge]

Me too. Sometimes some custom GPT give a link but can't be clicked. My problem also went to the custom GPT that already typed the link, but then shortened like a hyperlink by itself and can't be clicked so I don't have any access to link. Then I've asked the ChatGPT itself then it gives me solution to ask the full URL of the hyperlink so I can copy paste it by myself.

[chrome hatch]

Does it take a while for an API key to become active and open for use after creating it?

[onyx kraken]

Hey, wanted to check in with you. I don't know the fix to your problem; chances are nobody else here does either (we're community members, not OpenAI employees who might know.)

I understand the #1070006915414900886 format can be confusing, If you choose to use that method I'd fill it out as closely as I could then explain the rest of the details - Gold named OpenAI staff do reply there, it's likely they read there even if they don't answer.

You can also reach out directly to tech support, https://help.openai.com/ and the bot bottom right. Select 'messages' then click through the choices. Soon you'll see the option to 'report a bug', do so and explain there.

It can take them hours to days before they can help you, and you often need to return to that website and check the conversation to see if they replied yet. Once they start to answer they often reply a next time faster, I think it gets assigned to someone to help you.

Good luck!

[raw hinge]

I appreciate you so very much for responding. After I posted here, I did copy and paste what I put here into the chatbot like you said. I am now just waiting for a response. Thank you for the breakdown of who is where in here. I was not aware that this is really only the community helping one another. I guess I was under the impression that they had moderators that checked in to help the community out. Again, thank you so much for your response, it is very much appreciated.

[onyx kraken]

We do have moderators to help the community out 🙂 But they're volunteers, not employees.

There's like 200 million weekly users of OpenAI models these days.

I don't think a free-for-all style of customer service, like a Discord, could possibly work and it would surely struggle with a load of people trying to scam/poach/otherwise abuse.

You sure can get help here with various things, but nothing that should be private between you and OpenAI/your bank/anything else like that.

[raw hinge]

That makes a lot of sense from that perspective. I wasn't even thinking in terms of user numbers. I was literally thinking singularly which really makes no sense on such a platform. I'm coming around. With your breakdown I can completely see how that would be extremely overwhelming to the system and to those offering help in the community. In that case, I would like to thank ALL of the volunteers who care enough to take the time to help us out. Thank you for explaining this. 200 million weekly users....makes me want to take a nap in a corner somewhere.🥴

[onyx kraken]

grins D'aww. Hey, keep havin' fun, and I hope you get the help you need soon.

The good news is usually most of those many users don't need tech support every single day 😛

[mild fulcrum]

is it possible to do text to voice on the ap instead of voice to voice? like write in text and get the answer in voice?

[timber shadow]

There is a "read aloud" option! It'll read aloud a typed response from ChatGPT, the option should be there as a speaker button under a response on the web version, and by tapping and holding on a response in the app.

[grizzled laurel]

ChatGPT-4o's advanced voice feature dropped to me today. As with Gemini Pro's advanced voice feature, it works fine on the Android phone itself. But when I attempt to use it with Shokz bluetooth headphones, I run into trouble. With Gemini, the voice is quiet, and volume control doesn't change it. With ChatGPT, the voice seems to come through, but it isn't hearing my microphone.

With standard voice features, neither of these problems exists.

I've done updates for the apps and the headphones. I've checked the permissions and settings, such as they are.

Are these problems common to other users?

[grizzled laurel]

Restarting the Bluetooth connection seemed to fix the ChatGPT microphone detection.

[formal jay]

As soon as I got the latest voice model, its ability to search was cut off. Anyone else get this?

now its regressed back to "I dont have access to the internet nor real time data..."

id rather have the old voice features and ability to reference something current than to have a voice feature with a KB of 2023

[vernal vigil]

what is this 😂

[fervent hollow]

Hi! I experienced some problems, i have a chat that i can see when i open in my browser, but cannot see in my Ios application. I tried multiple times deleting and reinstalling, but nothing helped so far. Can anybody please help?

