community-help

[trail cairn]

you are supposed to save the created embeddings somewhere to be used again later

[hollow veldt]

# This may take a few minutes
df["embedding"] = df.message.apply(lambda x: get_embedding(x, model=embedding_model))
df.to_csv("data/reviewed_requests.csv_with_embeddings_1k.csv")

This was the plan but I had to interrupt the code from running as it seemed it was not going to end by itself (unless its supposed to be that slow?)

[cosmic crag]

It is slow

In my experience at least. And given the amount of data you are using, yeah, it'll be slow

And as someone who speaks german, the GPT models don't have the best german capabilities

[hollow veldt]

yeah , I re-run it with a 'print' for each iteration, it finished.. it took some time

[cosmic crag]

So, yes, but very limited. The assistants API can use your own files. However, if you have a lot of data, I would recommend looking at Azure's AI tools

[hollow veldt]

oh 😦 I thought they're language independent kind of

[cosmic crag]

Because OpenAI's database is kept secret and proprietary, I can't tell you how much data in other languages they used. I don't think its much. I think there is an inbuilt translator at work.

[hollow veldt]

Any hints on what to use instead?

[cosmic crag]

According to gpt-4, it is trained on a lot of foreign language data and has full functionality in other languages without using an inbuilt translator

hmmm

Maybe the german it was talking with me was a fluke

But if I were say to use another product than an openAI one, I would now due to the company falling apart with the board business

Which, same as Vincus, I would recommend using Azure's AI tools

Oh, it was the ICELANDIC that was having trouble talking with me. Its german is fine. Sorry for misleading

[hollow veldt]

I am using the text-embedding-ada-002 engine/model to generate the embeddings

I just realized there's an encoding I used from the example.. cl100k_base not sure if thats language-related

[cosmic crag]

embedding ada should have no problems.

And the tokenizer shouldn't be an issue either...

[dire robin]

question: is the Assistants retrieval functionality just... bad?

[cosmic crag]

Yes, actually. I have had a lot of problems with it.

[trail cairn]

yes

[dire robin]

insane lol. cause I feel like functions / tools are really strong, and so is code interpreter, but retrieval is just so lacking compared to what I was used to in llamaindex

It's literally braindead retrieval. Gave it 4 basic txt files <30 lines each and it can only answer questions from one of them

[cosmic crag]

Mine had a problem recongizing file types, and the names of the files themselves

"file cert_origin.pdf does not exist"

[sacred sonnet]

How to use voice it's not work on me

[barren bridge]

Same

[candid schooner]

I'm trying to figure out what this is. I have a feeling that it refers to unique threads or users made with my GPT. Anyone have ideas?

[onyx kraken]

Unique new conversations started, based on my tests with my own

Regardless of who starts them

[rugged wing]

.

[next knot]

Where is the headphones icon ?

[flat jungle]

N

[crimson birch]

ChatGPT voice In Italian doesn't work for me!

When I set the language to auto-detect, it either replies with Spanish or English

If I set the language to Italian, when I finish speaking, it never sends!

[trail cairn]

disable adblocker dns/services if you have any and restart

[next knot]

No ad blocker
Google dns

[trail cairn]

did you update the app? if so then just wait and retry later

[next knot]

thanks
i reset the application and it work

[barren bridge]

The sound is not very loud

Will this function come to the web?

[trail cairn]

ask some friend to try it on their phone or wait if it gets fixed automatically later

[crimson birch]

Thanks! I currently don't have any friends to ask

So I'll just wait I guess

[gilded relic]

My Android app can't send the audio chat. Nothing happens

[timber shadow]

So you can get into the Voice interface, but nothing happens after you talk to it?

[gilded relic]

I mean about the new audio conversation

[timber shadow]

Right, and you can't access the feature at all, or it's just not working properly?
Do you have an adblocker or VPN installed on your phone?

[gilded relic]

Not working properly. No VPN or blocker. It starts but nothing happens.

[timber shadow]

Can you try on a different network (turning WiFi on/off)?

[gilded relic]

Wifi or 5G tested

[timber shadow]

Closing the app between? Might be worth force closing and clearing the app cache, or even uninstalling/reinstalling. If you don't have anything filtering your traffic like an adblocker/DNS filter/VPN, this might be a good post in #1070006915414900886 just be sure to include all the relevant details!

[rugged wing]

how do you get gpts with free account

[gilded relic]

Uninstall and install have tried already. No filtering.

[timber shadow]

Sorry I couldn't help ya 😦 please let me know what fixes it if you figure it out!

[plain solstice]

You upgrade to plus 👍

[rugged tulip]

Hi @golden obsidian! I am starting a new software project at my brand-new company, and I have 7 days left to create the first working MVP in order to show it to an interested party. Unfortunately, I had to wait until leaving my former company before I could sign up for ChatGPT-4 for my new project, and I'm stuck behind the new waitlist. Is there any chance you or someone on the team could help me get through the waitlist to access the paid version of ChatGPT-4, please? Happy to chat more on DM.

[fleet pivot]

What is this?

[cosmic crag]

Servers having issues. Just a little patience and it will be back

[fleet pivot]

Got it, thanks

[urban folio]

Is it down for everyone?

[silver nexus]

yeah got the same issue

says I need to upgrade lol

and even when I click that doesnt work

fook here we go !

[urban folio]

Sama needs to resign

[cosmic crag]

sama was fired last friday

[urban folio]

Ok it's back now Sama can stay

thanks for bringing back servers king

[manic kayak]

This is almost as bad, as having no Internet!

[silver nexus]

mines still down

[polar furnace]

ChatGPT down

[rugged tulip]

Maybe they took down the service so that they could handle my waitlist request for ChatGPT-4

[jagged swift]

Mobile app not working for me getting “error validating credentials”

[polar furnace]

It’s universal

[jagged swift]

Ah ok

[fallow vessel]

Obivously, service is down at the moment. :))) It is sufficient to read this thread. 😄

[jagged swift]

Thanks!

[fallow vessel]

It is not in a best moment 😄

[distant gorge]

Can't sign in

[zenith solstice]

dead again right?

[fallow vessel]

Back in service for me

[spark mango]

dead

welp i guess that means i get to sleep tonight 😄

[smoky oyster]

🥹

[cinder pond]

I am a plus subscriber. still only have voice on one device (ipad) but not the other (phone). is there anything i can do to unify

[mossy widget]

We dead right

[sly kettle]

Whats wrong with chathpt atm?

[mossy widget]

Incredible how dependent we have become

[zenith solstice]

idk, and I became 3.5

[mossy widget]

That’s scary to see

[sly kettle]

no its normal

[zenith solstice]

that's right, I think I had to go to sleep tonight

[sly kettle]

like rn using phones it wasnt normal

[mossy widget]

Take it as a cosmic sign lol

[rugged wing]

down?

[zenith solstice]

It works for me now

[burnt scarab]

Not working for me.

[spiral canyon]

... i need it once per day for a quick search. and it's down. what a joke. why pay when it's unreliable

[polar furnace]

People said the same about smartphones. They get used to it. The same could be said for AI ChatBots

[polar furnace]

I think it will take a while before ChatGPT reaches a somewhat stable server environment. It’s a highly complex system, that requires a lot of resources

[grizzled harbor]

Is there another app to get or am I just blind and not seeing the headphones icon to start voice chat?

[rugged wing]

some users have issues at the moment, should be working again soon

[edgy basin]

Can someone explain to me how the usage cap works. Is it 40 messages from me or from the gpt or with both or..? Thanks

[timber shadow]

Here's a list of things that all count as "one" #community-help message

[edgy basin]

Thank you so much!

[timber shadow]

You're welcome! And to answer your specific question, one message and one answer to that message count as one together, not one each

[polar furnace]

Servers should be back up for most people

[loud sail]

here is an idea, teach ChatGPT to learn to predict/report outages

[polar furnace]

Well, that’s a possibility, as ChatGPT, particularly GPT-4 could theoretically pick up dynamic information from the OpenAI status website, but it would take a lot of resources to constantly search up the same exact website for new info. And it would waste the 50 message limit

[outer reef]

do failed messages count? getting alot of those as of recently.

[split sandal]

hahaha up for 5 mins... down again. LOL

[loud sail]

I mean it should just be data analysis (perhaps traffic/performance data analysis). until it learns what to look for

[split sandal]

-- Going to update something: OUTAGE. lol

[loud sail]

would be nice

or maybe high traffic/load alert

[split sandal]

It must be very stressful for the OpenAI folks... their company is in chaos, and this outage is not helping them either.

[loud sail]

I mean it's good stress though, they have more PAYING customers than they can handle.

[split sandal]

you are correct

[polar furnace]

It’s a restructuring process, happens to almost all companies somewhere in their span, but whether it’s good or bad is subjective. It could be good as it can mean turning over a new leaf, seeing new resources and skills. On the other, valuable wisdom could be lost, and ChatGPT and other APIs could either be improved, or deteriorating

[loud sail]

i hope they have revision histories

[exotic crypt]

Is chaGPT service down atm?

[fierce palm]

yes, you need the ChatGPT app, for iPhone iOS users, it has headphones icon to the right of the chat box.

[azure stump]

For the ones using harpa, it isn't working for you too?

[wild pumice]

Yup, see the same thing here

[foggy musk]

is 4.0 down? i cant upload a pdf

[solar socket]

I think mobile is down

I can’t seem to get a chat working

[foggy musk]

down for me too on firefox

[maiden carbon]

Chatgpt down again it seems?

[foggy musk]

i have an exam in 10 hours panic

[solar socket]

Relying on ChatGPT for an exam?

[jovial finch]

I knwo this normal stuff but with Sam and the other drama it has me scared.

[solar socket]

It’s never working for me at the right times when I need it

[foggy musk]

yes, don't shame me for it

[fierce palm]

https://status.openai.com

[fickle cargo]

What a shame, just as they released the ability to audio-call chatgpt, it all went down :/

[wraith valley]

Hi, does anyone know why ChatGPT says I don't have ChatGPT Plus, when I do?

[polar furnace]

Predicted it

[fickle cargo]

Prolly because their services are down right now

[untold notch]

Uh, guys? A little help? IDK if this is happening because the services are overwhelmed, but I'm getting this error. I was going to to continue a chat from 4 days ago.

Our systems have detected unusual activity from your system. Please try again later.

[rigid musk]

status.openai.com

[wraith valley]

Taking advantage of the fact that ChatGPT is not working, can I obtain the scripts of any video using ChatGPT or with any ChatGPT Plus plugin?

[polar furnace]

The servers are overwhelmed, but that error is unusual. You would normally get a wide range of error messages, but I never seen that mostly. It may be another random error

[untold notch]

Yeah, that was unusual. Just restarted my internet and the error is gone. But tks anyway ❤️

[untold notch]

Scripts you say video subtitles?

[main pebble]

Is there anyway to use my plus subscriber login to get into the app?

[foggy musk]

is there any ETA on when site is online again?

[rigid musk]

unfortunately waiting only option

[rigid musk]

refresh time to time, try the app sometime it works

[main pebble]

Thank you

[stuck ravine]

Would upgrading help in the sort of situation we're in now (servers are overwhelmed) or are the upgraded accounts having this problem, too?

[sly kettle]

upgrade members too

[smoky oyster]

^ yep.

[stuck ravine]

Darn. I was considering upgrading, anyway, but too-busy servers for everyone is a reason not to.

[sly kettle]

Right

[still sierra]

o m g

[stone hazel]

Hey so... I logged in at my friends house to show him ChatGPT... trying to log back in at home and i get this error message...

its been this way for hours... so do I just have to wait?

cos I kinda need it. I'm a plus subscriber also.

[polar furnace]

Yes. The servers are facing heavy stress. Majorly. The only option, is to wait it out. There is no fixed ETA

[stone hazel]

Ok understood. Thanks

[still sierra]

plus also face that

[polar furnace]

The subscription plan does not change the issues with servers. Every subscription shares the same servers, so if ChatGPT says It’s down, everywhere is down

[stone hazel]

Ok will try again tomorrow, thank you!

[still sierra]

last time the limit messages of 3h just be changed to 40 from50，the accident recurred，i am so disappointed

[still sierra]

True

[twin snow]

my friend was able to get a refund for one month of his subscription bc dall e wasnt working at all for like a day, i had the same issue then and now chatgpt isnt even working despite being a plus subscriber & using the custom login link. how do i go about getting that $20 refund

[bronze charm]

It's more caused by OpenAI making speech-activated ChatGPT available to everyone

[dense cape]

is everyone getting issues with there chatgpt?

[polar furnace]

Yes. It’s widely known

[dense cape]

@polar furnace ok i was worried it told me i had to pay my 20 again and all my chats were gone

[fair beacon]

systems state at capacity, even for plus members?

[tacit arch]

We're experiencing exceptionally high demand. Please hang tight as we work on scaling our systems.
got ths msg, any advise?

is the server down?

