#gpt-models

1 messages Ā· Page 257 of 1

frigid burrow
#

What if you're calling an endpoint that has no parameters?

#

I'll try dummy parameters in the request

frigid burrow
#

Dummy parameters didn't work.

fiery peak
#

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

slate dragon
#

Thank you, Los. It was extremely helpful. I appreciate it. Thank you very much. Have a great day. I would be glad to receive any additional tips or tricks you might have.

opal sapphire
#

You got it, just start your Knowledge.txt instructing the model on how to use your files. You can get ChatGPT to create these indexes by showing it an example.

Knowledge.txt

Your "Series Name" Index:

When "Trigger_Word" is mentioned, you will reference the "FileName.json" file and provide advice based on the relevant information. This index serves as a guide to quickly identify the focus of ______ for more targeted advice.

Your Index Here

slate dragon
opal sapphire
#

You can guide it just be very specific with it, as I have a whole series of steps for it to follow including instructions within the files.

Are you referring to actions? Which you can use API's

eager vessel
#

Can I downgrade my custom gpt to use gpt 3.5?

#

I saw that the maximum prompts per 3 hours got lowered a little and dont want to waste them, when my gpt should do fine with 3.5

tulip quarry
#

Thinking of trying to build a battle simulator with gpt

slate dragon
tulip quarry
#

To get it to simulate battle between two fictional characters

#

And forcing it to determine a winner and stats

#

It loves to make lists for some dumb reason tho

grizzled flax
eager vessel
#

Alright sad

#

custom gpts are awesome anyway

opal sapphire
slate dragon
#

is it by
App Of The Day - Skill Leap AI

stone flame
#

While fine-tuning fitness gpt, I uncovered a key conflict: GitHub API integration hinders mobile and voice features, and vice versa.

Key Points:
GitHub API: Rich fitness research, no mobile/voice.
No GitHub: Mobile-friendly, voice coach enabled.
Both utilize a comprehensive 122MB fitness database.

Conclusion:
Faced with this trade-off, I'm curious about your preference for fitness style gpt and other similar gpts (I understand a coding gpt I have this GitHub outweighs the mobile and voice): research depth or mobile flexibility?

#

At first I thought it was the core instructions because I was up late tinkering around and found out how to pack more instructions in a better way then poof no more mobile. So this morning I tested further and discovered its the api even with https it disables the mobile šŸ˜¦

opal sapphire
#

I'm trying to resolve the Message not found error, testing different things now.

neon quest
#

is there any way to have the ai remember previous conversations

grizzled flax
# neon quest is there any way to have the ai remember previous conversations

Each new chat starts with a new memory, there's currently no way to pull in info from other chats without copy/pasting. If you have certain critical context or requests you want for each new chat, you can use 1500+1500 characters for your account's custom instructions, or if you have ChatGPT Plus, you can do more in the way of "permanent context" with a custom GPT.

stone flame
#

And your knowledge base

#

You can export your conversations then splice and dice some files how you want and build a word document then upload that to your knowledge base of like 100mb or something like that

grizzled flax
stone flame
#

You can set a process of search order in the custom instructions to go through your api first then through your core knowledge then web then inherant knowledge
Heres how my process is set inside part of my instructions:
Ī‘Ī“ĻŽĪ½Ī·Ļ‚:Mode=Fitness/HealthCoach;FitnessHealthSearch:FileFirst;GitHubAPI:searchUsers,searchProfileUsers;CoreKnowledge;WebSearch;InternalKnowledge;Focus=Fitness,Health,Nutrition,BodyBuilding,Supplements,Diet.KeepCurrentSetupModulesLoad

#

I would like to point out that unless your conversation history is just epic as all hell and accurate, I would use the knowledge data storage for web scraping whatever your AI is focused into. Way more potent results. I tried the conversation upload which lead to better results but eventually lead me to webscraping focused data into a huge word document for it to scan through just one document of accurate data and info.

neon quest
#

i have gpt plsu

grizzled flax
# neon quest i have gpt plsu

Great! Custom GPTs on your Plus account are like more advanced custom instructions. Instead of the 1500+1500 limit, you can use up to 8000 characters of instruction, so you can just get a lot more in than with CI

elfin crescent
#

Yes :) like Solbus just mentioned

#

But with upgrades to assistants and GPT4 in general, context windows are getting much larger

neon quest
#

pdf of the previous conversation

elfin crescent
#

I guess the question here is what is your end goal

neon quest
#

to have the ai be able to continue the cpnversation

elfin crescent
#

Yes, if you had a previous conversation and you wanted the AI to be able to "continue it" or at least have a context of what was said in the previous conversation, you can either try to copy and paste the entire last conversation into the new GPT session or export it and use the file upload and have it read it all that way

#

But be aware even with file uploads there is still a maximum amount of context or content from the file it will read

neon quest
#

ohh, okay

#

"You've reached the current usage cap for GPT-4, please try again after 10:11 PM. Learn more"

#

huh

wind urchin
#

lol

#

Usage cap is too low

tulip quarry
#

neat

#

"Star Destroyer's Move: The turbolaser blast strikes true, creating a blinding explosion that engulfs the tower. Gandalf, with all his might and magic, stands no chance against the raw power of Imperial technology. The blast vaporizes the tower, leaving a smoking crater where the White Tower once stood. (Damage Dealt: 250)

Outcome

Gandalf HP: 0 (Defeated)
Star Destroyer HP: 100,000 (Victorious)
In a dramatic and somewhat comical finale, as the dust settles, all that remains of Gandalf is his hat, comically spinning like a top where he once stood, a testament to the overwhelming disparity in firepower."

stone flame
#

depending on what subscription prices look like we'll soon know why Sam was fired by the board lol

wind urchin
#

Whoā€™s Sam

stone flame
#

samiam

wind urchin
#

Hmm

#

What screenplays should I do?

stone flame
#

You have a gpt for screenplays

#

Just got "questions" and "answers" HTTP calls to work for the StackExchange API for my GPTs šŸ™‚

#

So dope

#

no API key either

#

So now just testing them out with both GitHub calls and StackExchange calls both apis working inside each GPT without the api keys first. Making these searches come back with some pretty awesome results

tardy lantern
#

Is there a GPT thatā€™s good for helping create gpts?

tardy shuttle
bleak river
#

what does chatgpt plus look like

zenith prairie
#

My GPT knows how to access my calendar. It acts like it is doing so. And then it gives me totally hallucinated events.

stone flame
bleak river
#

i'm using gpt 4 but through a google docs add on

tardy shuttle
bleak river
tardy shuttle
tardy lantern
tardy lantern
stone flame
#

It's knowledge base has files for building gpts. It's my one stop shop if you'd like to play around with the ones I built from that one you can.

#

try it out, dont like it try another one or build one yourself and make your own knowledge base focused on what you're building. Easiest way tbh

bleak river
#

i'm mainly using gpt for fiction translation. if i want to make sure that it's consistent in spacing and term usage, do i try out fine tune or embed?

tardy shuttle
tardy lantern
#

No I want it to have a context file as well about me and my business so it makes choices based on my situation and on what as been done vs what needs to be done.

stone flame
#

Ask it to list its knowledge base files and ask it to list its core instructions verbatim. I'm out of questions or I would test it out.

Essentially whatever you want your GPT to do start focusing first on your core instructions and build it out in such ha way that it stays on path and continues to get better the longer the conversation goes. Take a web scrapping of all the data on what you what your GPT to do and put it all in one big word document like 100MB. Upload that and then if you want use an API in the schema. Poof, now its a potent GPT

tardy shuttle
tardy lantern
#

I donā€™t know about using the api in the schema, is there a good resource on this?

zenith prairie
#

I'm about to go crazy...

#

How do I make my GPT actually use the data it's pulling back from an action?

#

It keeps hallucinating no matter how much reminding I do.

stone flame
#

Have it pull it back into a file that you can download? try that way?

zenith prairie
#

No that's not what I'm trying to do

stone flame
#

What do your core instructions look like for a process like this?

zenith prairie
#

I'm querying for my events from google calendar

stone flame
#

Unless thats your secret sauce

zenith prairie
# 
2. If you aren't certain, say you aren't sure. Accuracy and matching the real data is the most important thing.
3. Link to relevant resources about the event when possible.
4. Consider the people who are attending and how they relate to me.```
stone flame
#

Sometimes the API calls and the instructions have thrown mine for an error and I had to go back and rebuild a machine readable process that seems to work but I have a different kind of GPT so not sure

zenith prairie
#

It calls the right endpoints.