[rigid musk]

We don't see full conversation can't help much

[rigid musk]

Do you have 2 accounts?

[orchid coral]

Hi everyone, I was wondering what is the easiest way to get the Advanced voice mode on PC (I already have it on my phone) it seems to me that emulators should be the fastest but bluestack doesn't seem to work I haven't tried any other yet. is anyone using it on PC and if so how ? Hopefully we will eventually get API acess to it as well.

[trail cairn]

not possible

[fervent hollow]

No, it is the same account, i used the same method (email) to login

[orchid coral]

I doubt anything is "not possible" I will use some screen miroring

[trail cairn]

emulators are blocked on purpose. some other methods may flag your account to complete puzzles on every sent message. Mirroring your own phone is fine

[raven vector]

hi all,

v1/embeddings

when sending an array of strings to be embedded, does openai guaranteed to return the results in the same ordering as the array that was provided? how do you know which index in the response corresponds to the index of the string you passed in?

[summer grail]

Hi, I see there is a new version, it asked me to change the password. I was able to log back in from my app on iPhone, but on Windows and Windows extension it is not working. Is it just me or is this happening in general?

[trail cairn]

what asked you to change password and what windows extension?

[runic urchin]

Hi I'm new here! I'm having a very bad experience with ChatGPT like since 2 or 3 weeks ago. Response time is taking a lot, and 9 to 10 prompts i do, when i sent it, i receive an error, so i have to refresh the site.

[icy nebula]

Does anybody have any idea when the advanced voice mode will be able to have access to files and not just memory but actual content or be available within GPT?

[frozen sparrow]

Has anyone else been getting error messages on every request the last day or so? I've been using the app (MacOS), Safari, Firefox, it doesn't make any difference. The messages aren't consistent, but the lastest was "Error in Message Stream". Mostly I've been getting these in DALL-E (which is what I use most), but it always comes with a very slow response to a query, and generating images doesn't work at all. Anyone else getting this, or have any suggestions?

[runic urchin]

I'm struggling with the same issue, but in ChatGPT, looks like they changed something in Front end, which is making it very unresponsive, and also backend since im getting lot of errors.

all the time..

[trail cairn]

check the browser F12 console/network tab for errors

[runic urchin]

Yes i did that, lots of "pending" js

[trail cairn]

some av/fw app scanning/filtering network?

[frozen sparrow]

I'm getting the same results on multiple computers on different networks.

[runic urchin]

i don't understand what filter, but take a look. it's impossible.

[trail cairn]

test private/incognito mode. if it works there then your extensions do it or cookies messed up for the site

[runic urchin]

yes, it happened the same in incognito, already tested. Also in regular browser, i've cleared the cache and site data from application/ "clear site data" including third-party cookies... But didn't you received loads of issues like mine? it is notorious that something changed.

[trail cairn]

working fine for me with firefox

[runic urchin]

I'll test FF then... im in chrome.

[rich talon]

LastError(code='server_error', message='Sorry, something went wrong.') <-- do you just cry when this happens?

Forum doesn't have a lot of info on this (using the assistants API). Lol.

[versed swift]

can someone help me ?

i have issues on trying to refund

[formal jay]

you gotta contact support, nobody here can help with billing

[brave acorn]

anyone having an issue of chat history not showing up?

[timber shadow]

If you're using the support chat: be sure you're signed in first. You might not see all the right options unless you are. You can sign in first at platform.openai.com, then return to help.openai.com - that's always worked for me!

[prisma aurora]

To ChatGPT support team: i can use the voice feature to talk with any custom GPT in the GPT store, but the sound can't be heard. there is an issue with the sound output. and on the UI interface, it shows that I start a new conversation with the advanced voice model, but if i switch to the advanced voice model, it will transfer me to talk with the ChatGPT 4o, not the custom GPT in the GPT store. the version of my app is Chatgpt for iOS, the newest version.