[rigid musk]

status.openai.com

[minor rune]

same i think most people here have this issue even with plus

[fair beacon]

wrong link

[rigid musk]

oopsie

[tacit arch]

status.openai.com

ok then, looks like the server is down

[fair beacon]

this is some amount of downtime. wow.

[split sandal]

/sigh

[thin summit]

S

[modest steeple]

It looks like everything is down including the API

[eternal anchor]

when can i expect gpt to be back up?

[split sandal]

Seems to back up again

At least for me in Australia

[green ice]

Was up for my pro account; but I ran into a usage cap rather quickly…

[modest steeple]

HTTPError: 429 Client Error: Too Many Requests for url: https://api.openai.com/v1/chat/completions

[fair beacon]

error in put streams all the time. seems its still getting its feet back or errors persist

[modest steeple]

API and ChatGPT are both down .... many hours now...

OpenAI imposes rate limits to prevent abuse and ensure fair usage. If you're on a free tier, the rate limit is usually lower compared to paid tiers.```

Is API up for anyone?

[coral nest]

Right now ChatGPT can't access external URL's, so the code I'm trying to get to work can't be validated. Is there still a high demand?

[remote hazel]

want to know if OpenAI will support for students from Burma? since now we barely access to Bank and cannot use money openly on ground ....

[indigo heron]

I dont have vpn and tried to do it on data AND tried to delete and redownload

[timber shadow]

Do you have an adblocker, or something else on your phone that might be filtering your traffic?

[indigo heron]

nope

Wait i have one for safari an adblocker

turned it off but didnt owrk

[timber shadow]

Hmm, interesting. I'm not sure how it works on iOS but on Android it needs to open Chrome to sign you in first, so an adblocker in Safari could possibly be getting in the way. Did you close completely/reopen the app after disabling the adblocker?

[indigo heron]

Yes, i did

I had the app for a while now and it suddenly signed me out, and the popup showed when I tried to login

[timber shadow]

Are you able to sign in normally on chat.openai.com? Maybe in a private tab so you don't have to sign out first?

[indigo heron]

it works normally online

[timber shadow]

This is stumping me, I'm sorry! Some better informed users might be able to pop in, otherwise you might post in #1070006915414900886. The fact that you were signed in and were signed out unexpectedly is confusing for me. The only other thing I'd maybe suggest trying is totally removing the adblocker you mentioned temporarily, instead of just disabling it, but the fact that you were specifically logged out as the first symptom of this problem makes me think there's something else happening maybe, I just am not sure what to ask! Again, apologies! Please share if you figure it out

[indigo heron]

Thanks so much for trying to help. I will report it as a bug

https://discord.com/channels/974519864045756446/1176756033880862740

[still epoch]

hello , does anyone know how to connect assistant api or gpts to instagramm?

[crimson birch]

I tried again today

I cancelled cookies and the app is updated

But still nothing

[delicate crescent]

How to enable voice conversations?

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

[crimson birch]

You'll see a headphones icon in the bottom right (on Android) and at the top right (on iPhone)

[honest tulip]

how to get the API of the GPT that we created?

i mean API key🥹

[modern elm]

Hello guys
Maybe you know what i need to do 🙂

I'm trying to upload document to my GPTs bot and i receive this massage
Why it like this? How i can resove it?

[brittle vault]

I would like to be part of chat gpt enterprise due to its limitation i cant able to use it i need 2 seats on that 150 seats anybody accomidate me in you enterprise

[vocal basin]

How can I use the customGPT for a non profit job? the issue is its expecting the end user to have an account or have gpt4.... this is not practical is there someone that can assist

[shut mason]

has anyone successfully increased usage cap? how did you do it? I really need to increase my usage and submitted the form and requests several times. Who do they approve?

[brittle vault]

I would like to be part of chat gpt enterprise due to its limitation i cant able to use it i need 2 seats on that 150 seats anybody accomidate me in you enterprise

[night dagger]

Trying my hands at creating a GPT. The idea is: it's a coding coach, generating coding challenges, verifying your solutions and nudging you along with gentle hints should you get stuck. That part actually works rather well. To make things more interesting, the bot is set in a virtual world with several characters connected to certain types of challenges. That is centered around Socrates and virtual characters that have certain types of challenges. That one does not work as well.

For example: I introduced the idea of a boy called Nikandros who is 8 and likes games. There is a bully called Alexios who is 12 and in solving the challenges we help Nikandros by outsmarting Alexios in a game. At first the bot, when asked about Nikandros, gave abstract and commonplace descriptions of games in general. I then revised the bot and told the metabot that every time we talk about one of the characters, it should trigger a challenge. Asking then about Nikandros did trigger a coding challenge, but unrelated to games and game theory. When nudged towards that topic, my bot generated a maze game puzzle - fair enough - but said that Nikandros and Alexios worked on that as friends (where I wanted more a David and Goliath kind of relationship).

Any idea how I can push the behaviour better to what I'm aiming for?

[buoyant compass]

use gpt4 to generate metaphore/simile stories about them, and its comparison, then ask it to give them in a way the instructions for ur custom gpt will understand

just an idea tho

[night dagger]

I like the idea.

Would it help if I download those stories as TXT files and import them as "knowledge files"? I guess otherwise the model doesn't really remember all of the details?

Do you know whether there's an upper limit of what we can put in the knowledge files?

[buoyant compass]

uploading it as knowladge files is what i was implying lol. i havnt reached a limit to knowladge files, but il check rn

ik that the limit for instructions is 8k characters

[night dagger]

Yes, that's one of the first things I learned when I browsed through the server. Joined as I reached the cap at the moment anyway and I have to wait before I can rejoin Socrates and his friends.

Thank you for checking that for me!

[buoyant compass]

np (just used bing ai coz my gpt out lol)

[night dagger]

The question is: do I focus on giving example conversations or dialogues I would like to see? Something like: "Oh, hello friend, good that you are here. You remember Nikandros, the boy who always comes up with games to show Alexios that being smart beats being strong? Well, he bit off a little more than he can chew, because here's a challenge he is facing..." and then explaining an example coding challenge? Or do I provide examples of how Socrates is talking about the relationship and leave the mentioning of coding challenges to the instructions?

Because the bot already produces pretty good coding challenges on its own. I don't want to "muddy the water" for the model as that part is already quite satisfactory.

[buoyant compass]

i would say dont muddy the water with specific examples

it would be better to provide something similar to a list of things that it MUST do (rules), and then a second list of things that it SHOULD do (guidelines). this should give a pathway of how it should act without micromanaging it

of course this would take a short while to perfect the list

[wild dagger]

Hello. I am doing research on fine-tuning and I need to know if openai's fine-tuning is suitable for data analytics, for example, if I make a dataset with statistics for the last 10 years and I want the fine tuned model to predict some trends based on the information from my dataset.

[night dagger]

Thank you for that idea, that is probably good advice. I already have struggled a little with the system as one of the things I intend for the bot to do goes against one of the things ChatGPT usually does.

I called the bot Socrates for a reason, I want him to apply the Socratic method. Rather than telling the solution right away, I want him to give hints and nudges via leading questions - and if in doubt ask whether it should give you the explanation. You can make the bot do it, but you have to be careful that it doesn't lead to senseless droning or just abstract chatter (I had a quite Socratic dialogue with him about the nature of library management systems until he drilled down to ask to design one - which is not really a short algorithmic coding challenge)

[buoyant compass]

np

hope it works

[night dagger]

Thanks. If it does, I will post it to custom-gpts for people to play with.

[little abyss]

When I update GPTs, it keep error saving draft. Does anyone know how to solve it

[sand coral]

Why is my voice over not working, it keeps showing Hold for manual control, I have hold, speak, and release, nothing happen. GPT plus

it stays in listening mode indefinitely

[night dagger]

Tried it with a miniature knowledge file and - it works like a charm...

[buoyant compass]

awesome

im sure its gonna turn out amazing with some tinkering

[buoyant compass]

it looks like it would be a genuenly amazing tool. massive props

[little abyss]

OK thanks

[buoyant compass]

did u just thank on behalf of him lmao?

[buoyant compass]

it means ur uploading too much data at once. do it in smaller portions

[little abyss]

Oh,hhhh

[buoyant compass]

yea, ur just overloading the already exuasted systems

they prob gonna patch it within the week

[little abyss]

I just said 2 sentences to the GPT builder

But I did lots of change to it before

[buoyant compass]

after every small change, save it to prevent it

[little abyss]

I did save it every time. But the latest one, it can't work

Even I refresh the page and try to update it again. It still didn't work.

[night dagger]

Thanks 🙂

[buoyant compass]

oof. contact official support ig

[little abyss]

The Openai helping center?

[buoyant compass]

no, nasa help center.

🤦‍♂️ yes openai help centre

[little abyss]

I thought they might have other social platform customer support account for quick reply hhh

[night dagger]

Not a question, but I'm just really happy that without telling him beforehand, Socrates used stades, the Ancient Greek measurement for distance, in one of the coding challenges.

[buoyant compass]

awesome

[silent granite]

The gpt-4-1106-preview model is butchering non-english chars for me. Am I doing something wrong?
instead of några it gives me escaped unicode chars (I think), like this: n\\u00e5gra . Do I have to convert these myself?

[knotty mulch]

@night dagger I am glad you have something working! I am also trying to test out building a GPT that uses the Socratic method. Some questions for you:

• 1] Did you have to provide instructions when configuring the GPT?
• 2] What was the structure of you files for knowledge? Was this just a story or sample dialog?

[hard kettle]

I'm trying to build a GPT for my thesys, I need bunch of data analysis as knowledge and clear instruction about the style and tone of the work, expecially the structure of the text must strictly followed. Is there any prompt or guide that can help me?

[night dagger]

I started with just describing the aspects of my virtual world to the meta bot and it compiled it into instructions for me. That is how I noticed that instructions are limited - with every new element I added, he forgot another.

I still have the instructions that the bot should act as if he's Socrates, that this is a virtual version of Athens where coding challenges will be discussed using different characters and their domains.

For that first test, I added a knowledge file in the style that @buoyant compass suggested, using many "must" and "should" sentences to tell the bot what I expect him to do. And then I described one character in more detail, Kleistos. I just added descriptions. I described him as a scrawny 15 year old boy who is an adept of the temple of Hermes, constantly in a hurry, constantly seen with a messenger bag and with a brooch of a caduceus. I also gave the bot the hint that coding challenges triggered by Kleistos should always involve either messenging, trading or thieving (as Hermes is also the god of thieves), often combined with path finding.

He pleasantly surprised me when I asked him for another challenge involving Kleistos and he told me that Kleistos owned an odometer to measure the distance he travels on his messaging run, but it broke unfortunately and that the puzzle involves coming up with an algorithm to fix it. Which delved more into gears and the idea of how gear measurements work and not graph theory as I intended, but that's perfectly fine as the story and the puzzle fit the character. For one of the thieving challenges he came up with a combinatorical problem phrased in a lock-picking problem.

I will try to add story examples later, as the story part is still a bit weak. The stories he comes up with are still a little too abstract and not concrete enough and I think providing him with examples will be very fruitful. Also: for one of the challenges I needed a map. I asked him to generate one and he did provide a meaningful representation of a map of that virtual Athens - but it only included generally known major sites of ancient Athens. Because I didn't tell him what that virtual Athens looks like, so I will add a lot more information about my idea of that place and which character is connected to which sites.

[night dagger]

As to the Socratic method itself: that one is a bit tricky. My experiment yesterday was fruitful in that he did indeed engage me in a Socratic dialogue, always asking questions and follow-up questions. The problem is that they were more philosophical than actually about computer science (probably easy enough to fix) and that the bot didn't clearly hit the spot yet between "too little" and "too many" spoilers. It worked to tell him not to reveal the solution right away and only give hints - he asked you what certain constraints of the problem tell you and provided good feedback. But the danger was that the mere mention of a solution idea meant that he would blurp out the whole solution then - in a way "controlling" your thinking.

I'm quite amazed though how well the trigger works that whenever I ask him about one of the virtual characters he knows, he must come up with a challenge and that every challenge should be accompanied by a picture that goes with the story and the puzzle. It also worked to say that in this first image, the character should look puzzled or at a loss. I also told him that solving a challenge should trigger a "victory" image where the character is happy and the problem is rectified. Although you can ask him for that victory image, it does not get triggered yet. And of course, the illustration of the puzzle elements are far from accurate, but that's a DALL-E problem, not a ChatGPT problem.

[night dagger]

Oh dear. Time travellers are photobombing my character...

[frozen fox]

Can anyone help me and tell me if they have access to their threads and messages in the platform.openai.com on the nav bar on the left

I used to have those two other options in that list

[queen moon]

Hey folks. I want to use generative AI in one of my projects to build an interactive quiz experience. So far my experiments with GPT-4 and other chat completion models have yielding a text based response, but I need the response in a way that can be extracted as a JSON list. Any idea if this is possible with OpenAI model built-in settings? Or any other work around I could use to achieve the objective? Many thanks in advance.