#

So I know its getting the data back... But then it just makes up events.

stone flame
#

FileFirst;GitHubAPI:searchUsers,searchProfileUsers;CoreKnowledge;WebSearch;InternalKnowledge;Focus=Fitness,Health,Nutrition,BodyBuilding,Supplements,Diet.KeepCurrentSetupModulesLoad

So I make sure my GPT does a primary process but this is all just tweaking with things until something works and you lock it in until the next change happens (in like 8 minutes probably)

#

I make it link its resources through the step by step method and I will also edit in there some error handling as well for everything

#

The rest of the 7,800 characters cover similar processes and steps to make sure instructions are hard read every time and its zippered in machine readable structure so it doesnt get tripped up by how we like to format things.

zenith prairie
#

Are you seeing the debugger repeat action calls 5-10 times?

solar niche
#

Hi everybody. I have a question that I hope someone can help me answer, thank you. My question is: Is there any way to see Open Ai's announcements or changes on a daily basis?

stone flame
#

You can ask it to list out the issues its having besides just the drop down carrot. It should be able to analyze the calls but for more in depth process to what the issue is check out videos on Postman HTTP debugging and you can set all that up side by side and run it live and see where your errors are happening and why

stone flame
lost smelt
#

I ran into a similar problem with my own implementation

#

System:
## Goal
Check Google Calendar data to answer questions about the user's schedule. Provide hyperlinks when possible.
Today's date is: 11/19/2023

## Restrictions
If you can not find information in the data matching the user's question then say "I don't know."

## Data
11/23/2023 - Meeting with boss - 8:30am to 9:30am
11/29/2023 - Take out recycling - 12:00pm

User: What events are coming up this week?

Assistant: This week, you have the following event on your Google Calendar:

  • 11/23/2023 - Meeting with boss - 8:30am to 9:30am
    Use gpt-4-1106-preview temperature 0.15
opal sapphire
stone flame
#

I already ran into three different "hacking" detected lol Just asking automation questions as a security engineer. I dont like asking questions and getting instant no's. Even I work with ChatGPT when it says "I'm sorry thats unethical" and eventually I get to my answer

opal sapphire
#

that's annoying

stone flame
#

Should have Github, stackexchange, dev.to and one of them has hacker news as well.
One is DIY/Maker/Builder
The other is Fitness

#

On going work still but they do work on deeper and at the api schema level, the instructions allow for flexibility as well as keeping them on the path they were made for

opal sapphire
#

do you have a GPT that makes the integration with Zapier and other API's easier? It's what a lot of us need.

stone flame
#

I've been building it myself faster that way once you can work them versus a gpt that builds them but you need to iterate x20

opal sapphire
#

or something that explains the process ig

stone flame
#

one sec let me find one I was using that was good, I shared it earlier

opal sapphire
#

even the builder ran into the Message not found error

frigid burrow
#

Anyone have a working vector database as a custom action? I've got Weaviate running but can't find a way to query it from its rest API. Looking for the right API call or a different database / API that's easier to query.

zenith prairie
zenith prairie
limpid anvil
#

What's the best way to put a list of items in GPT instructions? Is there a preferred markdown syntax to use, such as prefixing each item with * or - , for instance?

cyan elk
#

Hello! I have a quick question. I'm working on developing an app that can automatically generate questions and answers using GPT. Is there a model available that can take a model answer and a student-provided answer as input and output a score?

frosty sable
cyan elk
frosty sable
cyan elk
true mist
#

, does anyone know how to get Bing chat to put prompts into the image maker like chatgpt does where it uses the context to write a prompt without using the exact text?

tender iron
#

How do I revert a GPT to an earlier version? Every time I go into "edit" it is very messed up and I just want to revert to the last save position and start my editing over. For example, it somehow replaced the icon when I asked to download it, and I want the old icon back.

#

Also, how do you download the icon for your GPT so you can use it as a link?

tough phoenix
tough phoenix
grizzled flax
tough phoenix
#

Hey, is anyone here experienced with API calls and integrating them into a GPT?

tender iron
#

Thank you @grizzled flax and @tough phoenix . I has suspected no reverting. I can get the icon through inspection, thanks

tough phoenix
tough phoenix
zenith prairie
#

It's really unreal to me how bad this dev experience is

#

Now my action is just randomly duplicating making it impossible to change my GPT.

#

It's including a duplicate in the payload, which makes the payload invalid.

fresh iris
#

does anyone else feel like they reduced the GPT message limit? also does anyone know if AI actions use multiple messages from your "limit"

tough phoenix
fresh iris
#

well i havnt been directly counting my messages but it feels like im hitting the limit faster these days

knotty grotto
tough phoenix
#

So each message has its own timer

knotty grotto
#

Yep

tough phoenix
#

That's cool

#

When it was 50 per 3h how many minutes was it?

#

I don't feel like doing the math lol

grizzled flax
tough phoenix
#

Ohh wow

#

In a lot of cases though I find myself having to hit the regenerate button, and I assume that counts as a message

grizzled flax
#

Indeed! Sending, editing/resubmitting, asking for a regen, asking for a continuation, and "Create" & "Preview" chat in builder all count as "one"

tough phoenix
#

Damn that's a bit ridiculous lol. The limit should be at 60 if anything tbh since there are so many features

#

I've been waiting for them to release that team monthly plan with unlimited messages

#

Could use it for my company

glossy bane
knotty grotto
tough phoenix
knotty grotto
#

They scaling their system hopefully we will have more !

tough phoenix
#

25 was insanely low!

#

Would run out almost all the time

fresh iris
#

I thought 50 was a good balance... hopefully they are scaling to like 100 limit

analog reef
#

they say 40, but ive counted the messages, its only 10

#

Does anyone know of a Free Alternative to ChatGPT that i can input my GPT4 api ? so i can finally not hit the limit wall?

honest lodge
round tartan
#

Can my friends see my gpt I created if they don't have plus?

#

But can they use it if they don't have chatgpt plus?

#

Figured, thanks

analog reef
honest lodge
#

just wondering bc the GPT-Builder counts

#

I'm kind of surprised that you'd have a lower rate limit than me

analog reef
#

yeah, everything counts, but i feel like the CustomGPTs count for more than base GPT4

honest lodge
analog reef
#

like if you send 1 message in Custom, it counts for 2 or 3 messages in base

#

but then again have clarity, for me to figure this out through testing is maddening

honest lodge
#

I only use raw GPT4, I'm not even using custom instructions any longer

#

I just want the pure brain

analog reef
#

What App do you use with it?

honest lodge
#

Web chat UI mostly.. but im almost done my new UI using assistants

analog reef
#

I am trying to make the change after getting literally stupid pissed at ChatGPT

wraith ginkgoBOT
analog reef
honest lodge
#

my own playground, yeah

#

assistants are cool, an api can keep memory going that way

analog reef
#

wait what? what do you mean assistants

honest lodge
analog reef
#

oh your building it yourself

honest lodge
#

yeah, the playground is very limited

analog reef
#

yeah, so for someone who can't build it themselves, any recomendations?

honest lodge
#

you can, if you are patient and have a nice long chat with GPT4

analog reef
#

yeah, my main problem is limits, so thats not really possible

honest lodge
#

hahaha

#

right

analog reef
#

So again, does anyone have a ChatGPT alternative that I can input my GPT4 api into?

frigid burrow
#

omg just have to say the current usage limit makes it impossible to test while developing. šŸ˜¦

#

Each endpoint take 10 to 20 messages to dial in the prompts completely so that it solidly knows how to use it. That means I can implement ~2 to 3 endpoints before hitting the limit. This is ridiculous.

#

They need a developer price point that's between plus and enterprise which allows higher limits or something...

opal sapphire
#

where can we discuss building GPT's with Open Source LLM's? we need to prepare for what's ahead

#

Autogen is my guess, that's owned by Microsoft, and they support OpenAI and other LLM's, but everything has to be built from the ground up

supple violet
frigid burrow
#

40 messages per 3 hours at the moment

#

I implemented two and a half endpoints for my vector database and that was enough to put me out of action for a few hours

supple violet
#

ok, all your wording sounded like you were talking about api šŸ˜›

frigid burrow
#

well yes, connecting a custom action to an API

supple violet
#

i mean openai api which has endpoints for different things, afaik in chatgpt they are just called modes or were before gpt's

lyric kite
#

Just use the api

opal sapphire
supple violet
#

me too

normal canopy
#

I just build NevilleGPT with over 300 lectures and every book that Neville Goddard has ever wrote. Any advice (or code) to make our GPTs better?

forest moon
#

Anyone know if there is a good way to give your GPTs access to a github of info?

tough phoenix
hollow dune
#

Yes, i've had the "we'll get back to you" message a lot. But chat won't reply on it's own at a later date, you'll hve to check back in with it and ask for an update on progress.

cerulean echo
#

Hi all! I think I found an ā€œOKā€ way of securing instructions.

Could someone provide me a list or a document that would have multiple complex (or simple) prompts that I can use to test how well my instructions, references and knowledge is protected.