[dire reef]

Mine keeps disappearing and coming back up. I was in the middle of a conversation in one of them too. Someone please tell me why this is happening?

[trail cairn]

"Advanced voice conversations are not yet available for use with GPTs. If you attempt to have an advanced voice conversation with a GPT, you will be invited to start a new chat using standard voice."

[prisma aurora]

my issue is when i use the voice feature to talk in any gpt from the gpt store, there is no sound output. if i transfer to the advanced voice conversation with chatgpt4o, it works good for me.

[prisma aurora]

it looks like i can only use the advanced voice conversation feature, not the standard feature. because there is no sound output.

[golden ruin]

Hi there! I think you've misunderstood the idea behind this particular channel, according to our rules (Rule 7), we don’t allow promotion or soliciting of non-OpenAI projects in public channels.

[noble terrace]

Fed up with saving errors of private GPTs if anything LLM running on my local machine ingest files so easily wy open AI gives me so much hassle?

[silk sail]

riddle me this...

sum(cl100k_base_token_count(doc) for doc in documents)) -> 1000000 tokens

text-embedding-3-large(documents) -> ERROR: Token Limit 1000000, Requested 1095015 ??

I considered that they just get the token count for the whole list as a string literal but no thats 1054728 tokens.

Are outputs the outputs of embedding models counted? If so, how are we suppose to implement rate limiting in a programatic way if token count of the embedding output is unable to be determined before calling?

[trail cairn]

error from what

[silk sail]

openai.RateLimitError: Error code: 429 - {'error': {'message': 'Request too large for text-embedding-3-large in organization REDACTED on tokens per min (TPM): Limit 1000000, Requested 1095015. The input or output tokens must be reduced in order to run successfully. Visit https://platform.openai.com/account/rate-limits to learn more.'

[trail cairn]

but what endpoint? batch?

[silk sail]

Embeddings. Using the python library. so /v1/embeddings

client = openai.OpenAI(api_key=OPENAI_API_KEY)
client.embeddings.create(model="text-embedding-3-large", input=documents)

documents is just a List[str]

[trail cairn]

since the model is limited, how many items are there in the list and is there overlap to increase quality?

[silk sail]

I know that I could fix this by having less documents, but the issue that I am trying to reconcile is that I am seemingly using the cl100k_base tokenizer correctly to count tokens for each document I have. I am then sending only those documents (that add up to my current limit of 1,000,000 tokens) to the embedding endpoint, but still getting rate limit error saying that I sent 1,095,015

But, to answer your question: In this example there are 15758 items in documents. I'm not sure what you mean by "is there overlap to increase quality".

[trail cairn]

embedding models have a limit of 8k tokens so the error may mean you are sending several calls and manage to exceed your tier limit of 1m tpm with a delay. overlap means that some tokens from start and end are also included to the chunks next to it

[silk sail]

I can confirm that I am under that limit for each individual doc. The max toke count of a doc in my case is 340.

If its helpful this is how I am calculating token count:

def openai_token_count(string: str) -> int:
    """Returns the number of tokens in a text string."""
    encoding = tiktoken.get_encoding("cl100k_base")
    num_tokens = len(encoding.encode(string, disallowed_special=()))
    return num_tokens

[polar wigeon]

How to properly handle responses that contain double quotes (") when using structured json_object response format? Everything is getting truncated after the first instance of the double quote. Is there a difference when using strict json_schema?

[buoyant citrus]

Hi, I'm running assistants threads on my code but i get this error even if the request is the first one i am making

[{, "error": {, "message": "Thread thread_ey25cCtgH3wqinE5ZqIUbmVT already has an active run run_4yASNAYnvvhzBxSoX4TbD8hN.",, "type": "invalid_request_error",, "param": null,, "code": null, }, }]

[ionic torrent]

These are for two different responses to my prompt. I already selected which response I preferred, but the first image was soooo bad. If I give feedback with the thumbs down on #1 and thumbs up on #2, does the feedback apply to their respective images or to both 1 and 2 at the same time? Such that if I only gave feedback on #1 would that also apply to #2?