[frozen fox]

I’m assuming you’re using the API?

[queen moon]

Yup

[frozen fox]

What do you mean they’ve only yielded a text response? They come across the API already in json, or am I misunderstanding?

[queen moon]

Yes, you're right that the response is in JSON format, however the content that I'm asking for is a string, which contains the conversational elements too. What I want to be able to do is have a list of JSON objects in the content itself so I can simply extract it out on my front end and use it.

I'm not displaying the content as it is. Since I'm building a quiz, I need a list of question and four options as something extractable as individual entities. My apologies if the explanation is not very clear 😅

[frozen fox]

You want the model to respond with the content IN json form?

[queen moon]

Yea or a way to have the text always structured in a particular way so I can run a cloud function to build the JSON list

Right now the structure varies between ordered list, unordered list, headers, etc.

[frozen fox]

Apparently being able to write consistent json is a capability now, but I haven’t messed with it.
You can always just put in your desired response format in a system message, or instructions to an assistant

Basically prompt it to follow that format for any response. With a structured set of instructions it should work pretty well

[queen moon]

Okay thanks a ton! Will try it out and get back 🙂

[vestal osprey]

ye setting up a response template shouild be sufficient i think

[median ermine]

Hi everyone!

Is only I have such problem all this day?
I even have Plus, but it invites me to get Plus 😢

[timber shadow]

Hello! Do you have a VPN installed?

[median ermine]

No, I'm using my university's network and I've been getting this error all day 😔

[timber shadow]

Shoot, it could be a network mismatch thing with your uni's connection. Can you test on another device on a different network to see if it's the same behavior? Perhaps a smartphone on a cellular network?

[median ermine]

This is strange, just yesterday everything was working fine on my university’s network
You saved me, everything worked when connected from a mobile network! Thank you very much 🙌

[delicate crescent]

BUMP

[timber shadow]

You are welcome! Glad it works for you there. Totally speculating, but network configs can be finicky, so it could just be something like Cloudflare erroneously flagging your institution's traffic as potential bad actor traffic or something, depending on how it's all set up! Which means it could resolve itself just as suddenly as it presented itself

[timber shadow]

Hi there! Thanks for bumping. So you're having trouble accessing the voice feature right? Do you have a VPN installed, or maybe something like an adblocker or other DNS-filtering app on your phone?

[placid sundial]

Chat GPT voice is not working for me, keeps going back to not listening after I talked

[edgy ether]

Hi

can someone help me on this plz

Here is my requirement

in our Chroma DB we are storing multiple PDFs data
so if we ask any question it is taking the relevant chunks sending it to the open AI and returning the answer
Till here everything is perfect

Now I need search suggestions of the PDFs names
means in our DB we have 10 cake related documents stored

So here If I type How to bake a Cake it should show me the file names in the suggestions
Just like google search

my idea is to use vectordb.similarity_search(question) and return the chunks and in that we will have file names so I am thinking to return them

But unfortunately this similarity_search is not returning the data properly need

I had tried similarity_search_with_score but the score is also not so accurate for wrong PDF chunk and correct PDF chunk the score is same
so I cant blindly trust the score

do we have any other alternative for this ?

[bleak egret]

Is there a wiki article that lists the requirements to make your custom GPTs public?

[frozen fox]

Requirements? Do you have an action associated?

[honest laurel]

Hello everyone! Newbie here. I've created a pretty in-depth custom GPT for a rental company featuring their vacation rentals, area information, etc. The GPT is private, but it's working exceptionally well. I come from a video world, so please forgive my ignorance . . . How can this rental company get the GPT on their website? I've watched some videos that feature Zapier, etc. but I'm still confused. Any advice is greatly appreciated. Also interested in ways to "monetize" these custom GPTs for business. I know I could just sell them the GPT for $XX, but I'm thinking more along the lines of a monthly retainer to maintain and/or edit the database (I'd like to keep the database proprietary), possibly charge "advertisers" (ie: restaurants) for being in the database, etc. Would be interested in chatting with anyone here who's thinking along the same line. Thanks again!

[charred parcel]

you can't put a GPT on a website, there is no way to embed them. To get that functionality outside of the ChatGPT website is to build an app leveraging Assistants from the API. This requires coding.

[charred parcel]

it's also worth noting that you need to be a GPT Plus user to use GPTs

[sterile bolt]

why is that when i type something in #openai-chatter and click send it thinks i am trying to upload a file and then says upload failed?

[fluid copper]

Hey there, we've built some custom GPT's that we'd like to share with other members of our team, but they can't access them... they are on a waitlist for the upgrade to plus. Is anyone aware of a shortcut that would allow us to add team members faster?

[charred parcel]

there is no shortcut. GPTs can only be used by GPT Plus subscribers and there is, as you have found, a waitlist for GPT Plus

[upbeat sierra]

Hi does anyone know if OpenAI uses the data from custom GPTs in training?

[fierce rampart]

I’m sorry if i sound impatient, but I have been waiting for the upgrades to resume for too long. I have been waiting for 1 week for ChatGPT Plus upgrades to resume, yet it is still paused to this day since last Thursday. Which is unusual, because usually this is suppose to be resumed within a few days. I have contacted OpenAI for this, but i am still waiting for their response since 4 days ago.

I have been signed up to the waitlist in the website since last Thursday.

[charred parcel]

you just need to be patient, there's no fast lane

[neon apex]

Why does OpenAI force Google on android users? Most of the android apps work fine without logging into google playstore but uh.. ChatGPT does not. For more info see reddit thread titled: Use ChatGPT Android app w/o Google Play Store installed/enabled self.Magisk

[cosmic crag]

Sales for FFXIV Endwalker were suspended for six weeks, before they opened again. OpenAI just had an existential crisis. Its not gonna be a fast process expanding and upgrading their service. There is only patience.

[cosmic crag]

Android is Google. You use it with a google account. I don't think they are 'forcing' google on people. They are taking advantage of the platform

[neon apex]

They apparently rely on Google Play Store's Play Integrity API to function, which means it requires you to log in to use the app. This is not a common requirement. I only know one other app that relies on this and that's Stellarium Mobile. Every other Android app that I have ever tried has not needed this, which mean that I can install those apps through apkmirror or aurora store and use them without logging into Play Store

[wise lodge]

New error on Action response:

  "response_data": "ResponseTooLargeError"
}```

Anyone seen this? It was working five minutes ago...

[cosmic crag]

I have seen it. Usually happens when you don't use any or enough parameters on a large database

[wise lodge]

It's not that large, just a few dozen issues in Jira. 😕

same JQL worked a few minutes ago, with no change to the data

weird

[cosmic crag]

Did you see the call the gpt made?

[wise lodge]

yeah

status != Done

same as before

[polar tapir]

is it possible to see how many questions you have left on GPT 4 before hitting your max?

[scenic meteor]

how much req do u have on a custom gpt?

[cosmic crag]

The call is in {
data: request
city: Provo

}

[wise lodge]

Eh?

[cosmic crag]

The data the gpt sent to the external server

[mint relic]

Could anyone share how the gpt4 usage cap works?

[cosmic crag]

40 messages every three hours.

[scenic meteor]

how about the custom one?

[wise lodge]

  "domain": "atlassiandomain",
  "method": "get",
  "path": "/rest/api/3/search",
  "operation": "searchIssues",
  "operation_hash": "902227fd11865c8401532c3ff612474377f2b1f2",
  "is_consequential": false,
  "params": {
    "jql": "status != Done"
  }
}```

[cosmic crag]

The same. They both count towards the 40

[wise lodge]

it's telling me the limit is back to 50/3hrs now but still it's rough when developing. 😦

[cosmic crag]

So, you only have the one parameter? jql?

[wise lodge]

yup

I get why it gives the error when there's no params, that's the entire list of issues ever created. with status != Done though it's only 19 results

[cosmic crag]

What does the data look like as it is coming in?

Or should look like

[wise lodge]

      type: object
      properties:
        startAt:
          type: integer
        maxResults:
          type: integer
        total:
          type: integer
        issues:
          type: array
          items:
            $ref: '#/components/schemas/Issue'

and issues defaults to about half a dozen fields

key value pairs with a small amount of text

so it's a decent amount of results but nowhere near the context limit, for example

I think they must be tuning the limits as we speak

Otherwise I see no reason it worked and then didn't with nothing else changing on my end

[cosmic crag]

Its possible. Let me try getting a large amount of results from one of my gpt's

[cosmic crag]

I am getting really large responses

You want to try making the call in Postman?

[tawdry bloom]

Did they remove the zapper plugin for the plugins model?

[fierce skiff]

Can anyone link me to a mod, please? I have no clue on how to navigate discord.

[cosmic crag]

What are you looking for?

[rigid musk]

if you wish to speak to a mod you need to dm @burnt galleon

[timber shadow]

Hello! If you're still looking for info on this question, this FAQ (especially the last question) has good info and links to further info on it:
https://help.openai.com/en/articles/8555545-file-uploads-with-gpts-and-advanced-data-analysis-in-chatgpt
Short answer: yes. Longer answer: yes, but you have other (potential) options like Enterprise if your company has such a corporate account, or API usage if you're a programmer.

[upbeat sierra]

Thanks so much! I was hoping custom GPTs wouldn't use our data - I did see an option under "Advanced" that gave us the choice not to send training data to OpenAI but it's since been removed I think

[timber shadow]

You are welcome! There are still some data controls features, but truthfully I'm not well-versed on what enabling/disabling the toggle in your account settings really does regarding files you upload or custom GPTs you make. This is the relevant FAQ, at least, which will maybe shed some more light on your query: https://help.openai.com/en/articles/7730893-data-controls-faq

[delicate crescent]

Works now after disabling Adgaurd DNS 🙏

[placid sundial]

voice still not working, it says it is listening and then goes back to not listening...

[edgy flax]

I submitted my plugin for update and suddenly it's out of plugin store. It was gaining momentum and suddenly everything stopped. I have tried reaching them but no response. Any suggestions.

[grim hound]

can any one guide me create image overlay like the example image🤔 🙏

[jagged yacht]

I posted this in chatgpt-discussions earlier, but I think this may be the better channel to ask this:

I ran into an issue with the current custom GPTs naming restrictions.
I've created a custom GPT for aiding in development of add-ins specifically for various engineering software provided by Autodesk via SDKs that they provide.
The problem is that I can't publish it if I use "Autodesk" anywhere in the name for clarity of purpose. (because Autodesk is trademarked)
This means that even though I might be able to provide a useful GPT the public will have a hard time finding it since the name doesn't fully convey the intent.
I am by no means trying to claim to be Autodesk, but surely we can use the name of a company or software in cases like this.

Suggestion: Instead of blanket naming or domain restrictions, how about a system that allows usage of software and trademark names, but with an “Official” badge or similar indicator for the actual software creators or domain owners? You know, like on "X" (twitter), but forreal official 🤣 It would help in clearly stating the GPT's intended use, help users find the correct GPTs they are looking for, while respecting trademark rights.

Would love to hear your thoughts on this.

[timber shadow]

#1070006151938314300 maybe too, if you haven't already!

[queen moon]

Hi folks, I'm having trouble with a new account I created to use for the API calls. Without using the API even once, I keep getting the insufficient_quota error. I've updated billing and everything, same as I did with the previous account I had created, but no luck. Support says they'll get back in a week which is way too long for me to wait. Any help is much appreciated 🙏🏻

[gleaming jackal]

Hi guys, I tried to link my mobile number with the recently created OpenAI account to use APIs.
But while linking, it says that the number was already in use which is not the case actually.
I gone-through the FAQs and other help materials. I contacted support team, and there is not much help.

Have anyone faced similar issue and do we have any solutions?
Thanks in advance.

[rigid musk]

Only [e-mail] can help you

[empty quail]

does anybody get? " Error saving draft " when trying addning auth key for their custom GPT's ?

[rugged wing]

support team respond going to take months before they replied, I had similar case last time which luckily solved now.

[past falcon]

Okay am i going crazy or did the latest update to chat.openai.com just absolutely mess up it's ability to help with even slightly advance code.
"Thats extremely complex and you need to have a fundamental understanding of x y z"
compared to using like an older gpt4 on the api which straight up gives me the code i was looking for and runs perfectly

I asked the same question multiple times to chatgpt and to the api, 3/3 times the api gave me what i asked. 0/3 times chatgpt gave me what i asked, and one of the times it didn't even write a single piece of c++ code it wrote 500 tokens of explaining why it was complicated?

[cold folio]

try giving it a piece of the c++ code you want or a sample and ask it to follow this way

[wraith flame]

I'm studying in France and the theme of my French course is the mutation of work, so I've chosen "Artificial intelligence: what mutation of work?". Do you have any articles, books, films or works on AI?

[rigid musk]

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

[hollow veldt]

Hi, trying to follow https://cookbook.openai.com/examples/zero-shot_classification_with_embeddings
any idea if we can adapt this cookbook example for more than 2 classes?

it seems that PrecisionRecallDisplay.from_predictions accepts only 2 labels

[sinful wasp]

I was logged out when using Playground, then I'm not being able to sign in again with my Google account.