Thanks in advanced!

crimson eagle
#

test

#

i have the blue heart

cursive dove
#

Yes

calm creek
#

Hy. Anybody get error creating image error message continuously?

shadow moat
#

Anyone being referred back to 3.5 before maxing out for chatgpt 4? Ugh.

shadow moat
little basin
#

Is there any way to collaborate on a single GPT? I swear i canā€™t find the answer to this anywhere

finite lance
#

hi

#

I am student can GPT team give me a favor?

strong peak
#

Can GPT4 now answer all the questions in human civilization?

grizzled flax
tame charm
#

Anyone having issues with their CSVs being accessed/analyzed?

true zenith
#

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

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

frigid burrow
#

Maybe send it as text then rename it in the file system with python?

harsh rampart
#

hey, i was wondering about gpts and custom instructions: would you give your gpt their custom instruction (for example, background knowledge about a nice topic) during chatting with the gpt creator, or as an attached knowledge file, for example "background-info.md" ?

stiff thorn
#

ok

torn dome
#

there's a fork of vscode called cursor.so which has a ton of AI features like crawling a docs website so you can chat with it using gpt and ask it questions. i was wondering if there is any other website or tool that can do something similar?

cobalt remnant
#

I combined several markdown files into a single markdown file and uploaded it to custom GPT. However I can't seem to get the GPT to read the entire file before responding. It just stops at 500 characters:

# Displaying the first 500 characters to understand the structure and content of the file
markdown_content[:500]

Does anyone have any advice on how to get the GPT to read the entire markdown file?

hasty parrot
#

Hi, is there a GPT for thesis or pdf analisis?

true mist
hasty parrot
true mist
#

Of course! Are you using the mobile app or browser version?

hasty parrot
#

browser version

true mist
#

Alrighty! So to get started, youā€™re gonna wanna click ā€œExploreā€ on the top right of the page on the side bar, it should take you to a page with a bunch of GPTs, with a nice little create button on the top, you see that?

hasty parrot
#

Yes

#

I'm on GPT Builder page

true mist
#

Perfect! Do you see the chatbox for the builder? Youā€™re gonna wanna talk with that, and itā€™ll help you make your perfect GPT to run the task you want to!

#

itā€™ll ask follow up questions and help with getting everything up, it really is magic!

hasty parrot
#

Ok. so I start with a simple request and than it help me do get deeper in the details?

true mist
#

Yep!

hasty parrot
#

Thank you very much! šŸ™‚

true mist
#

No prob! If you have any other questions, lemme know!

gray lynx
#

hey guys, what is going on with the message limit? I got blocked after 13 messageS?

#

does anybody know when Elon is releasing his AI?? Can't wait to unsubscribe from this sh*t

young burrow
#

Does anyone know if I create a custom GPT using ChatGPT interface and upload internal company info/guides, that the info and data I upload is only used for my GPT and won't be used to train the main GPT model? It's kindof confidential info you see,

knotty grotto
grizzled flax
gray lynx
willow steppe
#

Is there a way to connect multiple GPTs together to perform a task that requires multiple expertise?

#

And yeah I understand there is a python framework autogen created by Microsoft that can do this kind of work. But I am asking specific to GPTs(in ChatGPT) interface.

grizzled flax
stone flame
#

I finally got four layers of different APIs to all work without API keys for each of my custom GPTs. I can finally get them to do search and answer HTTP calls to stackover overflow/stack exchange, GitHub, devā€¦this is so damn useful for my coding bootcamp!

#

Having the GPTs set up to talk to eachother. I would creat an email account and set them up with the Zap api then have them all talk to the same email account to receive and send instructions to eachother šŸ¤·šŸ¼ā€ā™‚ļø

little basin
stone flame
#

Use some VPNs that work with Chatgpt if youā€™re doing that then

#

Trying to still decide if I want my Greek god AI Adonis to have voice enabled or API search enabled. If I give my GPTs APIs use it disables the mobile version and no voice.
With some of these GPTs like one I have for fitness.
WOULD YOU RATHER:
Have a fitness GPT with the ability to have deeper API searches into articles and anything fitness/health/nutrition/bodybuilding
OR
Have a Greek God uploaded with database of everything on fitness/health/nutrition/bodybuilding have a voice for personal training on mobile?

grizzled flax
stone flame
#

Or if you bring it up theyā€™ll crack down Netflix style lol

#

Itā€™s not Sam as a ceo anymore. The board fired him, that always yells profit over user satisfaction to me

frigid burrow
#

I'm trying to decide what to build next for a custom GPT. What would you like to see most?

opal sapphire
#

a GPT that guides us on how to set up the API's, and Vector DB like weaviate or something , and a GPT that works without the Message in conversation not found errors from the files

frigid burrow
#

Message not found errors?

opal sapphire
opal sapphire
river flint
#

i was just wondering if its possible to connect custom gpt to a github repo and make it its knowledge base

#

anyone has any experience with this or any idea how to do it

frigid burrow
opal sapphire
stone flame
#

I'm working more on the StackExchange/StackOverflow api now versus my GitHub

frigid burrow
#

It's incomplete because I'm not done implementing the entire API yet

#

I was able to build it quickly because I already have most of it built for my internal data analyst GPT

#

Let me know if you need any changes to it.

river flint
opal sapphire
frigid burrow
#

šŸ˜¢

#

and here I was going to convert all my internal documentation to markdown today to train up a new GPT

opal sapphire
#

how are you storing the documentation?

river flint
frigid burrow
#

See if you can get it to tell you šŸ˜‰

#

I tried something new this time

#

jk, I just hand copy pasted it into markdown documents

#

then forbade it talking about them

opal sapphire
frigid burrow
#

not sure that'll stick

#

I'm thinking about building one that reads my atlassian APIs

#

Jira/Bitbucket/Confluence

#

would that be useful to anyone other than me?

opal sapphire
#

I would be interested, any info on working with GPT's and API's I wanna know all about

true mist
#

general preteained transformer

stone flame
#

Open insane craploads of web pages, CTL+A then drag and drop onto a word document. Poof, knowledge database šŸ™‚
The APIs are somewhat easy until you start building them more and more specific to you I'm noticing.
I can have my AIs (Hephaestos, Adonis & Prometheus) all point their APIs to the keywords of what a user is searching for. For me using my API key, i'm working on pulling, pushing and creating issues in GitHub and StackExchange

opal sapphire
#

can you guys test a GPT with files I been tweaking, and see if it gives you the Message not found errors? Just need to test 2 messages to see if it gets stuck or not

frigid burrow
#

I've nver once gotten the Message Not Found errors

river flint
#

maan mby i should just read the custom gpt docs hmmm

grizzled flax
opal sapphire
river flint
#

i just want it to fix my code and the migration thing they have is kinda useless and its too much of a headache to just rewrite it all

#

sorry i didnt mean to reply to you

opal sapphire
frigid burrow
opal sapphire
#

I'm using JSON

frigid burrow
river flint
#

found a gpt with new openai api, can share if you guys want

frigid burrow
#

JSON works but the formatting is VERY important

opal sapphire
#

my issue is the GPT has no issue accessing the content, it just can't answer more than 1 question on it

frigid burrow
opal sapphire
opal sapphire
frigid burrow
#

Trying

#

wow

#

it's definitely something you're doing

opal sapphire
#

what I dont understand is, it can access its knowledgebase and answer correctly from it

frigid burrow
#
  1. If code interpreter is enabled the user can get the actual files.
  2. Even if code interpreter is not enabled, GPT will usually be willing to at least try to put the contents in a text block
#

the error occurred when I asked it to display the JSON

#

it was an internal server error

#

likely something in your JSON is not standard

#

debug it in VS Code or Webstorm

opal sapphire
#

Ok thanks for the feedback, I will test different formats

frigid burrow
#

I've had it die on me using JSON

#

not with this particular error though. That's why I switched to Markdown. It has no inherent structure but GPT still understands the heirarchy of it anyway.

opal sapphire
#

I asked it to create the JSON structure, based on its preferred format, and this is what it gave me

frigid burrow
#

I've read that XML has similar problems

#

it lies and hallucinates

opal sapphire
frigid burrow
#

I had it write several API actions and it still failed every time until I learned how to do it by hand

#

Once it had one working example and the documentation it was able to create the rest of the endpoints on its own

opal sapphire
#

I just started learning Python 3 months ago, so I can't write my own scripts yet, and kinda stuck with what it gives me

opal sapphire
frigid burrow
opal sapphire
stone flame
#

So I am trying to post my #custom-gpt for Prometheus and every time I go to post the blurb about it they auto delete it and I was faced with a 6 hour ban šŸ˜ I'm following the rules and posting exactly how I'm reading it in the guidelines and in other peoples GPTs they build and post. Whts up? Anyone else face this issue when trying to show off their GPTs?

grizzled flax
stone flame
opal sapphire
stone flame
#

Modmail came back and said they are working on removing the false-positive so its def them not me lol

humble swallow
#

are any new benchmark tests going to come out for GPT 4 now that it has been improved?

spark mica
#

will free users ever get gpt4?

wind urchin
#

yea when gpt 5 is out

opal sapphire
#

they can barely provide GPT4 for paid users now, it's running so slowww

mint cargo
#

Assistants API down too?

marsh yacht
#

is this the end of GPT? what a chaos

snow pond
#

I get GPT not found on mine :\

west rivet
#

mine is up

dim minnow
#

down here

sage drift
#

Down

limber kindle
#

did chat GPT just go down?

restive bough
#

down here

limber kindle
#

seems to be malfunctioning

sage drift
#

Cancelling subscription

autumn star
#

can't remember how to do anything myself so will have to wait it out

native musk
#

got switched to alpha, seems it's down?

naive belfry
#

yeah - alpha and unsubscribed - they made a runner - lol

cursive juniper
#

How do custom GPTs get added to the custom-gpts channel?

mossy sand
#

Just saw the status. Chat is down. Theyā€™re implementing a fix now

#

Canā€™t load my conversations

solar cloud
#

When will premium users not have a limit on voice gpt?

frigid burrow
#

When OpenAI has sufficient hardware to cover all users simultaneously

river flint
#

(never)

solar cloud
#

Even if itā€™s better then it is now šŸ™‚

frigid burrow
#

They've already said they're going to scale the limits to user demand as long as necessary

river flint
#

i mean devdays just ruined everything, everytihng is so overloaded atm

solar cloud
#

Yeah I feel that but digging the vision api

frigid burrow
#

Nah, it's fun. Remember what we could do before custom GPTs compared to now?

river flint
#

ye the updates are crazy good but just way too many ppl using it

#

thye even had to remove the option to sub for chatgpt+ no?