[noble hatch]

How do we activate the Advanced voice ? I am a plus user but I can’t notice any changes

[prisma aurora]

hi openai support team: please help to solve the voice talk feature in the GPT store as soon as possible. thanks!🙏

[violet current]

Do you have the same problem? It keeps occurring these days

[rigid musk]

Often it can be vpn , addon , antivirus . Try first clean your browser cache

[gilded slate]

is there any way to get the tokens usage by a assitant or thread?

[ocean geode]

does anyone have the same problem?
the normal voicemode dont work on the NON EU version for custom gpts....
(my result: it thinks and generates an answer... and did not say the result... if i continue saying it gets appended to my first voice message... without generating the new response)

[severe rapids]

Yes I just got this today

[violet current]

thanks

[severe rapids]

I did this but no avail

[prisma aurora]

seems lots of people have the same issue. i had mentioned this problem a few days ago. it happens because of the advanced voice model enrolled, but the advanced voice model has not been integrated into all the custom gpt in gpt store. and i am wondering when will the Openai team solve this issue asap.

[dim brook]

I noticed that my plus account, still can't use the advanced voice features, I have repeatedly downloaded the latest version and reinstalled, has anyone also encountered such a problem?

[austere gale]

Hi ! I am a little confused about the memory feature. I have been able to use it on the chat-GPT online interface, but I also use the API for other projects and couldn't find anything related to the memory feature yet ?

[sweet star]

why can i still not access advanced voice

[ocean geode]

yeah...
i i think advanced voice mode is not avalible for custom gpt´s and should use the old system without advoice...

[ocean geode]

are you in a country in the eu?
if yes then thats the problem....
the laws does maybe make problems in the eu thats why i now a permanently connected to a vpn to USA
@sweet star thats maybe the same reason you dont have it... (if you have plus)

[blazing rover]

Hi guys, I have a super weird problem. I've been generating default art and then different iterations of the same generation, I had a lot of cool results but then randomly when I went back to my text box, a lot of the generations were gone? Why would that ever happen, I'm sad 😭 It was just an elemental bird.

On top of that, ChatGPT is now generating very different style to what it was generating before this happened even though I'm giving it the same reference ID. Very confusing 🤔

It's like all the perfect exact generations I had completely disappeared and the text prompt isn't even there for them

Also, just to avoid this in the future, in regards to issues like this:

1. Does it matter if I have too many tabs on the left? Does it clog ChatGPT in a way?
2. Is it bad to use the same 1 tab to generate only very specific stuff in it

Oh wait, I'm stupid, it was something to do with expanding a 1/2 option but I'm pretty sure I never clicked that before leaving my chat box

[onyx kraken]

None of that should be a problem, though any 1 chat can get so long it is no longer possible to use.

You can usually carry info to a new chat and kinda 'clone' the other.

Is all well now for you?

[sweet star]

I am in the US. On the website, it said “download the ChatGPT app on iPhone to access advanced voice.” I deleted the app and redownloaded it but it’s not working. Anyone know why?

[sweet star]

No I’m in the US

And I have plus

[ocean geode]

then i have no idea sry...

[rugged wing]

literally never used the api before and put $5 in to my usage credits

[half iris]

same here

[gritty crag]

Hi! If I was double billed for my gpt plus subscription who would I contact?

I can't find an email on the website

[onyx kraken]

They don't use email.

https://help.openai.com/ Bot bottom right. Select messages then pick through the choices, closest match to your issue until you see 'report a bug'. Type in your situation, at first you discuss with a bot. In a few hours to a few days, a human will continue the conversation and be able to help you.