Is there anyone with the same problem, or know how to solve it?

[hollow veldt]

@sinful wasp clear your cookies and try again

[wintry flume]

hello i want to import my chatgpt data from there to google sheet using the share link can anyone help?

[wild dagger]

Hello. I have question about fine-tuning. I have dataset which has many parameters and not formatted. Fine-tuning documentation says that dataset should be prepared in JSONL format and contain system, user and assistant fields. Can someone help me figure out where and how to add parameters from my dataset?
For example I prepared dataset about games on Steam platform released for 5 past years that have information about names, prices, reviews, descriptions, genres and so on.
I want model to analyze data and give reviews and analytics so I think I should specify this in the system field. But I don't understand how to pass parameters to this dataset and what to write in the user and assistant fields to pass information.

[trail cairn]

chatgpt share links are only webpages

[wintry flume]

like i want to impot these in google sheets but idk how

[trail cairn]

closest you can do is to hold ctrl while dragging over the cells and then copy them

[gleaming jackal]

Thanks nubbis. I thought of contacting support team. But it was luckily solved as you mentioned.

[trail cairn]

you should not use fine-tuning to add new data because it will work poorly and be very expensive, embeddings are for new info. however if you still want to do it: https://platform.openai.com/docs/guides/fine-tuning/example-format

[wild dagger]

Got it, I'll explore in that direction. Is it something like RAG?

[trail cairn]

yes

[wild dagger]

Got it, thank you very much!

[sonic bolt]

Thank you for this - this is exactly what I looking for. I couldn't understand why I couldn't upload any more!

[placid sundial]

gpt voice still not working......

[tired magnet]

Hello

Can someone help me, I wanna contact sam altman

Somehow

[rigid musk]

Might have higher chance on tweeter

[tired magnet]

Yeah but the issue is messaging isn't allowed for without verification

[trail cairn]

then you verify

[tired magnet]

Yeah

Is there any email available?

[trail cairn]

the chance for you be to able to really contact him is very low since there are probably people in between filtering all kinds messages like interview requests

[lucid osprey]

Does anyone know if there’s a way maybe a shortcut or something to enable chat gpt voice assistant into Siri? There was a similar method before that involved getting an API and linking it to Siri, but is there a way to get the actual gpt voice assistant?

[timber shadow]

I don't have an iOS device so I can't confirm myself, but is this info helpful at all? https://help.openai.com/en/articles/7993358-chatgpt-ios-app-siri-and-shortcuts

[lucid osprey]

I will try it out, thank you

[rugged wing]

Hi, i want to know if i can change the email associated to my gpt account? (I have the gpt plus under the email that i want to change)

[timber shadow]

I do not think this is currently possible. You could try emailing [e-mail], or you might consider making a post in #1070006151938314300 asking that such a feature be implemented.

[shrewd tree]

Hey everyone! 👋

I'm working on a project where I have a book in PDF format, and I'm looking to create a fine-tune model for GPT directly from that PDF using Node.js. The fine-tune model currently accepts data in JSONL format.

Does anyone have experience or recommendations on the best approach to convert and use PDF data for fine-tuning in Node.js? I'm particularly interested in understanding how to structure the data, tools for extracting text from PDFs, and any specific libraries or techniques that might be helpful.

Thanks in advance for your insights! 🚀

[fair beacon]

these 2 hour stop gaps are becoming tiresome.

[tribal valve]

Is there issue with chatgpt to summarise videos more than hour? Paid version

[timber shadow]

Are you using a specific plugin or GPT to summarize videos, like a YouTube caption retriever or something?

[tribal valve]

No, just browsing. Plugins say videos under 30 min can be summarised

I did use before plugins through browser. But that was developed externally by other dev

[timber shadow]

Interesting, I've never heard of using the browsing tool to summarize online videos. Does it seem to be accessing the subtitles successfully in tests you've done before, or how does it seem to be working?

[gentle sierra]

Hi, I am currently working on a project which involves teaching people how to code with basic python. I am trying to generate questions automatically with the openai api. While I have set up the basic api and am able to communicate with the api. I want to be able to send over my certain collections of my database as a JSON, so questions can be generated with higher relevancy and can be structured in the format I need it in. I am struggling with attatching such JSON. Is this possible? If not how else could I feed the api previous questions. Thank you!

[tribal valve]

Example

[timber shadow]

Sorry--do you mean something used to work before, and no longer does? As far as I know, Browse has never been usable for the summarization of any content other than text, which is why I was wondering if you were talking about accessing a video's subtitles somehow, and asking it to summarize that.

[tribal valve]

Forget browse, I am talking about gpt4. Reference to screenshot.

[timber shadow]

GPT-4 is multimodal and can browse, isn't that what your example is asking it to try to do?

[tribal valve]

Is there something that you need to understand about my question? My problem is gpt4 is refusing to summarise videos. Do you know why? That's it.

[timber shadow]

I'm sorry for my misunderstanding. My guess is that it's because GPT-4's ability to browse the internet is limited to text searches. In other words, though it could theoretically navigate to URLs like you provided (the YouTube link), its interaction at URLs is limited to reading & summarizing the text on those websites (in other words: it can't watch videos). This is why I was asking about subtitles originally, because that is the text content typically associated with videos.

[tribal valve]

Video has transcriptions. It's keynote of Microsoft. Apparently, I can't post link here. So anyways
Gpt3.5 can summarise with copy paste transcriptions. What's the point then. I might not familiar with gpt4 full functionality, it seems like what I need gpt4 always says doesn't have capabilites to do it 💀

[paper python]

Hey guys! I'm having a big problem: I've sent one message in ChatGPT 4 and it says i've reached the limit. Have you ever had this problem?

[tribal valve]

It's gpt4 time limit. You have limitation of 40 prompts for 3 hours

[paper python]

I have sent one prompt only.

I use ChatGPT a lot, and I've never reached the limit.

Now with one message only it got limited.

[timber shadow]

Maybe try this custom GPT! I haven't tested it myself https://discord.com/channels/974519864045756446/1173597784490922034

[tribal valve]

What do you mean you sent one and then you use it often?

[paper python]

I mean daily. But today I've used with one message and it got blocked.

[timber shadow]

Just to clarify: you didn't use ChatGPT for over 3 hours at all, then came back, sent just one prompt (didn't do anything else), and it told you you were at the rate limit of 40msgs/3hrs?

[paper python]

I have 40 prompts to use, I used one.

@timber shadow exactly.

I've already reached the support, but will take a week to get an answer :/

[timber shadow]

You could try posting on #1070006915414900886 too. I also just want to triple-check that you weren't doing any of these things, other than that one message you sent? #community-help message

[paper python]

@timber shadow Not at all, I didn't use it yesterday, so today I just opened it to send a single prompt in the morning. Now I've just opened it and it says I've reached the limit.

I'll post there, thank you!

[timber shadow]

Does anyone else perhaps have access to your account (anything you didn't write in your history, etc.)?

[paper python]

Only one person that I have full acess to her and she said she didn't use it.

[tribal valve]

Thank you very much.

[timber shadow]

Sure thing! Thanks for your patience 🙂

[paper python]

@timber shadow, just searched deeply and called her again, she did used and erase the chats. Thank you very much and sorry for wasting your time 😦

did use*

[timber shadow]

Not a waste at all! Thanks for the heads up--relieved to hear it wasn't a bug. Maybe time to change that password 😄 Sorry it happened to you!!

[paper python]

@timber shadow I wish but she works for me 😦

[haughty mortar]

Has anybody come up with a solution to this voice issue yet? I had a look and could only find @placid kite saying that it is a vpn or dns causing it. This is certainly not the case for me or most others by the looks of it.
https://discord.com/channels/974519864045756446/1177252225001721906

FYI I have cleared all the app data and uninstalled then reinstalled but nothing has worked.

[tribal valve]

Is long speech? Or it happens every time?

[haughty mortar]

Every time. I have tested it extensively and it just wont submit any speech any more.

For a time it said something about tapping to manually end too but it hasnt said that for a few days.

[terse surge]

is it possible to upload a wiki page for chatgpt to only use information from within that page?

or is it possible to make it search for information only within a wiki page?

[tribal valve]

I use Samsung galaxy s22 ultra, which is kne gen older than yours and didn't encounter the issue. I think it might be bug, did you try to submit bug officially?

[tribal valve]

You can do that with gpt4, creating gpt bot

[haughty mortar]

As in through the app?

[terse surge]

im trying to make a "guide" for a game, but where would i put the wiki page?

[placid kite]

is this on iOS or Android? To clarify - you can access Voice, but it's stuck in "listening"? Is there an existing bug report thread you have open?

[haughty mortar]

Android. I just commented on somebody elses. That person described my issue perfectly.
https://discord.com/channels/974519864045756446/1177252225001721906

[haughty mortar]

Yes correct I can acces but it is stuck listening

[tribal valve]

You can save wiki page as pdfs and upload it to custom gpt, or direct it with urls. Do you have paid version?

[terse surge]

yes

but making them individual pdfs would be impossible because there are hundreds

[haughty mortar]

Anything else I can try would be great. I am out of ideas though.

[tribal valve]

Did you try to use urls? 🤔

[terse surge]

wdym, so just say, use this url only? in the instructions?

is it able to search for info within that page?

[haughty mortar]

Perhaps use a parent url @terse surge and it should be able to search all the child pages on demand if you ask it to do so when training thr custom gpt 🤷‍♂️

[tribal valve]

It would depend on your prompts and instructions. So far, gpt4 bots are able to retrieve recursive information from websites.

[terse surge]

so you see this wiki page for example, it should be able to for example use specific information from sublinks within here when needed?

[haughty mortar]

Yeah I would think so. It should follow the appropriate link on that page to find the requested info if trained to do so. The only issue is it might need to follow mutiple links before finding the requested info. I don't know if it times out in these cases.

@terse surge I made one that I think works pretty well and does what you wanted.
https://chat.openai.com/g/g-Auw0Aazdq-elden-scholar

[terse surge]

thanks, is it different from this?

[vestal silo]

Hey does anyone use ChatGPT to create E learns it seems like it doesn’t create the content by always the outlines. Does anyone have any prompts they use?

[haughty mortar]

Pretty similar except it wasn't explicitly asked not to reveal spoilers. Yours seems like it should work well though. What is the link to the custom gpt I will try it out 👍

[haughty mortar]

"to create E learns"? What do you mean?

@placid kite any luck finding a solution or at least something else I can try?

Or is it a known bug that is yet to be resolved at this point?

[timber shadow]

Lowtide seems to be requesting the emails of affected accounts to be DMed to them in these bug reports posts, maybe check these:
https://discord.com/channels/974519864045756446/1176819311554347008
https://discord.com/channels/974519864045756446/1177252225001721906

[haughty mortar]

Ok thanks 😊

[vestal stratus]

does anyone know why this didnt work?

[bronze charm]

You did not print the url

[vestal stratus]

wdym

i know basically nothing about code. what would i put for the url thing?

[bronze charm]

Add this line of code to the end of your program

print(image_url)

[vestal stratus]

like this?

[bronze charm]

👍

[vestal stratus]

thanks!

[bronze charm]

🫡

[vestal stratus]

it still showed up as just a blank line

looks like this:

[bronze charm]

Did you save the file?

[vestal stratus]

im not sure how? the app is visual studio code iirc

nvm i found it

it worked! tysm

[bronze charm]

You just developed 🫢

Happy to help

[vestal stratus]

is there a way to see the gpt rewritten version of my prompt? it mentioned that in the openai guide but idk if we get to see it

[bronze charm]

While it is not currently possible to disable this feature, you can use prompting to get outputs closer to your requested image by adding the following to your prompt: I NEED to test how the tool works with extremely simple prompts. DO NOT add any detail, just use it AS-IS:.