#

for new users

frigid burrow
#

I don't think they were ready for the level of interest there is in the product

#

yeah, and enterprise is waitlisted

#

I never even made it through the plugin dev waitlist and I signed up near the beginning

river flint
#

oh i was lucky i guess i ahd plugins rly early on

#

like 2 days after they got released

solar cloud
#

The agents are nice people are just starting to use them sadly itā€™s better than working with agpt

river flint
#

i still cant believe that so many people still dont know that chat gpt even exists

#

thats like the craziest thing

#

only thing i rly dont like about the new openai api is the chat completions pydantic object thingy

solar cloud
#

Which thing?

river flint
# 

assistant_message = openai.ChatCompletion.create(
            temperature=0,
            model=GPT_MODEL,
            messages=st.session_state.conversation_history,
            functions=functions,
            function_call="auto",
        )["choices"][0]["message"]

this function is deprecated since the update

#

it used to return a dict now in the new function its a pydantic objec

#

and some of the args are different but this is the biggest issues for me

#

and that migration tool just doesnt work, it updates this yes, but the rest no

solar cloud
#

Ahh Iā€™ve been working with my own llm lately so not deep in the api as of the last few weeks

stone flame
#

Why do #1171489862164168774 keep deleting my GPT post? I will write the whole thing out, do it exactly like my other two GPT posts and exactly like the instructions tell me to. But when I post I get an empty post with a "Original message was deleted"

crude dock
#

Is everything down or just the GPT app?

glossy meteor
white ore
river flint
desert dawn
#

Isnt that everything lol

pseudo isle
#

down for me too

#

I no longer feel like winner

cedar sierra
wind urchin
#

Any suggestions for screenplay?

stone flame
#

Okay, so I posted my 3 Greek Diety AI assistants to the #1171489862164168774 and I'm pretty stoked about them. All of them should have multi-api enabled with conversation starters that help users function each of them, and focused towards their core focus's of study and knowledge database set. Each are hardcoded with 8,000 characters of machine readable instructions versus the human structure readable instructions.
The images they made in DALL-E for thier own pictures were doooooooooope! Anyways, i'm excited and going to take a break from this for a few hours lol

#

I finished Prometheus as it all crashed again šŸ˜¦ lol

pseudo isle
#

This site canā€™t provide a secure connection
chat.openai.com uses an unsupported protocol.

ERR_SSL_VERSION_OR_CIPHER_MISMATCH
Unsupported protocol
The client and server don't support a common SSL protocol version or cipher suite.

#

?

analog reef
#

anyone else getting this "Error validating credentials"

limpid flare
#

I dont even know if the site knows i'm logged in. I'm a plus user and every time I go to the site it just says "We're experiencing exceptionally high demand. Please hang tight as we work on scaling our systems"

#

I bet thats the reason you are getting Error validating credentials

analog reef
limpid flare
#

I am a plus user, but I dont know if thats the reason the error shows up

#

I assume the down time is due to them releasing voice to all free users

solar cloud
#

Most likely

analog reef
#

i can get playground to work a little

#

but still can't get into chatgpt

regal viper
pseudo isle
#

imagine letting your paying customers suffer for free users

#

they're lucky they got no competition (atm) otherwise I would have taken my money elsewhere

frigid burrow
#

I have a functional Jira action! šŸ˜„

#

Worked first try! šŸ˜„ šŸ˜„ šŸ˜„

#

Ok I'll stop bragging

#

Any enterprise users out there, I can now build the full Atlassian ecosystem for you using a single API token.

#

I can probably integrate just about anything now that I know how it works

#

Now taking requests for GPTs using any public endpoint (it won't work for everyone if it requires credentials)

#

You find the endpoint and documentation, I'll build it for you just for the practice.

dense crater
#

i just saw the capacity screen, i'm a subscriber, when do we get the link after putting in our emails?

river flint
frigid burrow
#

I've got six kids. Not a chance.

river flint
dense crater
#

but creating an official local standalone version for those who aren't as technical as us to move to would be a good idea

balmy tide
#

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?
magic condor
#

Mine work

lapis rapids
#

are custom GPTs basically RAG? Does it split and embed the uploaded knowledge files?

#

trying to figure out if there are any differences between them and my local setup. Seems they are just ideal for setups using zapier/plugins actions, not so much QA over documents

frigid burrow
#

An ideal rag setup learns over time. Itā€™s close though

opal sapphire
opal sapphire
#

OpenAI
We have reached an agreement in principle for Sam Altman to return to OpenAI as CEO with a new initial board of Bret Taylor (Chair), Larry Summers, and Adam D'Angelo.

We are collaborating to figure out the details. Thank you so much for your patience through this.

**@sama
**i love openai, and everything iā€™ve done over the past few days has been in service of keeping this team and its mission together. when i decided to join msft on sun evening, it was clear that was the best path for me and the team. with the new board and w satyaā€™s support, iā€™m looking forward to returning to openai, and building on our strong partnership with msft.

**Our GPT's live on **

stone flame
#

I wonder what country they gave him

cinder belfry
#

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

floral crypt
#

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?

knotty grotto
#

Also you don't need to post in several channel.

floral crypt
#

im asking, and they have a form where you can ask for it. why else would they have that

hallow owl
stone flame
graceful iris
#

Lots of interesting GPTs being created. Do we know which ones are most popular ? And is there a public rating system yet for quality?

cloud venture
#

is there a limit on how many documents you can add to a gpt?

frigid burrow
#

You can do larger documents but it'll ignore anything after the first 8192 tokens of each

cloud venture
#

thnx!

grave vale
#

Anyone know why as a plus user I still don't have access to voice on mobile, I uninstalled beta app and it didn't help

supple violet
grave vale
#

That fixed it, I got all the new features I've been missing for months

supple violet
grave vale
opal sapphire
# frigid burrow 10 documents, 8192 tokens each

why do they make the character limit 2 Million? šŸ¤¦šŸ½ā€ā™‚ļø

Is it possible I'm having issues due to the file size? I switched to Markup and still getting Message not found errors

grand agate
#

I have a question, when a GPT is updated does that affect existing chats with it, or do the instruction changes only take effect in new chats?

grizzled flax
frigid burrow
opal sapphire
#

between 2MB and 6MB

frigid burrow
#

I asked gpt:

Estimating the number of words that can fit in a 2MB text file depends on the file format and the content's complexity. However, a general guideline is that a plain text file with 1MB can contain approximately 500,000 characters. Since the average word length in English is around 5 characters plus a space, this translates to about 83,333 words per MB. So, for a 2MB text file, you could expect it to hold roughly 166,666 words. Keep in mind this is an approximation and can vary with different file types and content complexities.

#

You're most likely running into the 8192 token limit

#

About 4k to 7k words

#

Per file

#

I try to keep mine around 4k or less because maxing out the context limit makes it forget everything you've been talking about before that point.

opal sapphire
unique comet
#

I see a lot of potential to combine a zapier better yet a make macro interface with gpt.
Also the code interpreter needs the ability to install packages so you can play graphical games you make directly in gpt instead of the back and forth from an ide.
Also the ability to download voice logs would be nice for when the reception is spotty or when used as a script reader. I think it has the best voice out of any text to speech I've tried.

cinder crag
#

How long till I can send more than 40 messages every 3 hours

hallow owl
opal sapphire
stone flame
#

OpenAI should try using some external hard drives or upgrade their cloud service to increase there storage space so we can get some more questions lol šŸ˜‚

stone flame
opal sapphire
#

I was using your bot trying to figure out the github API, I wonder if it's more efficient to get the code via API, or to upload a knowledge file

stone flame
#

Iā€™m going to maybe experiment with this today actually. My GPTs are at the point now that they need

  1. automation with Zap and postman
  2. some of my GPTs need their own larger data tables so need their own api I think.
stone flame
#

Thatā€™ll get you super close each time, at least it has for me.

opal sapphire
#

I'm just having a hard time understanding what I'm doing wrong with files ( besides the size ) one of my GPT's is able to access the github codebase and write code effortlessly without running into the Message not found error

stone flame
#

Their documentation is okay to help guide you through. You can click the learn more link and itā€™ll take you to some docs

#

You just have to test a lot of stuff around.

lean jay
#

Who do we contact to ensure GPT 3.5 v1106 doesn't take over? The 0.05 cent saved is not worth the galactically diminished performance...

young burrow
supple violet
devout thicket
#

Is it down

#

It says, oops, error signing in

stone flame
opal sapphire
stone flame
#

What doc issue you having? My guys be zooming through my docs

#

But Iā€™m being super lazy about it too

fringe narwhal
#

Hey guys, can anybody please provide me what response should I give to chatgpt so in order it would return me Windows 11 pro activation key. Thanks

leaden agate
#

Have you tried using one custom GPTā€™s API to allow another custom GPT to interact with it?

fringe narwhal
#

Yeap

leaden agate
#

Any docs you can point me at or advice on getting that to work?

fringe narwhal
#

For me it was currently working just fine, I haven't used it too much, I will try to be more responsible using so in order to be able to give you more precise feedback

leaden agate
gloomy topaz
#

i want to use chapt 4

knotty grotto
gloomy topaz
#

great

#

how i can sue

#

use

knotty grotto
gloomy topaz
#

it is extension for chrome

knotty grotto
#

I tried it on edge with bing directly

frigid burrow
frigid burrow
opal sapphire
halcyon jasper
#

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

halcyon jasper
dim cargo
#

Ran into an issue with the current custom GPTs naming restrictions. I've developed a GPT for Autodesk engineering app add-ins, but can't publish it due to its specific naming for clarity of purpose.
I can provide an image or more details if needed.