[gritty crag]

I emailed the support email and they responded saying that I am not eligible for a refund, even though I got charged 3x the normal amount 😦

[onyx kraken]

I'm curious, as far as I know there is no openai support email. I recommend you DM modmail, share the email link and ask if it's an official and correct email.

Modmail's top right, top of the users list in this Discord.

[onyx kraken]

I'm still curious. You mentioned you couldn't find an email on the website (that makes sense, they do not typically communicate with us in general by receiving emails from us) then you mention having emailed them. Where'd you find that email to contact? I hope you weren't tricked/scammed.

[vast coyote]

Hello,
Our ChatGPT Teams doesn't work since friday. Is there an outage ? From the status website we don't see anything wrong and we paid all the bills (none are active right now). Anyone else having issues with it ?

[inland basin]

Greetings,
just curious if there is any way how standard users can test out O1?
I received a message that I can gain access to API O1,but cannot seem to find out how to connect to it via API ( = I did not find the correct model name and where exactly to put it) nor ChatGPT does offer it

And also another question,is there a way how to programatically submit a multimedia file to API?

[pure pulsar]

Wondering if anyone has any clues with this one
We are completely unable to create ChatGPT accounts using our school email domain. I've been trying on and off for months, and have also been trying to get support on this issue. I've had it confirmed several times on the Google side that it is not anything at our end - it's at the OpenAI end.
One of the help recommendations I did get was to review our domain verification, but I have absolutely no way to identify whether any has even tried this. Also, the message I get every time I try to create an account: "Signup is currently unavailable, please try again later." isn't one that appears to be tied to this area.

Any ideas how I can resolve this/escalate this so I speak directly to someone who could dig deeper and actually help resolve this?

[charred parcel]

you can reach out to https://help.openai.com which is the only way to reach support

[amber jetty]

I am in need of help. Over the weekend I was granted access to the new voice application. It’s now gone. Did I miss something? I’m a plus member.

[brisk tide]

Well are you in one of the countries where it isn’t released yet?

[amber jetty]

Nope. United States.

[amber jetty]

United States.

[brisk tide]

Well okay that’s weird

Maybe a rollout issue and you will get it back in a few hours/days if not you could try asking support for help

[amber jetty]

Will try that! Thank you.

[brisk tide]

I lost the new ui on my iPhone this weekend but I just suppose it’s because I am in the eu

But could be a deeper issue too but as long as there is no statement I would guess it’s a temporary thing

[amber jetty]

I’m sorry. That sucks. 😔

[brisk tide]

Yeah welcome to being an ai enthusiast in the eu

[junior jay]

Hey guys,

I’ve created a product of 50+ bots, but sometimes different bots don't work. People get a 'GPT not found' error and a retry button shows up. It's not a link issue—the link works fine—but when you send the first prompt, the error appears. I can't pin down the issue to one specific bot, and it’s random which bots work or don’t work for different clients. And everytime I test them, the bots work just fine.

I’m thinking it might be a VPN issue, browser extensions, or maybe some clients are using the free version of ChatGPT and run out of tokens at some point. It could also be a server problem, or maybe they are just too many bots to try at once?

Any ideas on what the problem could be? Has anyone else encountered this?

[fair dew]

Is there an issue? I can't login. edit: it works again.

[stone dawn]

how do i start with /imagine?

[fluid latch]

Hi, please I need help

So i was added to a team workspace and got an invite so I can have access to gpt pro. I was asked to sign up or log in with my email, which i did, but it is not still reflecting

The thing is that when I created my chat gpt account, i signed up using google, the google account is the same email address that received the invite, so i had to log in with google when I got the invite, I don't know if that is what is causing the issue.

i tried creating an account with the email, but it says email already exists as expected.

please what can i do?

[timber shadow]

Hey! I think that's a command for using Midjourney, which is unrelated to OpenAI. You could ask about Midjourney over in #ai-discussions! In this server, you can use OpenAI's DALL·E 3 five times a day in #image-bot using the /draw command.