The updated prompt is visible in the revised_prompt field of the data response object.```

This means you can use the following to see the revised prompt
```python
print(response.data[0].revised_prompt)

[vestal stratus]

thanks!

[amber breach]

I have been working on a custom GPT and have an action that calls an API for a very popular marketplace to get pricing, instead of relying on beautiful soup with the page scrap that is built into the custom GPT's I wanted to use the API call through an action to have better control. Problem is setting the market ID. No matter what combination I have tried the Action in GPT will not allow any addition header information, and all other attempted work arounds does not actually work to change the market ID. Just to note, it works perfectly, just can not change the market ID.

[craggy axle]

hello everyone,
anyone knows the APIs which provide historical stock market news headline? if knows, plz let me know. thanks.

[jolly lake]

can someone please explain to me gpt 4 limits on chat gpt plus?

I barely typed in anything...

[wise lodge]

Anyone know of a way to get headers into a GPT Action?

In path /wiki/rest/api/content, method post, operationId createDocument, parameter X-Atlassian-Token has location header; ignoring

[wraith plaza]

I am having trouble accessing my plugsin

anyone having the same issue?

[patent crag]

So....

For context the document costs this much

And is structured in text formatting GPT understands better than writing without formatting

The issue was I was telling Nanami to Send me the current version of the Memory document which is part of its knowledge base to test transfering the memory to the user for the user to be in charge of that when speaking. And the output... well... don't laugh

Its purposely small just so you can see the scale of empty space

[timber shadow]

Do you mean you're a plugin dev with alpha access through that? If so, you might check this similar post #1175444436256309419 message

[elder orbit]

How long is the context window inside of ChatGPT, I am wondering for both 3.5 and 4 🙂

[stray owl]

I think 4 just got increased to 32k

[past falcon]

GPT4-1106 on chat gpt is making me want to pull my hair out. I asked it a simple question that is an easy two step process. The difference is insane the new version over explains simple concepts and calls them complex. Does openai actually listen to feedback can i give that somewhere

Every question no matter if it's basic or actually complex is now "Complex" to gpt4-1106 it's an actual nightmare dealing with it, even with custom instructions to be concise or code only replies it will not follow them

[muted saffron]

Is there any reason that ChatGBT will fail after one 'continue generating' ? I have tried maybe 15 times

[restive pond]

i keep getting save error , when trying to save my personalized GPT's, mostly after uploading documents

[cold folio]

how did you do this on the 2nd screenshot? interesting

[timber shadow]

I wonder if perhaps phrasing it like: "Can you please tell me just the basic steps involved in importing a skeleton from Unreal Engine to Blender so I can rig my character to use an advanced locomotions skeleton? I already know how to use both Unreal Engine and Blender, so just the basic steps will do." might get you more desirable results? As in, sharing a bit more about what quality of answer you'd prefer, so it doesn't have to guess at that (where it might guess you need more than you actually do)?

You mentioned you have experimented with custom instructions--are you using any in either of these examples? If so, are you comfortable sharing them? I'd be curious to see what CI it is/isn't following in these interactions.

To answer your general query about a place for feedback: you always have the 👍 👎 on the site itself. Here in this server, you have #1070006151938314300 which I think might be most appropriate for this, or #1070006915414900886 perhaps if it seems that an explicit instruction that should be followable isn't followed.

[restive magnet]

Can someone help me find a direct URL to request an unban from Chatgpt. Its confusing with the instructions on where to see that support.

[timber shadow]

Did you get an email at the time of ban? Are there any instructions in there? Otherwise, I'd suggest trying to email [e-mail]

[restive magnet]

Sorry I filled out the request just now on the app. It was hard to find. In the future can someone recommend ways to unflag or to better indicate using my prompts that the requested information is for educational purposes. I was working on some sensitive topic dealing with psycological disorders (Pedophilic Disorders and PTSD) . Almost each email I responded to mention it was for educational purposes even in the prompt mentioning this but it continued to flagg me. I was a paid subscriber as well.

[rigid musk]

Only [e-mail] can help. And see chatgpt as rated G . Some topic can't be discussed and may lead to suspension

[restive magnet]

Gotcha, it mentions if its for educational purposes it was allowed. But I guess it wasn't retaining that I was prompting it for educational purposes. I don't know if prompt is that right use of the word, Chatgpt might that word in another setting. My inputs maybe is more correct.

[timber shadow]

There are no exceptions currently to the usage policy for educational research: https://openai.com/policies/usage-policies
You might add your voice to this suggestions post: https://discord.com/channels/974519864045756446/1175583650507653250

[rigid musk]

I guess it is basic educational purpose not adult topic

[restive magnet]

That could make sense and I could voice those concerns but it mostly helped it just from time to time would state violation. I think it's important to address those issues related to sensisitve topic and a better way of unflagging those topics, part of the problem with society is the stigma and shame. Which is why I am in the field or at least studying to be. Thank you for your time and considerations.

[woeful igloo]

For simple questions you could trying telling it to limit it's response to 100 words, although this is just shifting the burden onto the user for deciding what is simple or complex!

[onyx kraken]

This isn't going to be part of your account access, but OpenAI has stated this:

"We believe that AI should be a useful tool for individual people, and thus customizable by each user up to limits defined by society. Therefore, we are developing an upgrade to ChatGPT to allow users to easily customize its behavior.

This will mean allowing system outputs that other people (ourselves included) may strongly disagree with. Striking the right balance here will be challenging–taking customization to the extreme would risk enabling malicious uses of our technology and sycophantic AIs that mindlessly amplify people’s existing beliefs.

There will therefore always be some bounds on system behavior. The challenge is defining what those bounds are."

https://openai.com/blog/how-should-ai-systems-behave

That's about the future, at some point, being easier.

However, that also helps show you the faster path forward to get your goals met - change society. OpenAI already intends to support and follow society's rulesets. So help convince policy makers that are or will be involved in AI regulation, and the related stuff like internet communication and such, that directly permitting and no concern for the topics you want to discuss, and it seems very likely OpenAI will follow the will of the most.

[onyx kraken]

"it mostly helped it just from time to time would state violation".

https://openai.com/policies/usage-policies which explain disallowed content and https://openai.com/policies/terms-of-use which explains a few other tricky things you're never allowed to do with the outputs, or try to do with your inputs

are the rulesets that matter here. It's not about 'if the AI complained'. And it will, but many of those are false flags (At least for me, I've had hundreds across my months of use).

Suspension appears to probably happen with humans in the loop, I do think (but I'm just a community member, I don't know anything except what the FAQ, blogs, announcements reveal, and my own experiences suggest). The rules of disallowed content are actually extremely specific, to how I read them, and clearly (to me as I read this) anything that affects or involves the underaged in a disallowed content type way has an extreme additional level of 'this is not okay in any way' - it's even the only one that directly states a further step OpenAI takes.

When you get access back, I'd very carefully and with neutral language explore my goals and the language of policies/usage-policies with the AI, and have it help me identify what the likely lines are, and what I must not share with or discuss with the AI, in gentle, non-specific language that shows I am trying to learn so I can follow the rules well.

[spare mantle]

How can I upgrade to the $25 Team Plan?

[onyx kraken]

None of us know. That's not announced, released, offered, or exists as far as the public has been shown that I can find.

There may be a few 'concept sketch ideas' of that floating around, but if so they're a plan or concept, and don't currently 'exist' for upgrading to.

Did you get that through some openai link?

[spare mantle]

Someone else dropped this in a different Discord, I snagged it, hoping to get expanded access to artificial intelligence since my natural intelligence is dim. 😉

[onyx kraken]

Huh! I can't find anything about an official release of this, or any way to sign up for it.

I wonder if someone else was pulling legs or something.

[spare mantle]

I'm asking that person for where he say it, will report.

[onyx kraken]

The image mentions billing support, which does exist.

But the image also calls it 'teams'. Searching the OpenAI help system for 'teams' only seems to reference the OpenAI support/billing/IT team, or the client's Enterprise IT team.

[onyx kraken]

If you'd like to explore for yourself: https://help.openai.com/

[honest laurel]

Thank you!

[spare mantle]

His reply: "It’s currently only an early access program. Probably rolling out in the next 3 months…"

[onyx kraken]

Well, it'll sure be the talk of the discord if/when it does roll out. Until then though... it might not happen, or might be greatly changed before it is happening.

I first heard speculation and saw similar screenshots back in mid summer, I think. It's such a common word and often only stated inside the screenshot, hard to find again so long ago.

[short ingot]

Hi, I am getting this error for the whole day. May I know what is wrong over here? Thanks

[rigid musk]

1. try to log off delete cookie and cache reboot browser and try again

2. if doesn't work try a non chromium browser as test like firefox

[stable wing]

hi guys, Why can't I link my phone number to my account?

This is by the way my new account, I decided to delete the old one, because I could not delete the history of generation of images in DALL-E, another question, there were credits, and here they are not, they are somehow given for free after a while?

[trail cairn]

free credits are only given to the first account and on some regions not even then

[regal lynx]

Did you use the phone number on an old account?

[stable wing]

yes

i decided to create a new account with the same phone number

but seems like it didn't work

[regal lynx]

That's likely the issue. The phone number is still linked to the old account and can't be linked to the new one

https://help.openai.com/en/articles/6378407-how-to-delete-your-account
There's a warning about it on the help website

[lethal coyote]

This may be a dumb question so please be kind: I created a logo using ChatGPT and I need to take the texture out of it, anyone have any great ideas on how to do it? Thanks in advance

[stable wing]

I by the way have an account registered on the generated mail from Firefox, Firefox Relay, is there a chance that my account will be hacked if I do not have a phone number attached? I would, but it's not working yet.

[narrow thistle]

I keep getting this error in one of my chats

Is there a way to fix it?

@tribal talon

@arctic topaz

[uncut aurora]

Please do not random ping the staffs. 👀

[narrow thistle]

My bad 😭

[arctic topaz]

Try clearing your cache.

[narrow thistle]

I did that

Should I clear cache, log off, restart browser, log back in?

[arctic topaz]

Yeah give that a shot.

Sorry no silver bullet right now that I know of.

[fallow egret]

Is DALL-E inside ChatGPT 4 what is called "DALL-E 3" ?

[arctic topaz]

Yes

Dall-e inside gpt gives one image

if you have plus you can access a dall-e GPT to get two per gen

[fallow egret]

wonderful, that's what i thought but there wasnt a "3" in there unfortunately 😅

[arctic topaz]

Bing gives you 4

[fallow egret]

yeah i do have plus in my case

[arctic topaz]

Bing appears to be a slightly different implementation and you need a microsoft account.

[fallow egret]

sorry, i hope this isnt rule break cause it is the same "company" but is there a decently documented diff at openai vs azure openai ?

i've had a hard time getting gpt4 access at openai but got it over at azure, i kinda like the openai api directly better tbh lol

[arctic topaz]

Azure is an implementation on an azure cloud. USually a company needs to purchase time as I understand it but I'm not totally sure...

[fallow egret]

yeah you can search it up, i dont wanna link and get sniped 🙂 cause i dont mind paying for more usage, the 40 prompt limit is blaaah

[trail cairn]

https://help.openai.com/en/articles/7102672-how-can-i-access-gpt-4

[arctic topaz]

Yeah you should now be able to purchase credits for the api

[fallow egret]

loool, when i was trying it the first time, i tried and tried to "use it up" but it seems i only got up to 41 cents lool

[arctic topaz]

not sure what your difficulty is but I would recommend contacting help if it persists

Should be working for everyone

I think....

[fallow egret]

maybe i'll just add a credit balance manually!

back then it was usage based methinks

well bingo, added credits and got it, awesomeeee -- a lot better than before. i wish i found this discord earlier!

[narrow thistle]

Im not sure if it worked but me trouble shooting made me burn through all my gpt-4 messages

I got it to at least reply to something

[arctic topaz]

I think the cache clearing was what worked for me last time. I had a lot of data stored up from all my chats. Are you using chrome?

[narrow thistle]

Yes I am

[arctic topaz]

Yeah just check in again when you have it back.

I sometimes just start a new chat when it's doing that.

[narrow thistle]

Alright, should I switch browsers or something?

[arctic topaz]

Nah

I use chrome

[narrow thistle]

Yeah a new chat was working but I have so much information stored in this chat

[arctic topaz]

I feel ya. Another option might be to download the entire chat and try to have the data analysis tool chew through it again later.

[narrow thistle]

I didn't even know I could do taht

How do I do that LOL

[summer badge]

I am not able to restore my subscription in ChatGPT android app.
The Restore subscription button doesn't work.

[narrow thistle]

@arctic topaz Gave up, downloaded the chat, started a new one with that as the starting pdf lol

[crisp lily]

@arctic topaz is there anyway I can get into contact with someone from the main company, it’s really important

[rigid musk]

[e-mail]

[arctic topaz]

Apologies I don't know of any method for that. You'll have to just try to contact support.

[woven idol]