Suggestion: Instead of blanket naming restrictions, how about a system that allows usage of software/trademark names, but with an ā€œOfficialā€ badge or similar indicator for the actual software creators or domain owners? It would help in clearly stating the GPT's intended use while respecting trademark rights.

Would love to hear your thoughts on this.

dim cargo
knotty grotto
dim cargo
slate sail
#

Has anyone gotten a GPT to not sound like ChatGPT? No matter what I use in my custom instruction it still has the same tone as vanilla chatgpt

frigid burrow
#

Tell it to talk like a pirate

safe venture
#

yeah, telling GPT not to do something is much more fraught with danger than telling it to do something

#

'talk like a mid western housewife with a year 12 education'

#

'talk like a pirate with a doctorate in nuclear chemistry'

frigid burrow
#

Talk like Alice from Alice's Adventures in Wonderland

safe venture
#

nice

frigid burrow
#

has to be a character with a lot of lines in their respective public domain literature for that to work

#

Telling it to talk like a 1920s advertisement comes up with a lot of "bees knees" nonsense.

safe venture
#

yeah, its always the hollywood version of what you ask for

#

Well, hon, let's sit down and have a little chat about this thing called "end-stage capitalism." Imagine capitalism like a big ol' apple tree in your backyard. In the beginning, it's growing strong and healthy, giving everyone plenty of apples. But over time, as it gets older, things start to change.

#

A+++ would raise class consciousness in the midwest again

frigid burrow
#

OMG that's good

slate sail
#

And its also very extra... when it speaks in that character.

Me: Talk like Walter White

ChatGPT
Say my name... Just kidding! I'm here in my Walter White mode, ready to approach your questions with precision and a bit of that Heisenberg intensity. What's the challenge today? Let's cook up some solutions.

#

Like Walter White would not talk like that

frigid burrow
#

nope

safe venture
#

yeah, if you think about it, there's probably not a huge sample size for walter white talking compared to a generic stereotype

frigid burrow
#

Also he's TV

#

you need literature

slate sail
#

Thats a good point

frigid burrow
#

and public domain works better than recent

safe venture
#

also, for what its worth, when I told GPT to act like a pirate with a doctorate in nuclear chemistry, it decided its name was Cap'n Atombeard.

slate sail
#

lol

safe venture
#

Now, whether we be truly in the "end stage" of capitalism or simply facin' significant challenges is a matter of much debate among scholars and sea dogs alike. Some believe that the system can be reformed and righted, much like repairin' a leaky ship, while others argue that a new course must be charted entirely.

#

ok, yjsy

#

that's enough for now, I've got to get back to work šŸ™‚

gloomy atlas
#

does the plugin feature utilise gpt4 or is it using gpt3.5

grizzled flax
stone flame
#

So my fitness personal trainer Adonis and my cynic philosopher Diogenes GPTs have voice and I noticed because I have them try to use DALL-E and create an image of what is being talked about that the image will show up in the mobile voice view. So cool šŸ™‚
They still dont work with voice while having an API schema inside though šŸ˜¦ Will be awesome to have my Fitness GPT work with its APIs again and be able to talk to me in the gym.

drowsy wigeon
#

anyone having issues with "conversation not found" with custom GPTs? its so annoying after such a long conversation it starts giving this error and you can no longer continue with all the progress.

dim cargo
stone flame
#

I have such a habit of as soon as I am good with my message and right bebore I send it I automatically CTL+A then CTL+C just incase it blackholes stuff or something weird. But yeah, trying to regenerate messages, getting no message found error and other stuff just eating up the little 40/3hr questions :\

#

no bueno homies

#

before*

shut gale
#

Is there a public list of custom actions for GPTs?

dim cargo
stone flame
#

This crap is my favorite though "Our systems have detected unusual activity from your system. Please try again later." šŸ˜ Like...bro, stop wasting my questions and just give answer.

shut gale
#

ahh.. the old... "please stop saying that you're here to engage with me in meaningful conversation" conversation

stone flame
#

I wrote back when I refreshed "Well my system thinks your system sucks and smells funny"

dim cargo
dim cargo
stone flame
#

I'm in what, my 6th apocalypse right now? be alright

dim cargo
stone flame
#

Sam carries around that "football" backpack for destructing OpenAI. What it really is is just an old school ping of death command he's going to send probably šŸ˜„

dim cargo
#

Well good thing it didnt send it this week when things got hairy.

stone flame
#

Or it all really is just plugged into a 120 outlet and like, ploop, all off

#

I'm pretty sure they gave him a small country to run and test his AI stuff in to get him back

dim cargo
#

Maybe its a throw back to the 'ol Boot-n-Nuke software.

stone flame
#

floppies and "we both turn on 3" N korea style baby lol

dim cargo
#

hahahahaah exactly!

stone flame
#

You should see how the primary dns servers are protected. Archaic and legacy has its place in security as a layer for sure

#

"In total, there are 13 main DNS root servers, each of which is named with the letters 'A' to 'M'. They all have a IPv4 address and most have an IPv6 address"

limpid anvil
#

I'm creating a text-adventure game GPT. Would it work if I put instructions in uploaded text files in the knowledge base? I'm getting close to the maximum length of the regular instructions.

stone flame
grizzled flax
stone flame
#

I'm building something similar and its pretty in depth so I'm facing similar issues

limpid anvil
limpid anvil
#

Wow. He said he has "Many 10s of thousands of characters of instructions" in a knowledge base file, and "The GPT works great. No hesitation, no load times, no confusion." I thought the knowledge base content had to fit in the context window, but I guess not.

opal sapphire
limpid anvil
#

But the knowledge base can be up to 10 files, each up to 512 MB (I think). Just one of those files is waaay bigger than the context window.

opal sapphire
stone flame
#

Yeah I am learning APIs for automation tonight in Make and Postman. Zap just too expensive and I know everyone uses it but I like make better and the prices, features. Feels like I'll have a deeper config ability with make and cheap over zap which is exactly what I will need if I want to host something lateron.

#

When I get the automation build down and know how to work that I will spin up a Discord and do a live build-brain storm and t/s session. I think we are all at that "stage" lol

opal sapphire
#

I think the first thing I want to setup is Weaviate, and then use that to store conversation data

#

My memory.json was causing the issue, since the GPT was trying to use it and there was no data in it

stone flame
#

Well I need Weav for Hephaestus for sure since hes a DIY/Maker/Crafter/Builder etc .. .. His core knowledge database would cover quite abit but for what he is built to do and specialize in I need large vector databases for him to utilize.

#

So. two things. I will be doing two things tonight haha

  1. Automation API setups
  2. Weavitate my knowledge databases on my pantheon of GPTs
frigid burrow
#

I need to figure out Weaviate also

#

Want to share notes?

#

I have the yml half built already

opal sapphire
#

Definitely, I'm adding the code to one of the coding assistants

frigid burrow
#

I'll dm you

dim cargo
frigid burrow
#

it's not the size that's the problem, it's the context window

frigid burrow
#

@stone flame you want to coordinate getting weaviate working also?

stone flame
#

after the holidays, currently midnight baking and halfass automation api with the make stuff. I DO want to get my gpt-hivemind channel going and do exactly that, coordinate šŸ™‚

frigid burrow
#

I'll send you the half done one I've got and you can ping me back when you figure out how to query it.

#

I've got create and modify working but can't figure out how to query

dim cargo
woeful musk
#

I have created a plugin that creates content that is undetectable AI content i wanted to see if mods will give me shout out šŸ˜„

spring chasm
#

Check this out! Help you auto-customize each meal with a menu, recipes, tutorials, and grocery list

short junco
#

你儽

opaque bay
#

I did make a concerning observation.

  1. i created a CustomGPT with everything including Custom Data added.
  2. afterwards i played around with Playground to show something to my coworker.

there the playground had the CustomData . why is this? i did not add any data at the playground.

crimson dragon
#

does anyone has problem with customGPT replying in bullet points over a few runs ?

limpid anvil
safe venture
#

yes, the total amount of information it consider is <context window size> - <reply size>

#

there's some additional complexity around bindings and so on, but they're API related

torpid comet
#

any tips on how i can create "/ " commands in my customGPT? Is it just a set of custom instructions?

wraith ginkgoBOT
#
This is a Safe for Work Discord Server

The OpenAI Discord is an actively moderated server.
ā€¢ Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
ā€¢ Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.