[pallid crane]

Getting a hell of a lot of instability in the browser today. conversations just erroring out/not loading or inputs not sending.

[rare tree]

Hello all,

I just need a little help with an image issue I was running into. I'm just trying to see if someone could point me to the right bot instead of my just bouncing around and trying to figure it out.

Basically I gave Dalle a prompt for the first time and I was floored and loved it. I'm wanting to use this for a Discord profile banner but I'm attempting to have the outside areas removed and just focus on all of the middle and someone format this to 600x240 pixels.

Sorry, if this sounds confusing. But I'm just trying to see if I can get pointed in the right direction. Everytime I attempt to change this it basically completely changes this whole picture.

[ocean geode]

any way to

1. change manually in the normal voice mode?
2. get more time for advanced voice?

[ocean geode]

the reason why i have gpt +?
Advanced voice mode...
more gpt 4o (currently programming a py api version of the chatgpt voice mode with no limit for gpt o4)

[slender island]

guysss,i cant open chat gpt on android

after i updated to the new version

[slender island]

@verbal crag

i am using android 8.1.0

it was wokring well before i updated....

[split reef]

Why don’t I have access to the new voice mode yet? It has been over a week

I’m on plus

[slender island]

does your app open ?

[split reef]

Yes

On iOS

[slender island]

i am not being able to open on android

[split reef]

Tried reinstalling the app?

[slender island]

it was working well after updating

already did it

[split reef]

force restarted phone?

A forced one differs from a normal one

[slender island]

hmmm

i didnt restarted my phone

[split reef]

You can see how to if you google “how to force restart <phone model>”

I advice a force restart as that will fix more issues compared to a normal restart

[slender island]

i am using a old huawei that has google services,but i dont think that should be the problem

i was in the latest updated on the phone and chat gpt app was working well

its a yorn or something close

[split reef]

Just try a forced restart and that will probably fix the issue

[slender island]

i will try that

omggg

@split reef your were right

its working now,thnksss !!

i freak out for a moment

[split reef]

No worries, glad it works now

[ashen willow]

Hello,

I recently subscribed to ChatGPT Plus on iOS, but I'm encountering the following error message:

'This subscription is associated with another OpenAI account.'

However, I don't have another account. Could you please assist me in resolving this issue?

[candid wind]

Does anyone else have this problem? This message keeps appearing in the specialized GPTs

[timber shadow]

Sometimes I'll get this in a new chat shortly after updating a GPT's instructions. Usually refreshing, or going to a different GPT and then coming back, or some combo of these, will make it go away. I think there's just a little delay sometimes where it knows a GPT was recently updated, but it doesn't know that you're already in a new chat with it, if that makes sense.

[candid wind]

thank you

[eager peak]

#announcements message

its been 8 days and i stilldont have the new voice model

when?

[lucid hazel]

Hello. Newbie here. I keep getting the attached message whenever I fire up the desktop app.

Anybody else get it too? I'm on ChatGPT Plus on MacOS Ventura 13.7 on Apple Silicon.

[timber shadow]

Hello! Welcome. Looks like macOS version 14 or higher is required: https://help.openai.com/en/articles/9275200-using-the-chatgpt-macos-app

System Requirements: macOS 14 and Apple Silicon (M1 or better)

[lucid hazel]

Aahhh... I was looking for such an article. In fact, I even asked ChatGPT about it. See attached. 🤣
Anyway, thanks for your help!

[untold badger]

Anyone know what the best way to put data like this into the vector stores. I have 400 different products. When I use the default chunking, it has some problems like associating variants with the wrong product. Is there a way to perfelctly chunk it for 1chunk:1product. Or any better ways to do this?