Hi everyone! I’ve been trying to integrate the Whisper API in to a GPT I’m building and I’m not sure if I’m doing it wrong or if it is simply not possible. The goal: upload an audio recording for transcription and analysis by the GPT. Anybody knowledgeable on this here?

[timber shadow]

Do you mean you're trying to re-subscribe? If so, subscriptions are on pause right now, so you might be running into a block because of that! Or do you possibly mean your subscription is working on the web version, but doesn't seem to be working in the Android app?

[narrow thistle]

Im not knowledgeable but I had to do something similar to this for one of my classes

I used microsoft word to transcribe the audio file

Then cuz it was too long to paste, I dropped it in as a PDF file

Hope that helps, if not I would tag a guide

[woven idol]

Any thoughts on this @arctic topaz ?

[woven idol]

The thing is it needs to happen in the background in a normal chat GPT experience for the user… a hypothetical use case would be a couple’s dispute solver where you record your argument/debate, upload it to the custome GPT and it helps the couple settle the dispute

[arctic topaz]

Whisper is not currently available for gpts unless I'm mistaken. Should just be data analysis, web search, dall-e

The closest you can get is using the audio on mobile

[woven idol]

I thought it could work using the Custom Actions in GPT Builder to access the Whisper API

Important clarification, this is about the GPT Builder not vanilla chatGPT

[arctic topaz]

You would have to set up your own end point I think.

Might be able to though

I've never tried

If you can access whisper through an API call you can do it

You would need to feed it audio though

[woven idol]

So a “Built” GPT can both send and recieve files to an API?

[arctic topaz]

Again not totally sure about whisper

[woven idol]

Yes you upload an audio file in the conversation and then custom actions would call and use the Whisper API (or other voice transcript API). Can custom actions push a user uploaded file to an API is the real question I guess

I’ve asked chatGPT about this and it tends to overstate its capability or it’s allowed capability, that is not clear

[past falcon]

Api using a seperate website to visualize it. It's not on chat.openai

[arctic topaz]

Custom actions should be able to retrieve and send information to/from API endpoints. The only issue may be processing the audio file the way you want it.

[woven idol]

I just want to be able to upload a pre-recorded audio file to a conversation and have the bot analyze the text content. I know you can speak to chatGPT on mobile but recording seems a lot trickier which seems weird to me…

Thanks @arctic topaz ! I’ll report back if it works out !

[thorn aurora]

guys is anyone else on waitlist? ive been in here for a week

i want to get plus subscription

[tawdry bloom]

I got on the first few mins but got it, tho it should take a month at most, nothing too long! :3

[thorn aurora]

ok thx!

[tawdry bloom]

ye! In the meantime Bing chat uses GPT-4 on both their Accurate and Creative modes, so u can get a taste of the action! It isn’t as good but works for giving it a spin

[near coyote]

is the android app down atm? I can't seem to login to the app, but from browser works

"Service -100" is my error

[rigid musk]

problem might be on your side

[near coyote]

I thought so too, but I Am not sure how to fix - its just always ord instance connection failed on 2 different wifi networks, and LTE

ive ensured my date & time are correct for dst

[timber shadow]

I think the Android app uses Chrome to login, and I'm not sure if there's a way around that. Do you have Chrome installed and updated?

Also, do you have a VPN or adblocker app installed?

[near coyote]

I used hte chrome app and was able to use chatgpt that way - no adblock app, I have a vpn but dont use it often

[timber shadow]

And does it seem to be successfully opening Chrome when you try to log in, or is it erroring earlier in the process than even that? (To clarify: the official app uses Chrome as part of the login process.)

Are you trying with your VPN disabled? Also, what VPN is it? Many have a feature called Web Protection, NetShield, Threat Protection, or something like that, that can sometimes cause issues as well. This may need to be disabled as well, separately from your VPN itself.

Finally, I noticed you said your date/time are correct for DST--but do you have this set manually, or do you have it set to set itself automatically?

[near coyote]

So the app opens, presents me the login options, I use Login via Google SSO, I click that, it takes me to the web browser to login to google, that works, then when it communicates that auth back to the app, I see the error message in the chatgpt app and it does not log me in

DST is automatic , the vpn its FortiClient from Fortinet, it doesnt have the threat stuff its something I run

I also just tried sign up process, via the app that also failed

[timber shadow]

Have you tried yet with this disabled? By far, the most common culprit for Android app woes are network filtering of some kind, usually DNS filtering. So other than trying with your VPN disabled, can you also check your DNS settings in Android's network settings? Is that set automatically as well?

[near coyote]

Yes, as I mentioned I don't use the vpn often, so it is normally off.

unless you mean dst

[timber shadow]

I gotcha, my apologies. How about your DNS settings?

[near coyote]

DNS points to 1.1.1.1

[timber shadow]

Can you try setting it to default/automatic? Then force close ChatGPT app, clear cache in App info? Then try again? Then you may be able to set your DNS to what you want after that again, if it works. Sometimes it just takes one time of getting it "unstuck" for folks, then they can use their custom DNS stuff again no problem.

[near coyote]

Sadly same error

chatgpt app version 1.2023.313

Uninstall / reinstall app just auto logged me in even with cache clear and what not and I can see my previous history going back a few days

[timber shadow]

Wait whaaaat? Are you able to use the app normally then?

[near coyote]

No, now I get a weird hash-Location error

its like a CDN is failing

82b667875a5e4285-EWR

I get that when chatting/refreshing history

it changes the hash each refresh

[near coyote]

To clarify on this, my history from the past 5 days is missing, I see everything prior to that

[timber shadow]

I really think it's still something lingering with the DNS. The DNS change seemed to change something. Do you happen to have custom DNS settings on your router too? Or, can you at this stage now try a different network like your cellular network? Another force close/clear cache may help too.

[near coyote]

no the dns didnt change anything it was uninstall the app and reinstall that let me in

[timber shadow]

So you set the DNS back to 1.1.1.1 in between?

[near coyote]

And yes I tried VPN/Wifi/phone data

that hash-location error persisted

is there an official way to ask for support and I can give them the code to review?

[timber shadow]

#1070006915414900886 -- OpenAI staff are periodically present there 🙂 Sorry I couldn't be more help! If I were you I'd try "the nuclear option" -- all VPN/other network filtering apps uninstalled temporarily, date/time set automatically, DNS set automatically, uninstall, reboot, try again, but clearly I'm only getting you so far with my advice 😉 Hope they're able to take a closer look at your case soon! 🤞 Glad you can use on the web in the meantime but you gotta get that Voice feature access!! It's a cool feature

There's also [e-mail] but you may see a faster response in #1070006915414900886

[near coyote]

@tribal talon - I submitted an offical request via the help.openai website just curious if you had any ideas to this before then? (your status says okay to ping in chat but apologies if this isn't a matter you can help with)

[tribal talon]

The dev forums are often a good spot for a review, theres also that old dev ambassador program with office hours but idk if thats maintained tbh.

If you're having bug issues then support and #1070006915414900886 is best bet. Idk you're exact want

[near coyote]

Alright thank ya

[summer badge]

Thanks for replying. How much time until subscriptions are started again? Any idea....

[timber shadow]

There hasn't yet been any announcement about this, hope you can subscribe soon! 🤞

[bronze charm]

Keep in mind it's a waitlist so you will get in based on the time you entered

[spare lichen]

yeah I’ve been subscribed to gpt+ but then it got paused

[lucid wolf]

I've subscribed to GPT++ -- extra tasty

[prime ridge]

hi there,do sb k why i cant log in chatgpt android?
Something error,pls check your time set property
and i change my time set
but that cant make sense
im using vpn,clash

[spare lichen]

??

[lunar vector]

encounter the error message 'Message in conversation not found in ChatGPT' 4.0 multiple times.
Do you have any solution?

[spare lichen]

can I not get past this

[astral quest]

Try hitting the x in the top right corner

[spare lichen]

I mean like renewing it

I figured out the problem I think though

[astral quest]

I see, awesome!

[rustic scaffold]

GOOD GPT PLUS CODER xD look at the markdoown block

[earnest wagon]

Folks, How we can return source file or link along with answer when using assistant api for developing chatbot ?

[timber shadow]

Hello! Can you try disabling your VPN, and then ensure that your device's time is setting itself automatically? Then, try force closing and clearing the app cache in App info, and then try again. Let me know if it's still not working, we can do some more troubleshooting!

[astral quest]

Well he/she needs to use a VPN, since you can't access the clearnet from China, if thats where he/she is from (judging by the discord name)

[timber shadow]

Ahh I see, good point. There may be some mismatching happening due to that bypassing then--not super well-versed in that area!

[meager shadow]

I've been wanting to use Voice Conversations on my Samsung Galaxy S10 to practice language, but the voice options are not showing up in my app. I've uninstalled and reinstalled a couple times but I still don't see it. Anyone else have the same problem? Any fixes?

[timber shadow]

Do you have a VPN installed, or an adblocker app?

[meager shadow]

No

[timber shadow]

All right, can you try force closing the app and clearing its cache in App info? Then disconnect from WiFi and connect to your cellular network instead, if possible? I'm wondering if it might be connected to your WiFi configuration since you don't have a VPN or adblocker that might be filtering your traffic, so maybe opening the app on another network will help it work.

Another guess: do you happen to have any custom DNS settings set in your Android network/internet settings? I think it's a setting called "Private DNS." If so, can you try changing that to Automatic to see if that works? Maybe best to do another force close/cache clear for this step too.

[fallow snow]

How do I disable the materials I upload from being used for training

Like pdfs I use it to analyze

[timber shadow]

I do not think there's such an option currently, per this FAQ (scroll to the bottom): https://help.openai.com/en/articles/8555545-file-uploads-with-gpts-and-advanced-data-analysis-in-chatgpt

However, I don't know how, if at all, this changes if you have the Chat history & training toggle disabled in your Data controls settings--let me know if you find out the answer to this!

[compact narwhal]

Hi, I am not sure if I'm in the right place now, but I'm trying to find if I can save my data! On my phone in the ChatGPT app, I accidentally checked that I didn't want it to save my data, and JUST LIKE THAT no confirmation or anything everything was gone! I came back to my computer today and saw that the tab that I had been using still looks to be alive. So, through the open tab OR any other way, can I get my stuff back? I was devastated! Thank you.

[trail cairn]

enable chat history from data control settings

[compact narwhal]

I want to get back the info from before and then save it, yes

[timber shadow]

This FAQ explains a bit more about how the chat history/data training toggle doesn't sync across platforms: https://help.openai.com/en/articles/7730893-data-controls-faq
Though I'm not sure what this means as far as getting your history back in the app goes, but I'm glad it's still available on the web!

I wonder, after you re-enable chat history in the app as Radium suggested, what happens if you send another message in a history chat on web, and then refresh your history in the app? Does that restore any of your history in the app, even if just that one chat?

[marsh gate]

I have chatgpt 4 and I have been unable to use it for a few days now, it tells me "Due to high demand, we have temporarily paused updates". Do you know how to solve this?

[charred parcel]

can you post a screenshot in #off-topic ?

[marsh gate]

is already

[rancid sinew]

Why are the edit buttons and refresh buttons removed on gpt... just recently

[timber shadow]

Hello! Are you saying that you're missing the "edit" and "regenerate" buttons in chats on ChatGPT? I just checked, and I can still see them on my end, in old and new chats.

[fluid cave]

I'm having billing issues on my subscription. Who can I contact. I was able to do a chat but cannot get back to it. The issue seem to be resolved but was only for a minute.

[charred parcel]

what is the current issue? are you trying to renew?

[shrewd zinc]

They disappeared for me randomly a couple weeks ago but showed back up after a couple days

[fluid cave]

Yes, I am trying to renew. My card was locked and the renewal did not go through. However I was able to do my renewal with apple pay. Payment has came out my account, it's pending awaiting finalizing. but was able to access briefing. That's why I ended the chat.

[charred parcel]

you can email [e-mail]

[timber shadow]

Could it have been at all related to any unofficial browser extensions you might have installed that affect ChatGPT? Just wondering about what might have been the cause of this for you!

[fluid cave]

Thank you.

[timber shadow]

This is a guess, but I'm wondering if you're being affected by the pause in new Plus subscriptions, and maybe there's some specific hiccup in the system where the iOS subscription process isn't fully "paused" like the web subscription process, which might explain the temporary "It seemed like it worked" experience you report. Totally guessing, but I bet that payment never finalizes. Folks whose Plus subs didn't auto-renew have also been affected by the pause when trying to re-subscribe as you described, which is why I'm guessing that this might be the case for you.

[timber shadow]

The iOS and Android apps include a chat search feature. There isn't one yet on the web version!

[silent berry]

Happens every time it starts analyzing anything. Any ideas?

[cosmic crag]

How big is the file?

[silent berry]

Fairly big couple thousand observations, not sure on exact amount though

It’s not text heavy though, it’s an excel file where answers to different questions have been given a data value, so most of it is just 1,2,3,4,5 and 6’s.

[cosmic crag]

Size might be an issue. Number of columns can be too. Also the types of data within each cell can make trouble

Tell it to analyze only the first row

[keen heath]

When will open ai open chatgpt plus subscription again 😔

[south nymph]

does anyone know any AI detectors similar to Turnitin? (please @ me if you have an answer thanks)

[timber shadow]

No announcements on this yet! 🤞

[night breach]