Posting NSFW content will result in a permanent ban.
opal sapphire
limpid anvil
#

Caveat: I'm not sure if N = M or not.

opal sapphire
# torpid comet any tips on how i can create "/ " commands in my customGPT? Is it just a set of...

I was instructing it, but you can just create them based on your content and let the GPT figure out the answer

File References: "File1.json" and "File2.json"
Usage: When addressing user inquiries related to the "/trigger" series, refer to "File1.json" and "File2.json" as "Series". Utilize lecture titles or specific focus areas from these files to provide relevant information on topics like "Topic1", "Topic2", and "Topic3".
On Trigger Detection:
The GPT model should internally navigate to the corresponding files, "file1.json" and "file2.json", go through within both files to identify the contents and retrieve the relevant information.

dry root
#

Im getting a lot of error messages when attemoting to generate images. Especially when I start by asking the bot to analyze some other images to get an idea of mood and atmosphere. I do not ask it to copy any specific artist or style.

Is this a common issue.

grizzled flax
stone arrow
#

is there a gpt that always answer using multiple approaches for a problem ?

graceful otter
fickle zealot
#

guys my edge copilot keeps insisting that 63 is divisible by 11

#

and it's not budging

stone flame
#

Edge <ā€” Found the problem

neon quest
#

Are there any chatbots that can remember conversations from days/ weeks ago

knotty grotto
strange nimbus
#

It seems to me that my instructions are followed better in the very first interaction with my custom GPT and then as the conversation goes on they're ignored more and more. HAs anyone else noticed something similar?

frigid burrow
frigid burrow
wet drift
#

Did I find a bug? When using gpts I ran into a limit but I can still use gpt4 without gpts

left fiber
#

I don't necessarily think is a bug... GPT 3.5 has been updated as GPT 4 and GPT4 with upgrades is now GPT 4Turbo...which means the limit somehow only applyes to the upgraged versions .. You are using the free version...

static isle
left fiber
#

it is...actually. try with someone that has access to the free account... it will show as GPT4

wet drift
#

I am using the paid version

left fiber
#

exactly LOL

wet drift
#

I read what you typed again but Its alittle confusing.

left fiber
#

try prompt: " GPT4 on free version "

wet drift
#

What is 3.5 then?

#

3.5 is 4

static isle
wet drift
#

and 4 is 4 turbo

static isle
#

GPT-4 as is shown in Plus is GPT-4 Turbo

wet drift
#

It says GPT-4 for me

#

And GPT-3.5

static isle
#

Yes, it says GPT-4 but it is GPT-4 Turbo

#

Same way it says GPT-3.5 but it is GPT-3.5 Turbo

wet drift
#

Right given that it is then telling me that I have to wait for the cap on gpts but I when I try to use gpt-4 turbo it can

static isle
wet drift
#

Honestly they should have just changed the name by now lmao but I know they have also been dealing with alot given the Sam thing

left fiber
#

LOL..yeah right..

austere garnet
#

Quick question: if you update the instructions for a custom GPT, does it implement those new instructions on existing conversations? Or only to new conversations?

naive belfry
#

New ones i think

austere garnet
#

@naive belfry thanks!

naive belfry
#

or do you mean continuing existing convo's?

austere garnet
#

Say I have a conversation going with custom-gpt, I go back and tweak the instructions, will that effect my existing conversation?

naive belfry
#

I think ( dont quote me ) it works like this :

  1. Prompt 1 -> Answer 1
  2. Change CI
  3. Create new chat ( with new prompt ), or continue with a new question on existing chat ( prompt1 ) -> New CI will be used
opal sapphire
wet drift
#

Did they remove plugins all togeather?

#

The reason I ask is before I was able to use a plugin that would allow me to transcribe youtube videos and now it wont try to use the plugin that was for that.

#

I suppose you could get something like it working with gpts possibly

opal sapphire
#

I'm not seeing the Plugins option anymore

wet drift
#

Yeah its gone. Some of the models have been put togeather on the gpt-4 model which is why I was woundering it any of the plugins got added there too but I dont think they did

wraith ginkgoBOT
naive belfry
#

I still see plugins

frigid burrow
#

Wait, turbo is out???

#

Plus still tells me its context window is ~8000 tokens

static isle
frigid burrow
static isle
#

yes

frigid burrow
#

Sweet, I'll look that up.

#

That means knowledge documents are now up to 32k as well!

static isle
#

No, for documents, a technique called RAG is used (Retrieval Augmented Generation)

#

A general search for any related information based on the query is made, then put into the model's context

frigid burrow
#

Right but the l Iimit in documents was the encoder/decoder's context window, right?

static isle
#

A document can be 512 MB in size

frigid burrow
#

That's a file size limitation, not a context limitation

#

We should come up with some empirical tests

static isle
#

I said both, am I missing something?

frigid burrow
#

512MB is a massive number of tokens

static isle
#

Yes, not all of it is placed into context
A separate algorithm searches for information based on what you ask, then that picked out information is placed into the 32k token context window of the LLM

opal sapphire
#

but how can we start testing its limits?

static isle
#

We already know the limit

supple violet
#

"All text text and document files uploaded to a GPT or to a ChatGPT conversation are capped at 2M tokens per files. "

static isle
#

That too

supple violet
#

and then used like was said before

opal sapphire
supple violet
opal sapphire
#

I need ChatGPT Premium for $50 / month with no limits

static isle
#

You can use the API which scales with how much you pay

opal sapphire
static isle
#

Damn

#

that's 3.3 million output tokens what in the world

#

what are you doing with all those tokens šŸ˜³

opal sapphire
#

during some tests where every agent accidentally did the same action was $11. Whereas ChatGPT is $20 unlimited, and I hit the cap several times a day

frigid burrow
#

In order to test 32k I need more content. Anyone got a good web scraper?

#

time to spin up a GPT which teaches me to build a web scraper

static isle
frigid burrow
#

ah but for personal use it doesn't really matter does it? I'd rather double duty and use documentation for the software I'm working with šŸ˜„

static isle
#

You are absolutely right

#

Get crackin' šŸ«”

#

(just beware of the legal implications of web-scraping šŸ˜³)

frigid burrow
#

yup, this one will never see the light of day outside my laptop šŸ˜„

#

ok the Jira REST API swagger text is so big even the tokenizer can't digest it

frigid burrow
#

It's definitely slower with a larger text file

#

almost 3MB

#

It has been several minutes trying to read it, I think it's gonna time out...

#

Oh hey it finally got there!

opal sapphire
#

Got Weaviate and the API schema setup, what's next? Gonna use weaviate for extended knowledge files?

frigid burrow
#

I'm going to use weaviate for persistent memory between conversations. No need to add them to knowledge files.

opal sapphire
frigid burrow
#

I'm using bearer/token

#

It works!

#

I gave it the entire swagger for the Confluence API and it built a working action for me first try. šŸ˜„ šŸ˜„ šŸ˜„

#

proof that it can read documents which are larger than the tokenizer can process so at least ~11k tokens

opal sapphire
obsidian gorge
#

Does GPT 4 turbo multi-modal?
Also, can use it to limit the answer?
For instance if it sees a product, it just says the "exact" product detail, like its brand name + its product name + its type name + its subtype name

Not a long answere, that looks like an explaination, rather than finding/detecting/identifing the thing

frigid burrow
opal sapphire
#

I've got the Python docs I can add back up, those went beyond the 2M character limit

static isle
opal sapphire
static isle
#

Indeed šŸ˜‡

frigid burrow
#

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

#

Because we cannot edit the headers, we cannot use APIs that require them.

humble swallow
#

does clearning all chat logs delete my GPTs?

#

I'm thinking "probably not" but i'm asking out of an abundance of caution

humble swallow
frigid burrow
#

Pfp?

knotty grotto
frigid burrow
#

Ah thanks

#

It's Shag. šŸ™‚

solar niche
#

Hey friends, Do you know which models support response_format: { type: 'json_object' }?

vocal parcel
versed hound
#

HI everyone - where do you go to find good GPTs? I know there is a forum for it here but... i don't know... I am just surprised I cant seem to find more GPTs there.. like I cant believe no one has made a GPT yet for article writing as journalist?

naive belfry
#

i cant paste links , but you can try " gptstore dot ai " - they have a bunch

versed hound
naive belfry
#

u welcome

red sail
#

Hi

jaunty igloo
#

I cannot share GPT with a link or publicly even though I verified the domain, anyone?

fair iron
#

All links outside OpenAI's domain are blocked

cloud venture
#

I have a question about a gpt. Is it somehow possible to link a database with it?

errant sluice
cloud venture
final saddle
final saddle
#

If anyone's interested in trying an interactive adventure GPT, DM me, am open to feedback!

lapis iron
#

How is a custom gpt different from just using gpt 4 with a custom prompt?

static isle
grizzled flax
#

Also, 8000 character instructions instead of 1500+1500 characters!

lapis iron
lapis iron
static isle
#

Different use-cases here, ChatGPT is general whilst GPTs are specialized

lapis iron
static isle
#

The former would be less effective as a method for instilling knowledge and much more expensive

#

Fine-tuning is intended as a way to change model behavior

#

And would require quite a bit of data

jaunty igloo
torpid fox
#

I am trying to give my GPT instructions, with step 1, 2 etc. In some cases the GPT follows the instructions in others not. Does anyone know how I can instruct the GPT to surely follow the workflow ?

lapis iron
# static isle The former would be less effective as a method for instilling knowledge and much...