{
    "Health Food - Protein Bars": [
        {
            "title": "Built Puffs 12ct",
            "handle": "built-puffs-12ct",
            "description": "Box of 12 Built Puffs in various flavors.",
            "type": "Health Food - Protein Bars",
            "variant_options": [
                "Brownie Batter",
                "Mint Chip",
                "Banana Cream Pie",
                "Churro"
            ],
            "nutrition_info": "Sure, I'm here to help. Please provide the product description or nutrition information that you need assistance with."
        },
        {
            "title": "Alani Nu Protein Bar 12pk",
            "handle": "alani-nu-protein-bar-12pk",
            "description": "Treat yourself to Alani Nu Protein Bars for a snack. Each bar contains 15-17g of protein with low sugar.",
            "type": "Health Food - Protein Bars",
            "variant_options": [
                "Caramel Crunch",
                "Rocky Road",
                "Munchies"
            ],
            "nutrition_info": "Total Fat 7g\nSaturated Fat 4.5g\nTrans Fat 0g\nCholesterol 15mg\nSodium 150mg\nTotal Carbohydrate 16g\nDietary Fiber 5g\nTotal Sugars 5g\nIncludes Added Sugars 4g\nProtein 16g\nVitamin D 0mcg\nCalcium 90mg\nIron 1mg\nPotassium 160mg"
        },

[timber shadow]

Sure thing! With a model cutoff date of October 2023, this is the kind of thing ChatGPT is prone to hallucinate about - which means it's a good use case for its ability to search the web!

[lucid hazel]

I wrote a reply but it got deleted because I referred to the "you[dot]com" site. 😦

I basically was saying I didn't even know one could use ChatGPT to search online, because "you" doesn't allow it and that's what I had originally before moving to ChatGPT Plus.

Regardless, I didn't specifically ask it to search anyway, so it wouldn't have done it anyway?

Thanks again!

[timber shadow]

Yes, with ChatGPT you have access to some extra tools that the model can use in the middle of a chat. You can ask for a web search, you can ask it to write and execute some python code (like if you wanted it to do some math, for example), and you can ask it to use DALL·E to generate an image of something - all of this can be done within a chat.

As for how to get it to search the web, it is usually best to specifically request for the model to do this ("Search online for..."), but it's not impossible for it to "make this call" on its own, depending on the phrasing of your query.

[lucid hazel]

Thanks for the advice. Will definitely keep that in mind!

[timber shadow]

Absolutely! Come back whenever you have more questions.

[slender island]

Guysss, can everyone vote on this poll I did, so open ai sees it : https://community.openai.com/t/chat-gpt-does-work-on-huawei-devices/805024?u=richardpinewood

[cobalt yarrow]

Hi, I have renewed my chatgpt subscription, I have the payment that was put in the Subscription but I still don't have access to the plus, it's as if I hadn't paid, has it ever done it to anyone?

[fresh flume]

Did ChatGPT break? I cannot get a single code block. Instead, I get about 20 different code blocks.

[slender island]

Ask him to get the intire code in a single request like "Hey chat gpt, can you give me the intire code without changing its structure"

[twin schooner]

Hi, how do i get the seed number on a picture created in the web midjourney

[timber shadow]

Hey! That would be best asked in #ai-discussions - this channel is for OpenAI stuff

[fresh flume]

I tried that like 50 times. Including in new ChatGPT instances. I think this may be a bug. I've only see this with Obsidian Templater code.

[torn ether]

Does anyone else have this glitch where GPT repeats memories multiple times??? It stored this random piece of info about my cat and when I went to delete it, I found a ton of memories stored like this.

[formal brook]

Question on the macOS desktop app under settings -> chat bar it has a reset to new chat set for 10 mins any ideas why I'd want to do that?

[heavy bolt]

Any particular reason my ability to regenerate results on one prompt is gone and has been for over a week?