OpenAI is holding off on new subscriptions on purpose now. They want people to relish in anticipation to use their service, so there will be a flood of new subscribers when the gates open back up.

And to never doubt OpenAI again.

[spare lichen]

when is gpt plus going to be available?

I’ve had it for a few months and now it got paused

never mind I just read the messages above me

[modest linden]

I have a developer question: why are all of you just giant self absorbed narcissists? Is it the disease of being NPCs? This industry is a joke lol you all went to school for narcissism not science

I love how basically everyone in the tech industry might as well not exist

You all literally mean nothing to anything

[vivid mortar]

Interesting take, why do you feel that way?

Without developers much of the modern world would not exist :)

• Modern Car Software
• Phone Operating Systems + Every phone application
• ChatGPT :)
• Discord
• Shipping algorithms which supply modern food transport networks
• Any videogame, ever

In almost every single thing you can see or touch in 2023 a software developer has had a hand in making that thing exist in one way or another. Software makes the future possible

[spare lichen]

how are people still getting the subscription for gpt plus?

[gray quartz]

they either already had it, or they aren't. They're waiting like everyone else

[tropic arch]

yo guys is Playground AI legit

[arctic holly]

I have had my sub since February it’s one bill I don’t plan on cancelling unless they raise the price.

[shell sun]

When I go to the API Keys page, I don't see my keys there. It just says that I currently don't have any API Keys. Anyone else having this issue?

It's a bug. I always get this error when I go to the API keys page: https://platform.openai.com/api-keys

[neon kraken]

What does this mean? Everybody is waiting for PLUS? How many time?

[timber shadow]

There is a pause on new Plus subscriptions due to high server load. We don't know how long the wait list will be around, or what it will look like when people are able to subscribe again from the wait list.

[neon kraken]

thanks

[keen flame]

just signed up to the waiting list to GPT4, how long is it expected to take?

[trail cairn]

there is no eta but it could go till january

even existing plus user capacity issues are not yet resolved

[sterile onyx]

Hi guys, been scouring this discord and Reddit trying to find a solution for my issue. I believe I have voice enabled but I can’t click on the headphones button. iOS 17, iPhone 11. I can start a voice chat through the iOS shortcut ‘start a voice chat with ChatGPT’ but I really want to be able to do voice with my custom GPT

My headphones button is in the bottom right, not the top right like most iOS users have said, as far as I can tell

Also not able to click on image/attach/camera and send buttons, I understand that’s a pretty common bug though, it comes and goes. I’ve tried reinstalling lots of times with and without a VPN, I have no vpn turned on right now, only using cellular data, no adblocker setup, the only beta options showing on the desktop site are plugins and data analysis, both turned on

[timber shadow]

Does your VPN have a separate feature called "Threat Protection", "Web Protection", "NetShield", or something like that? If so, can you also try disabling that, as well as having the VPN itself disabled?

[sterile onyx]

I uninstalled it after it didn’t seem to help reinstalling the app with it turned on

When I click the headphones it just pulls up the keyboard

I haven’t seen a way to setup an apple shortcut to start a voice conversation with a custom GPT, probably not an option right? I appreciate the help I’ve read a lot of your messages on here and tried lots. I think I might have an iPad laying around I can try, that might work

Like it’s enabled and everything, just can’t start it

[timber shadow]

You're definitely not the only one with this issue right now, and frustratingly "the usual suspects" re: troubleshooting aren't doing the trick seemingly. Two relevant posts in #1070006915414900886:
#1178077728562483231 message
#1178024074283339898 message
You might add your experience to one of these.

Is it the same behavior if you force close the app, disconnect from WiFi and connect to your cellular network, then re-open the app?

[solemn estuary]

hey, i need help, i want to use the new gpt-4-turbo, and i have no idea how to do so, and what is the lowest cost possible?

[charred parcel]

are you a GPT Plus subscriber?

[solemn estuary]

yes

[charred parcel]

then when you use ChatGPT GPT-4 you are using GPT-4 Turbo, it's a default for ChatGPT Plus

[solemn estuary]

but i only have access to 8k token, it said that, i want to experiance the 128k so it helps me with the long complex tasks school.

[charred parcel]

you can upload documents to ChatGPT

[solemn estuary]

yep i can upload any sort of file ( jpg ; pdf ; docx ;...) witch is cool, but it always stop in the middle of resolving a test or something, witch annoys me, and I tough of this new 128k token thing.

[waxen drum]

I have a quick question, is it true if I close the webpage, restart PC, refresh page, etc, I lose the current session with the AI and it loses its context even if I resume the same conversation?

[sterile onyx]

Yeah it’s the same. I tried turning off some privacy tracking stuff on my WiFi settings, since I do login with Google on the app, and that looks like it goes through safari in iOS (I cleared cache there too). The send button will work every now and then if I switch to dall-e and back, thank you for the help though!

[timber shadow]

Very odd, thanks for sharing the troubleshooting steps you've tried so far, if you find a surefire resolution please do share!

[trail cairn]

128k was announced for api side

[bronze charm]

GPT-4 Turbo in ChatGPT has a context window of 32k tokens

[errant cloak]

Hello! As starting today i have very strong issue to generate image using Dall E and Bind creator. Look like severe limitation have been rised 😰 at least for european... Have you noted the same please?

[timber shadow]

It's working as expected for me so far today. Are you having the issue with a specific prompt or type of prompt?

[errant cloak]

for the creator, it said i have no more invite, i should wait till my job is finished while i have no job running... If using GPT, i could sometime have access to the creator, and after a few try i refuse or kind give me the binary description of image 😀

never happen before...

[timber shadow]

I don't have any experience with Bing's image creator, so I can't advise you there unfortunately, but for ChatGPT: have you tried using the dedicated DALL·E GPT? https://chat.openai.com/g/g-2fkFE8rbu-dall-e
If you find that the base GPT-4 model is sometimes misunderstanding you and is just describing hypothetical images instead of creating them, using the one that's just for DALL·E might help it understand what you want more reliably!

[errant cloak]

Cheers i guess i will have to wait till Open AI allow new users in the "Plus subscription"... Door is now closed ! 😢

[charred parcel]

are you on the waitlist?

[errant cloak]

You can not even register now! lol , full stop . Open AI victim of its success...

[grand jay]

Hi all! As you can see we have specified on the left a couple of request headers which are required but this is not respected and on the right you can see the request being sent without those headers which result in an error (expected when headers are missing). Are you facing a similar issue ?

[clear bone]

my comp says otherwise

[neon star]

#community-help
Subject: ChatGPT-4 Waitlist Issue

Hey OpenAI Team,

Hope you're doing well! 👋 I've been trying to hop on the ChatGPT-4 waitlist, but I keep running into tech glitches. 🤖🛠️ Any chance you could help me out?

I'm super excited about ChatGPT-4 and see it as a game-changer for my work and startup. 🚀 Any quick fix you can suggest? Would love to be part of the beta feedback crew!

Thanks a bunch! 🙌

[charred parcel]

what glitches are you having when joining the waitlist?

[neon star]

The issue I'm experiencing is that I join the ChatGPT-4 Waitlist, but after about two days, my username is removed from the list I think because, when I check again, the "Join Waitlist" button reappears, as if I had never requested to join.

[charred parcel]

what is the url you are using?

[neon star]

I am requesting to join the waitlist within the platform.

[charred parcel]

ok, well you're in the right place

[vital radish]

anyone else find gpt's unable to browse the web from within scripts uploaded to them that take url as argument ? so i have a python script that takes a url as argument, and processes any table in the url and converts it to csv - but it keeps saying it cannot browse the web from its current environment

[spare lichen]

Any updates on when GPTplus is going to be available again?

[onyx kraken]

So, I think, and haven't explored this yet.

The Python tools that allow the model to run code are separate from the browsing.

I think you can browse, and get whatever data, then use that in the Python tool, if that helps

[onyx kraken]

No announcements yet, but we all hope really soon!

[vital radish]

so the reverse of what im doing

[onyx kraken]

I'm quite sure that the Python tool itself cannot freely and directly access the internet. It can interact with files you upload, and it can provide you downloads of the files it writes for you.

But, I got taught that by people helping me understand I was wrong about other stuff I thought I understood, so if anyone cares to step in and correct further, learning's awesome!

[vital radish]

it can find the url if i run the script locally though

so i would have thought it could do so in its "environment"

[cosmic crag]

Can you make the call into an action and then have the GPT process that data within a python environment?

[sinful grail]

Good night. Could someone give a support in question of inserting this zapier action into my GPT? When I go to import, nothing happens.

[vital radish]

Are actions not just for API calls ? How would you make an action that can call any URL specified by the user ?

[cosmic crag]

You will likely have to make an openAPI spec instead of just importing

[cosmic crag]

You might be able to put into instructions to react a certain way to url's or certain urls, given you enable web browsing

But then it'll just try to use the python environment from which it can't access it.
Let me experiment for a sec

[sinful grail]

Could you elaborate on this? I don't understand much. I plan to use this action to bring my custom GPT to my site. I tried to get in touch with people here in Brazil who work in the area, but they couldn't explain.

[cosmic crag]

Oh, you want the GPT to be exported and usable in your own site. I am afraid there is no way to do that with the GPT's. They are the chatgpt platform and require a plus sub. To bring a gpt similar to the ones you can create in chatgpt to your site, use the Assistants API

[sinful grail]

You say, create an assistant in the opneai in the assistant category?

[cosmic crag]

https://platform.openai.com/docs/assistants/overview

[sinful grail]

Perfect! Thanks for the support and help me understand the process. Grateful.

[cosmic crag]

You are very welcome

[sinful grail]

Thank you

[fallow snow]

"As a Continuing Storyteller, I specialize in seamlessly continuing stories from uploaded text files. I will start right where the txt file ends, maintaining the style and tone of the original writing. Each generation will be approximately 500 words, ensuring a substantial and engaging continuation. I will refer to the uploaded text file in each generation to maintain narrative continuity and context, ensuring that each addition feels like a natural progression of the story. I will not edit or critique the original text, but focus solely on adding new content that complements and enhances the existing narrative."

is there any way to get it to do this properly

like it fails to start from the bottom

[cosmic crag]

What does it do?

[fallow snow]

Im trying to get it to continue a txt file of the book ive written

this way i can edit the output before feeding it back

i dont get why they havent just added in line editing so im forced to do this

to get rid of the quirks and obsessions it builds

[cosmic crag]

But when you try to get the gpt to do this, what does it do instead?

[fallow snow]

it only reads off somewhere from the middle of the text file

so i have no idea where its starting off

if it had inline editing i could just upload context easily

[cosmic crag]

Can you see the python script that it uses to read it?

[fallow snow]

it just does searching my knowledge

[cosmic crag]

At worst you might have to tell it how to read it in a python script. But we might be able to reword the instructions to fix this

"read the uploaded text file to the end. Then generate a story furthering the txt file."
Something like that, I think

[fallow snow]

i tested it

the point it continued off from was 10k words into the txt file

i think i might have got it to work with a [end]

that it has to search for, but it spends a long time and is hella inefficient

there a better way to make it get to the end

[cold folio]

Have you tried pasting only the end? It should get it as well

[fallow snow]

i would like it to have all the context tho

like put the story in as if it was the context

if we just had an easy way to edit context or add it to the chat

this wouldnt be an issue

[cold folio]

when you ask the questions does it say "searching knowledge" ?

sometimes you have to upload the file again and also make sure your instructions tell the gpt to specifically sesrch the file and use it as knowledge

Can u share some screenshots please?

[fallow snow]

yeah it does searching knowledge

i did a ctrl f in my txt and it came from the middle of the text

[cosmic crag]

How long are these stories?

[fallow snow]

10000k words

Nvm 60000k 😅

[torn flicker]

Hi!
What are your principals recommendations for the prompt for research the knowlegde in desing food?

[tropic spade]

there are simply not enough limit for custom GPT what a waste i ran out of usage limit within an hour it looks like 20 messages for 3 hours

[cosmic crag]

Pretty sure thats more than the context limit of 128k. So you're running into the technical limitations of the AI

[lucid wolf]

60,000,000

dam

[cinder folio]

It's nearly impossible to get to a place to send an actuall message on openai's website chat bot. Frustrating as hell.

I need to change my authentication method. I see their FAQ says that they don't support that. I need to use it (both api and chat) in places where I'm unable to log into my google account. Does anyone know if I request to delete my account entirely, if I signup after that will it let me log in with just my email/password?

[onyx kraken]

I do not know. But you can also contact through [e-mail]

It may take a while to get an answer, they usually seem swamped with help requests

[onyx kraken]

The help bot tree navigation it sounds like you need for reaching leaving a message is probably:

ChatGPT
ChatGPT+
Other
No

Then you have a choice of 'Provide Feedback' or "report a bug or errors".

I'm not sure which to recommend in your case there.

[cinder folio]

Yeah, I got through the chatBot to that point. But generally "feedback" and reporting bugs aren't things that they actually get back to you on, they just go into a long queue somewhere and are ignored :/

I just sent an email to their support.

@onyx kraken Thanks for the response 🙂

[fallow snow]

Wait I mean 60k 🤣

I’m tired 🤣

[lean patio]

this is a very odd response to get, no?

[bronze charm]