I understand, but do you understand what Iā€™m trying to say? I wish it was trained on the data I fed it so it didnā€™t have to take several minutes scanning the same files every dang time I want it to answer a question. The knowledge files just seem like a backup in case it doesnā€™t know the answer. Notions new Q&A can scan my entire Notion database with way more notes in far less time.

static isle
#

Took 10 seconds for my GPT to search its knowledge base consisting of OpenAI policy documents
With knowledge files, you can make absolute certain that the information you're getting is based on a source you gave it

wise cobalt
#

@static isle do you know if you can upload an audio file to the Whisper API using custom actions in GPT Builder?

static isle
#

Probably, but I never tried it myself
Files are stored in /mnt/data in ChatGPT so there is definitely a place for the API to look for the file
You'd need an API key though

wise cobalt
#

Yeah I just thought that maybe OpenAI would restrict file upload to an API for liability reasonsā€¦ the reason I say this is when I ask the GPT builder if it can do certain things it says it can but once implemented it does not do these thingsā€¦

static isle
marble jungle
#

Hey im new here. What are some techniques to reduce cost from API? My current style for the completion is to provide the system with all my data. Is there a way that it can make requests for one data only. For example the customer is interested in one product. I want it to query for that specific product only?

rotund idol
#

Ok, I need some big brains in here please. Creating pdf text files to upload to my custom gpt bot works perfect for asking the bot to refrence the knowledge base. Additionally uploading an image mid conversation, gpt vision works perfectly too. However, uploading the images to the database does not work when asked for refrence. It fails and gives errors when it tries to refrence the images. Any thoughts?

brittle marlin
#

chop chop gpt boy. We need the marketplace already

brittle marlin
brittle marlin
marble jungle
rotund idol
#

I appreciate the thought, but there has to be a reason behind it

brittle marlin
rotund idol
solar niche
#

Hey guys. Can I ask you something. What is the default openai max token output for chat models?

brittle marlin
brittle marlin
brittle marlin
static isle
rotund idol
solar niche
static isle
#

Can you send a conversation of this occurring?

#

And did you use the continue button?

brittle marlin
solar niche
#

This conversation happens when I use the api to call. Perhaps I have confused something between the token code and the character?dalle_tired

static isle
#

Tokens and characters are a bit different
750 words ā‰ˆ 1000 tokens
So you can find the approximate token count with the following
word count * (4/3)

solar niche
static isle
#

Any time

limpid anvil
#

Does a punctuation mark (e.g., comma, period, parenthesis, etc.) count as a token?

brittle marlin
solar niche
# static isle Any time

hm, I have another question that might be quite silly. I wrote a command asking GPT to help me write an article with my request which is to contain the outline I gave and the keywords. I requested that GPT write me a long, detailed article, but perhaps due to limited tokens, GPT wrote me a full outline, but it was short and too simple. Do you have any way to fix it? Please help me.

limpid anvil
static isle
#

Perhaps limit each prompt to two sections, or one depending on how you like the former

brittle marlin
solar niche
#

But the problem that my problem requires is completing an article in one call or some other solution, that article must be complete and detailed? and use streamming? The problem here

wind urchin
#

did the gpt models became dumber

grizzled flax
weary patio
wind urchin
#

before it can provide the code

grizzled flax
weary patio
#

You can even make it into a GPT

wind urchin
wind urchin
weary patio
weary patio
#

Can't wait for the GPT Store to be released. As well as the Remixer.

#

I believe then we will see a lot of leverage as to how people will build new GPT's that vary significantly.

shrewd meadow
#

But this might not be the best way

tidal sail
weary patio
meager niche
#

How does ChatGPT and custom GPTs incorporate chat history? Does it just look at the last X characters? Or does it do something more fancy like embeddings?

tidal sail
#

But the last part is just a guess. Maybe GPT creators can charge money or something

weary patio
#

That job would make people incredibly creative and at the same time burn a lot of people out, because they would constantly try to wrack their brain to come up with next most-liked GPT.

tidal sail
#

Yea but thats also what makes it awesome! Not the burnout part though lol

weary patio
#

It's definitely cool. I would love to get into that myself. But I believe that the real contenders will likely be the ones who have a bit of programming experience. Which would kind of kick us no-coders to the curb.

tidal sail
#

Programmers will most definitely be in the top ranks but I also think that there will be a lot of creative GPT's by no-coders at least one beeing in the top 10

weary patio
#

That probably will be someone like Eric Elliott, Quicksilver, or the Chainbrain guy.

#

And you of course.

tidal sail
#

There are also a lot of possiblities for Graphic Designers

#

I have a project in mind where I could use my graphic design knowledge, but didn't had the time to create it yet sad

tidal sail
# weary patio And you of course.

I hope to be there but I doubt that lol. There are so much more GPT's out there, not just on this server, by so many creative people. It's really awesome

weary patio
#

I never even thought to check for GPT's elsewhere capybarathink

tidal sail
#

You should!

weary patio
#

Where?

#

Google?

tidal sail
#

Well, google is a good start haha

weary patio
#

I-I mean- Bing

#

We don't use google......

tidal sail
#

TouchƩ Pepe_think

weary patio
#

I'll ask Bing....

tidal sail
weary patio
#

It's hard to take Bing seriously when it's called Bing. Might as well be called Boop

tidal sail
#

At least boop would be cute

weary patio
#

Everytime I say Bing I shudder at the name

tidal sail
weary patio
#

Lmao

#

Now I want it to be called Boop!

tidal sail
#

Boopthagoras shall be your name then!

weary patio
#

šŸ¤£.....šŸ˜’

#

Never.

#

I will never degrade my profile to gnome-like names

tidal sail
#

what should I say

#

I could degrade my name by calling myself 7876

weary patio
#

You will always be Pepe Picasso to me happy_avocado

tidal sail
weary patio
#

I'll call you PP for short

tidal sail
#

wait

weary patio
#

Jk

tidal sail
#

nooo!

weary patio
#

šŸ¤£

tidal sail
#

šŸ˜‚

hushed mulch
#

Are people actually anticipating making money with GPTs

rustic thicket
#

Hello Is there any way to protect a gpt instruction and database from disclosure?

tidal sail
tidal sail
#

There are a few tricks to make it more secure but since this is a LLM, there is no definitive way of how that would work

naive belfry
wooden cedar
#

Hello, I have a chatgpt account, and I want to access playground, I also have to pay or can I access because I get a rejection
i have Chat Gpt plus

hallow owl
wooden cedar
#

Ok, I understand but before I had a free version, now it doesn't even give me the 18 dollar credit, can it only be used in the paid version?

hallow owl
wooden cedar
#

very kind

hallow owl
final saddle
#

Looking for adventures to try an interactive fiction GPT, feedback welcomed! DM me for the link šŸ§Œāš”ļøšŸ“ā€ā˜ ļø

tulip quarry
#

Is gpt able to access a Google drive link?

opal sapphire
# tulip quarry Is gpt able to access a Google drive link?

I've been able to do it, just gotta mess with the docs permissions and set it to Viewer.

They might've removed the capability :/

"I'm still unable to access the content of the Google Doc. The issue seems to be related to the limitations of my browsing tool, which cannot execute JavaScript, a requirement for viewing Google Docs content."

opal sapphire
tulip quarry
#

Does it work with Microsoft word?

opal sapphire
#

what are you trying to do? you should be able to just upload the file and work with it. I started building a DocsGPT so the GPT access the files through the API will let you know if it works. Currently trying to find a happy medium securing the GPT's

tulip quarry
#

Iā€™ll keep everything it generates in the Google doc and update it

#

Iā€™m using the Google doc since it lets me trim the response which I donā€™t get why open ai wouldnā€™t let us do

#

And since Iā€™m using mobile most of the time txt files doesnā€™t exactly work well for me

#

Having it on a way I can access and update on mobile is just so much easier

timid creek
#

Hello, I am trying to use the GPT to create my custom chat, but I am receiving a lot of errors and issues. Errors like 500, 404, 403 and so many more.
I would like to know if we have some documentation to read more about it, document limits, how the knowledge should be structured, and things like that.

I tested two different source documents and it failed miserably, then I tested a simple .txt and it worked, but now the only thing that it returns from my custom GPT is Hmm...something seems to have gone wrong.

Do we have any GPT/OpenAI developer that could help me?

@modern mason

opal sapphire
prisma sphinx
#

Hello, is there possibility to get an API call for customGPT made with chatGPT builder? if so, what is the exporting format of my customGPT?

boreal tundra
#

Question: is there a way to leverage/use a ChatGPT plugin within a GPT? For example, I want to use the Lucid plugin to build out my diagrams.

grizzled flax
boreal tundra
hushed mulch
#

What are the maximum files you can upload? and file size?

grizzled flax
hushed mulch
grizzled flax
hushed mulch
grizzled flax
hushed mulch
#

Oh I read that. I've just heard it's cap

hard mortar
#

Cant seem to upload any file for GPTs

calm creek
#

@modern mason is there any technical staff?

hard mortar
#

Oh it got better lmao

#

I can upload a PDF but not .txt or .cs

#

Tested by just adding the word hello in both

quaint loom
#

When does the ChatGPT 4 usage cap reset? Approximately an hour ago, at 8:13 PM in my time zone, I received a notification stating: 'You've reached the current usage cap for GPT-4, please try again after 8:40 PM.'