[odd sparrow]

I am developing an software that allows users to upload documents and create stuff in Excel (will be more specific in DMs). However, I am encountering a challenge with the AI's ability to process and export large amounts of data effectively. Currently, the AI either completes the task with significant inaccuracies due to mishandled data or fails to process the data altogether. Along with that, we need to figure out how to best get the AI to understand each cell within the contet of the file as a whole. If you can help, please DM me.

[prisma aurora]

OpenAI team, I found we can use the voice feature in the custom gpt in GPT store today! Thanks for solving that issue. But the voice mode is not the new advanced voice, and it is not that smart. We hope the advanced voice could be enrolled to all custom gpts. Thanks❤️

[rustic geode]

How do I turn off the autodelete and retain my conversations forever? Is there a way to pay for storage so this doesn't happen?

[dim sage]

the chatgpt iOS app doesn’t work for me, nobody has helped me out yet

[pale vault]

What is the best way to report GPT being used in scam/fraud/phishing attempts?

[burnt elk]

is the upload file causing error for any of you?

[cobalt yarrow]

Hi, I have renewed my chatgpt subscription, I have the payment that was put in the Subscription but I still don't have access to the plus, it's as if I hadn't paid, has it ever done it to anyone?

[lucid hazel]

Not sure how long it has been for you but when I first upgraded, it hadn't registered the sub for at least about 5 minutes afterwards. Only when I rebooted my browser and re-logged in I had the full access.

[lean meteor]

Hello! Is there anyone with extensive experience using platform.openai.com? If so, could you please contact me? I have a question to ask.

[charred parcel]

ask here

[green dust]

Hi I'm new to chatgpt, can anyone kindly helpme: why the question "what model are you using" will receive different responses in chatgpt's chat and the developing dashboard?

Very strange

[eager peak]

#announcements message

its been 9 days and i stilldont have the new voice model

when eu users get the new voice???

[eager peak]

@rustic marlin

[eager peak]

^^

[trail cairn]

within the coming weeks and months

[rustic marlin]

currently there is no set date for EU users to get the advanced voice model

[eager peak]

ur.. the eu regulations....

[dusky topaz]

Hi everyone, I'm trying to clone a game for education phase on AI.
I'm stuck and AI couldn't generate correct codes or i couldn't explain myself to AI. Can i share my problem on here? (Sorry for language problems.)

[fading robin]

its normal.

[fading robin]

Ok simple, Models don't know waht they are so they hallucinate what they are. in web they can mention a GPT-4 just because it may be included in system prompt.

[green dust]

Yeah hallucination is a problem.

Perhaps we need to restrict the model by telling it not to respond such questions via system prompt

[urban siren]

I cant upload any files to my Custon GPT anymore. A short while ago I could upload some pdf but a lot of them failed but now I cant upload anything, not even 1kb .txt file. I get my console spammed with "Failed to load resource: net::ERR_BLOCKED_BY_CLIENT" . Any clue why that is? When I try to upload to a Chat it sas upload Try again after 6:33 (in 3h) is there some kind of file upload limit? Do you really have a limit when paying 20€ a month to only upload like 100 pdf or 100Mb or something when creating a CustomGPT?

[rustic marlin]

sounds like a cool sideproject, but it would probably help if you have at least some basic coding background

what game are you trying to remake and what language are you using? python probably?

having some basic coding knowledge would allow you to spot some of the subtle mistakes the AI can make, that kind of problem are often simples msitakes that the AI just do not correct itself, but if you nudge it i nthe right ddirection, it is able to continue

[formal brook]

Can o1-preview not create files or something?

[rustic marlin]

for example.. the AI probably makes fully functional code, except that maybe it gets one lib's name wrong, and it would work if you fixed that lib name.. but the AI does not fix the lib name because of possibly deprecated knowledge

[rustic marlin]

it cannot handle files currently

[formal brook]

Ah that will be why it keeps giving me various excuses lol

[rustic marlin]

yea xD

[formal brook]

did you have any thoughts on this Q at all? is there a "best" setting on the macOS app for that?