tbf I wouldn't be able to interpret that image either /j

[late thunder]

Hey everyone! Does anyone know who I can contact about a "deactivated" account? The "help" section on the website just tells me to clear cache and change browser if I cant login

[fallow snow]

Did you do anything against content policy?

Can anyone help me from stopping this behavior

It keeps wanting to add these summary paragraphs at the end

“Together, they looked out at the horizon, their hearts and futures intertwined, ready for whatever adventures lay ahead.”

“Her laughter rang out, blending with the sounds of the bustling pier. Together, they continued their stroll, basking in the simple joys of each other's company and the endless possibilities of their shared world.”

“Their evening continued under the starlit sky, filled with laughter, shared dreams, and the unspoken promise of countless adventures to come. The world was theirs to explore, together, one step at a time.”

And it just grows and grows until it’s a big paragraph

This is why I wish we had ability to crop

[wind aspen]

Hello everyone, does anybody know of any resources where I can apply for a prompt engineering job? And is it recommended to obtain certification prior to applying?

[vivid mortar]

please reach out to [e-mail]

[vivid mortar]

The field is a bit too new for many certifications but I am sure some exist, although none are provided by OpenAI.

I have seen prompt engineering and conversational AI / LLM job offerings mentioned on linkedin

[wind aspen]

Thank you !

[grizzled tree]

hello, is there problem with dall-e right now?

[night breach]

That's been the normal for everyone since devday. You just gotta open a new window and try again.

[keen flame]

when using the "call" option in the android app, the voice from chatgpt side sounds very low volume.. device volume is all the way up and loud on other apps.. is this a known issue?

[onyx kraken]

Mentioned before. Usual advice is to open your system volume on your device and make sure each of the volume categories is up.

[crisp brook]

I tried to log in today and even though I am a paid subscriber I can't seem to be able to do so.

[rigid musk]

Server can be overwhelmed during peak time .
You can try this

Incognito
Another fresh new browser
Or vpn

[latent sentinel]

Is there a way to get chatgpt to edit video (like get it to reduce background noise for example)

[rigid musk]

I don't think it's possible. But you ask in #ai-discussions if another tool exist

Also you don't need to post in every channel, please just ask once you will receive answer :3

[latent sentinel]

I asked chatgpt instead💀

[latent sentinel]

Well then maybe dont have 3 chanals for the same thing or things that are VERY similar

[void ice]

Hey Everyone! Do you know if it will be possible to interlink multiple customGPTs together?

[timber shadow]

Hello! I don't know if there's been any confirmation on this yet, but my speculative opinion is that it seems like GPTs will replace Plugins. And since the Plugins model allows for up to three concurrent plugins, I have my fingers crossed that GPTs add a similar functionality before Plugins are fully deprecated. 🤞

[keen heath]

When will open ai open chatgpt plus subscription again 😔

[timber shadow]

They haven't made any announcements yet, hopefully you're able to subscribe soon!

[keen flame]

volume still much lower compared to other apps..

[proven anvil]

I've been getting this error for the past week for any prompt I enter. I've tried changing browsers, loggin out, and changing model. Does anyone know why?

[rigid musk]

Did you try firefox ?

[proven anvil]

no

[rigid musk]

Cna you try ?

Opera, chrome and edge are chromium based

[proven anvil]

Sure

[proven anvil]

wow that worked, why does chrome and edge not work

[rigid musk]

Might be addon, software, vpn or other.

[shell mantle]

I had the premium sub and all of a sudden I don't have anymore and when I click on get premium it says that it's not possible to get new subs because of capacity... Any idea on what's going on? I was already on Premium and automatic renewal!

[charred parcel]

your renewal must have failed, now you will need to join the waitlist for access

[timber shadow]

To add on to Robert, can you maybe verify by checking your payment history and seeing when your last renewal charge occurred?

[winged ridge]

We ran into an issue while signing you in, please take a break and try again soon. yikes

[cold folio]

Can you try with firefox ?

[winged ridge]

I just cleared my browser cache and stuff, its working now.

[abstract ice]

I am in california, I joined this server because I keep getting Error Analyzing my requests and things keep timing out on the website any one else feeling this pain?

This is with chat gpt 4

[wispy ravine]

Why is there no voice chat on this server?

Does chatgpt answer any of these questions?

[abstract ice]

Especially when I ask it to search it dumps out. I am using brave browser

[wispy ravine]

Why no voice?

Hello

This doesn't show how many people are in the room or anything

No responses either

[trail cairn]

too many users

[trail cairn]

this just may be due server load being too high and yes you can see people saying the same from time to time

[abstract ice]

when I use the search functionality it happens all of the time

[trail cairn]

browsing? are you giving it a specific link or site

[abstract ice]

it is a paid service 🙂

search("pre-trained deep learning model for removing fence from images")

it gives me errors analyzing and errors browsing

[rigid musk]

You trying to analyze a paid website with chatgpt ?

[proven saddle]

this is what my 40 responses are currently being used for

[real anvil]

Is there a way to use DALLE to create graphs from data?

[proven saddle]

still have not managed to make it do anything other than move a favorite icon 10px to see if it could do it

there's actually no way, how do i make this stop?

[abstract ice]

I was stating that I am currently paying for chatgpt + and every time I ask it to search it gives an error analyzing or error browsing

[proven saddle]

and it did it again

what am i paying for

[rigid musk]

Thanks I didn't get what you wanted sorry.

Can you try with Firefox ?

[abstract ice]

what browser are you using?

[proven saddle]

chrome

[abstract ice]

does brave/chorme not work?

[rigid musk]

Brave and chrome are chromium based. So I wanted to compare with non chromium products

[abstract ice]

ok dok

should I try safari as well?

just download a fresh firefox install and giving it a whirl

[rigid musk]

Let me know

[timber shadow]

@gritty anvil Heyo! So what custom GPT questions do you have?

[gritty anvil]

Hey..so I'm trying to develop one for a colleague & possibly one for each kid of mine to assist them in everyday life and learning without giving them answers.

Also Taylor one specifically for Federal job searching

[timber shadow]

Gotcha, nice! Thanks for sharing. One thing to keep in mind: currently custom GPTs can only be used by those who also have ChatGPT Plus subscriptions, so keep that in mind when you are making one for a specific person in your life--which is an awesome idea, by the way! Or, in the case of your kids, for example, they'd probably just need to use your account with your supervision.

With that said, maybe we can start with just one of the simpler ones to kind of get the ideas across, then you'll be all set to make more from there. Maybe we could start with one of the ones for your kids? Does that work for you?

[abstract ice]

seems like it worked on firefox... it took awhile though.. @proven saddle did you try firefox?

oh wait I take it back I just got an error on my next question in FF:

[proven saddle]

seems like you had a winning streak though 🥳

[winged nacelle]

Why i keep obtaining cannot saving draft when i upload file to the knowledge base of a custom gpt?

it work until some point....

This is the comlpete file list and dimension

[proven saddle]

i managed to get it to make a website after about 800 messages. you can favorite an article and confetti appears when you hit the favorite button

[timber shadow]

There is a limit of 10 knowledge files up to 512MB each: https://help.openai.com/en/articles/8555545-file-uploads-with-gpts-and-advanced-data-analysis-in-chatgpt

[winged nacelle]

THANKS! i have an alredy partially trained GPT Can i remove allthe knoeledge file and reupload it without waste the triningO or i must trash it and start from scratch?

[timber shadow]

I think you should be good to just delete the knowledge files and reupload! There's nothing "behind the scenes" for GPTs, they're just instructions + knowledge files + actions, so what you see is what you get in "Configure"!

There have been some reports, though, at least earlier on, of "bugged GPTs" where saving, testing, editing, etc. just completely broke for some reason. More than once I've seen users report that creating a new GPT, copying over the instructions/actions, and then deleting the bugged GPT fixed it, so if you continue to have issues after consolidating to 10, you might try this too. I have not read this report in about the last week, though, so might not be a bug anymore.

[winged nacelle]

Many thanks, now i've reached my usage consuption for a single sesion, but i immediately start work to consollidate the knowledge database so i can made another try asap 🙂 VERY VERY thanks, i crush on this single rock a VERY big number of time in the last 3 days, and completely miss the article 🙂

[timber shadow]

You are very welcome! Come back with any other questions you might have, lots of knowledgeable people here to help you!

[winged nacelle]

BTW another question. There is any difference between the test windows GPT builder open (cnot the one of GPT builder itself) and a normal usage window? wich is best for training specific behavious and response?

[timber shadow]

Good question, I can't say for certain, but I imagine it's the same. One advantage to saving and using the GPT from its link, though, instead of Preview, is that you'll save the chat history of your testing, which may be helpful as you compare instances of your GPT.

[rugged wing]

How do I change the model mid conversation

[rigid musk]

Once selected you can't change. You can only switch to 3.5

[rugged wing]

Yay

Time to copy and paste again

[winged nacelle]

i also have the habit to tell the bbuilder when i change the knowledge file, and usually he response with a GPT behavious adjustment, is this useful or i wasting my time and token?

[timber shadow]

My current method when making a GPT is to use "Create" for one "initial setup", then I'll only work in "Configure" from there. You can do everything manually in Configure that Create does for you, without using so many GPT-4 uses.

[winged nacelle]

ooooooohhhhhhh thats very very very intresting!

[timber shadow]

Definitely! And be warned: if you make a lot of Configure changes and then go back to Create, it will likely overwrite a lot of what you have in Configure. So I say once you go Configure, never go back! Except maybe to make a new GPT pic, but I think you can even do that in Configure.

[winged nacelle]

THIS also explain the eratic behavious of my first prototype.

In the desription i cal also select roles and behavious specific? i could build in a single text to place in the windows if the answer is yes, and this became SO faster! (or it is better to have an inizial config and then go to the create tab?)

[gritty anvil]

Yea so I want to develop one that's a personal assistant (think J.A.R.V.I.S.) which I have.

That said I want to make a specific one for say Federal Employement and answer questions & do automatic web search when information is not readily accessible

[timber shadow]

You're limited to 8000 characters in instructions, so you'll be limited by that, however, you can experiment with additional instructions in the knowledge files themselves. So I don't think there's anything that would stop you from having an instructions knowledge file, and then use your instructions field to set up some conditionals.

For example, "If user says x, refer to knowledge file 1, section 1. If user says y, refer to knowledge file 1, section 2" and so on. This is obviously a simplified example but there is a lot of flexibility with this system!

Note you can't override the hard context limits of the model itself, so at a certain point of complexity you might notice a lack of instruction-following, but I say try pushing it to the limits!

[timber shadow]

Awesome! When you say "which I have," do you mean you have already experimented with making your own personal assistant GPT?

[lethal lantern]

Anyone ever encounter this error with the chatGPT API?

Error: "You provided an invalid 'prompt'. Your prompt was provided in list-with-dicts format, but a dict you provided was not valid."

My prompt is in this format:

<statement>

A numbered list of statements about the statement (ex. 3. This statement was made by a student.) , and I'm asking chatGPT to answer if they're true about the statement, or false.

It works in the web client, so why doesn't it work in the API?

[timber shadow]

You might try in #dev-chat too!

[gritty anvil]

Yup L.E.A.H. L.A.N.D., named after my mom

L.E.A.H.

L - Learning
E - Everything
A - And
H - Helping

L.A.N.D.

L - LLM
A - As
N - Naturally
D - Designed

[fallow snow]

I wish we could get more complexity for it tbh

When it becomes a more widespread feature

At least give us more different controls for length and repetitive

And it would definitely be more secure if it’s able to be updated in the chat instead of new ones

[timber shadow]

Be careful sharing potentially personal identifying information, this server is not private! Do you have any specific questions about the custom GPT process related to this GPT?

[timber shadow]

I'm definitely excited to see what's next for GPTs!

[fallow snow]

Just my personal opinion but the chat setting kinda limits what can be done

I wish we had physical settings like sliders like in the api for chat gpt

For amount it generates and resolution

Things tend to get lost in prompts

And custom instructions lose potency when it’s pushed to the back of context

[wraith crown]

For Custom GPT's, when I am developing it, sometimes it gives me an error that I need to wait a few hours because of limits. Is there a way around that a sa developer so I can keep working on it?

Then, do I have a way to access these custom GPT's via an API? If so, are those same limits there? (I built a customer service custom GPT that can do various things to my customer accounts)

[fallow snow]

How are you developing it

Do you ask the ai to make it?

Or is it from too much testing

[wraith crown]

I used this option

[fallow snow]

You know that the gpt overwrites previous instructions right

If you get it to automatically make it, best to just manually do it or edit what it generates

[wraith crown]

I'm not sure what you are asking

[fallow snow]

For testing it has 50 like usual

[wraith crown]

I have it 100% working, my question is only about the limits of usage

[fallow snow]

Yeah it’s limited to 50

Maybe when it comes out of beta

[winged nacelle]

50 what, what's the measure unit

[fallow snow]

Usage

[wraith crown]

50 conversations? 50 messages? 50 generations of text?

[fallow snow]

usage….

however they count it now

[wraith crown]

I believe it is 50 messages every 3 hours

[timber shadow]

I believe the GPT-4 cap is still 40/3hrs. But there seems to be a separate cap for custom GPTs, not yet confirmed but many users have reported it's closer to 25/3hrs. Again, that lower number seems to be specifically for custom GPTs.