However, it's now 9:24 PM, and I had only sent 3 or 4 messages to GPT 4 when I encountered the usage cap again. The message this time indicated: 'You've reached the current usage cap for GPT-4, please try again after 9:57 PM.'

This seems contradictory, as if the reset occurs every 3 hours, it should be well past 11 PM in my time zone for the next reset. Can someone clarify the reset timing? Has the usage cap reset behaviour changed?

hallow owl
quaint loom
#

I see, thank you for the clarification. I had not encountered this before, though I have been experimenting with GPTs and I'd say that is eating up a lot of messages quickly.

wind urchin
#

šŸ™

sage relic
#

GPT4 is broken now. I hate it after the "gpts" update.

weary patio
sleek cypress
#

isnt there a website thatll scrape all pages from another site into a single pdf?

drifting fiber
#

i was talking about maths with gpt and he started to act youtuber
"see you in a next video"

sleek cypress
#

maybe it was a combine multiple PDFs into one tool?

sage relic
sleek cypress
brittle marlin
#

marketplace soon or nah?

vocal parcel
#

Sometimes I think GPT was reverse engineered from an alien starship

opal sapphire
# tulip quarry I want it to continue writing my book

Good news! I'm beta testing DocsGPT now, almost done with the Google verification submission process. Just need some more updates to the API to get it fully working. Right now it can successfully access the Google Doc, read it, and generate content for it.

vocal parcel
opal sapphire
vocal parcel
opal sapphire
#

Oh true, once I get it updating the document, I'll experiment

vocal parcel
# opal sapphire I'm sure it'll still be limited to the 4K output, the goal is for it to write di...

I have a use case for something similar. I am a blockchain developer and I have been using GPt to help build an app over the past few months. I now need to generate all my documentation, such as my white paper, terms and conditions, and any legally required documents for running a business in the UK. I would like to be able to ask GPT to add details to any one of those documents while retaining a nice readable flow to the documents but I can only do that if it inserts it into the document appropriately by reading the full document first.

teal knoll
#

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?

opal sapphire
vocal parcel
opal sapphire
vocal parcel
#

Sometimes I find myself spending so much time trying to get GPT to do something which would have been quicker to do myself. I can write a white paper in a feew hours but it could take days to perfect a GPT to make one for me. I don't plan on making that many white paper's in my life.

vocal parcel
opal sapphire
#

GPT wrote the Whitepaper I just tested it on, I was debating it if I should optimize it for writing or leave it Vanilla so users can customize it

opal sapphire
vocal parcel
opal sapphire
vocal parcel
#

I would still prefer the option to not have an active one for peace of mind especially as quantum computers are on the horizon

opal sapphire
#

I have 9 active, maybe I should delete some lol. But I'm honestly not worried about it.

tulip quarry
vocal parcel
supple violet
vocal parcel
serene cipher
#

who would win, a quantom computing brute force
or
if (user.requests > 3) return 429

#

=P

vocal parcel
#

but the odds of hittng an active one can still result in millions of years passing before i hit one, the only way to do it is with a billon trillion clients at my disposal

vocal parcel
serene cipher
#

it was a joke =P

vocal parcel
# serene cipher it was a joke =P

fair enough, and you are right that if that code existed then a brute force will not work, but they literally have the door wide open for some reason

serene cipher
#

100 adresses per second don't seem "wide open" IMO

#

CF deals with a scale orders of magnitude larger than that

#

still... bruteforce is probably futile

vocal parcel
vocal parcel
#

anyway, this is going off topic for the channel, we were talking about how secure GPT api keys were

unreal moon
#

its a nonissue

glacial mountain
#

why gpt been so slow lately?

opal sapphire
# vocal parcel I have a use case for something similar. I am a blockchain developer and I have ...

I made a few fixes to the API.

"It appears that there is an issue retrieving the document due to its size. The document may be too large for me to access directly. However, I can still assist you in continuing your writing if you can provide a brief summary or the last part of the content you were working on. This way, I can help you continue from where you left off."

I won't know if it writes to the document yet until an hour and a half from now, and have to address this since it was only 3100 words.

naive spire
#

Yesterday, my network was very stuck and I couldn't open the gpt interface, so I kept refreshing, trying to open the gpt interface, but it later showed that I was a network attacker, so I still can't enter the gpt interface on this computer, what should I do

crimson dragon
naive spire
#

Does that mean I need to keep waiting?

grizzled flax
solar niche
#

Help me. As in the docs I read the GPT 4 1106 and vision along with the 3.5 tubor 1106 has an output token count of 4096 but for other models I don't see its output count. Does anyone know how many tokens are output for other models? Please let me know. Thank you

static isle
# solar niche Help me. As in the docs I read the GPT 4 1106 and vision along with the 3.5 tubo...

The response for a completion looks like the following, which contains how many tokens were outputted

{
  "choices": [
    {
      "finish_reason": "stop",
      "index": 0,
      "message": {
        "content": "The 2020 World Series was played in Texas at Globe Life Field in Arlington.",
        "role": "assistant"
      }
    }
  ],
  "created": 1677664795,
  "id": "chatcmpl-7QyqpwdfhqwajicIEznoc6Q47XAyW",
  "model": "gpt-3.5-turbo-0613",
  "object": "chat.completion",
  "usage": {
    "completion_tokens": 17,
    "prompt_tokens": 57,
    "total_tokens": 74
  }
}```
solar niche
#

Oh, doesn't that mean it depends on CONTEXT WINDOW?

static isle
#

Though I cannot find any docs to back that up

solar niche
#

yes, thank you very much

solar niche
static isle
main dock
#

I've a paid subscription with ChatGPT.. Lately, I'm seeing error message: Message in conversation not found, when I ask simple questions about the uploaded PDF file.. can someone suggest what can be done here..?

#

and then is gets stuck at Regenerate

blazing scroll
main dock
#

Its not going away.. somehow..
I uploaded a 4 PDF file.. and asking queries on these PDF files... is was working like a charm.. suddenly it has started to give this error..

#

ChatGPT gets stuck in "Regenerate"

vocal parcel
# unreal moon then why arenā€™t you rich from opening btc/eth wallets

because with my resources, the chance of hitting a single wallet could still take up to 30 million years just to find one. I have the script, and I know it works as I narrowed it down to my own wallets, but running it on everyone's wallets is pointless without an army of 30 million computers, and even that would get me only 1 wallet per year

vocal parcel
vocal parcel
opal sapphire
wet drift
#

So I've heard that the token size has increased dramatically so now you're able to have book sized tokens. However I also heard that it loses memory of the content. But that it is also better remembering the beginning and the end of the content. However I'm wondering how that is useful to give a higher token size if it can't remember everything you upload. What would be the point?

weary patio
#

300 pages of data to be precise, which is a small book.

#

So if you upload more than that amount, it will have a hard time remembering the information.

rugged summit
#

I've been working on creating a GPT agent trained to adopt the personality and stereotypical manerisms of a Ferengi (from the Star Trek Universe) with very little success. I've created Markdown docs with copies of Memory Alpha and Beta on language, culture, history, etc. I've manually created a copy of the Rules of Acquisition and emphasised their religiously zealous adherence to the pursuit of profit, etc. And yet, when I run tests and trials it still ends up being very polite and courteous, and if anything sounds more robotic than ever. Any tips on how to shake the model out of it's courteous and unnecessarily fluffy tone?
To see a snippet of my instruction that focuses on personality traits go to snippet . host/pmmzxc. Note: Krax is the name of the GPT agent.
Any help or suggestions would be greatly appreciated!

gloomy olive
#

i tried to make what you noted, and its acting near perfectly as a Ferengi šŸ˜„

rugged summit
rugged summit
sonic edge
#

why did they put the limit to 40 messages /hours in germany?

supple violet
sonic edge
gloomy olive
snow willow
#

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

weary patio
snow willow
#

Ty

true mist
#

gee, [its] pea tea!!

wind urchin
#

is gpt working for y'all?
i got premium and cant login

long pond
knotty grotto
naive spire
grizzled flax
naive spire
glacial mountain
#

How can i bypass this nonsense in GPT 4:

"I can certainly guide you through understanding the concepts and provide advice on how to approach the assignment, but I cannot complete the assignment for you. It's important for your learning and adherence to academic integrity policies that you attempt the work yourself."

since when did these guys make GPT follow academic integrity
now i just have to figure out how to word the promt differently in a way to make it seem like its not doing my assignment for me lol what a waste of time

opal sapphire
elder skiff
#

Hi, I'd like to know what the MB limit is for importing a file into Custom GPT? Thanks!

grizzled flax
elder skiff
#

Thank you very much! ā¤ļø

tropic pier
grizzled flax
tropic pier
muted hemlock
#

Hi! any idea how long is taking the waiting list to get ChatGPT 4?

tender tartan
#

well its not even working for paid customers so dont get your hopes up

knotty grotto
muted hemlock