[3.2][3.3][RC] AI Labs with Midjourney, Mureka, FaceSwap, Pika, Claude, ChatGPT, Gemini, DALL-E, Stable Diffusion

[Miri4ever]

Wow, this sounds great and 75 Images per day is even more enough we don't use withing a month, from this daily free limit if they're free accounts as you wrote. But even if I can do a signup there, the current version cannot choose it for the bots' setup yet, right? I already paid a small amount for the normal Stable Diffusion Account, but Leonardo seems far beyond and even more promising images.

I would be bad if members loose money just because I needed to delete their posts, if for example, DALLE reaches the limit of 10 images while in a time. But I developed a way to get those out, without destroying my users' balance.

Sure I don't have to do more as only ounces, but we should use PHPBB Standards to pick usernames and not developing own things. I think it's simple to convert all usernames into lowercase, for example after input them.

[klingklang]

Hi, I love this idea and would really like to implement it in my board Unfortunately, no matter what I try, I only get the "AI is thinking..." message, with an empty log. I don't have the Login required extension (but I also tried to install and configure it as you suggest), nor the No quote on last post one; I have several more though. I upgraded php to 8.2 and modified the security session validation setting as instructed. None of this worked. Actually, under php 7.3 (which I had at the time of installation) I got the "Started at hh:mm:ss" in the log (and nothing else after that), while after upgrading to 8.2 I just have a blank log. Any ideas? Thanks!

EDIT: I got it to work with PHP 7.4. 8.x won't work though

[privet.fun]

Wow, this sounds great and 75 Images per day is even more enough we don't use withing a month, from this daily free limit if they're free accounts as you wrote. But even if I can do a signup there, the current version cannot choose it for the bots' setup yet, right? I already paid a small amount for the normal Stable Diffusion Account, but Leonardo seems far beyond and even more promising images.

Miri4ever, I'll let you know once this is ready, I still need to add UI and few more things before it can be ready for prime time.
Unlike paid Stable Diffusion/DALL-E it will be flat-fee approach where you pay $10/month and get up to 1500 images generated daily (which is a LOT and probably will suffice for very big forum).
There's also option to provide & embed original Leonardo AI image url so you do not have to worry about storing those images locally but trade off is that Leonard may decide to wipe them out on their own schedule.

I would be bad if members loose money just because I needed to delete their posts, if for example, DALLE reaches the limit of 10 images while in a time. But I developed a way to get those out, without destroying my users' balance.

Sorry, I'm not following, can you please provide more details?

[privet.fun]

EDIT: I got it to work with PHP 7.4. 8.x won't work though

I have it running on 8.x on my forum (production) and 7.4.x on my test local dev box.
Make sure that you installed and configured php with 8.x upgrade, my bet something was missing during upgrade.

[amorino]

Hello @privet.fun
1- any news to reduce the quote text ? like adding a new btn to quote only 1st line of answer

2- I tried to change to 16k gpt-3.5-turbo-16k and raised "max_tokens": 16384, and "message_tokens": 16000,
It worked but even with quotation bot lost context. New answers have no relation with the discussion

3- Gpt told me that text-davinci-003 is more powerful, is there a chance to integrate it oin your module please?

Thanks again

[privet.fun]

1- any news to reduce the quote text ? like adding a new btn to quote only 1st line of answer

I'm planning to release it this month along with leonardo.ai text-to-image generation support.

2- I tried to change to 16k gpt-3.5-turbo-16k and raised "max_tokens": 16384, and "message_tokens": 16000,
It worked but even with quotation bot lost context. New answers have no relation with the discussion
3- Gpt told me that text-davinci-003 is more powerful, is there a chance to integrate it oin your module please?

You can look at the log, locate request.message JSON array and verify that all expected context is there.
It will look something like this:

Code: Select all

"request.messages": [
    {
      "role": "user",
      "content": "Tell me a joke?"
    },
    {
      "role": "assistant",
      "content": "Why don't scientists trust atoms? Because they make up everything!"
    },
...]

If you see entire expected conversation in that request.messages array then problem is with OpenAI since this is what sent to their API.
If you don't see entire / expected conversation then there's an issue with AI Labs plugin code.
I did tested 16K models and it worked just fine since there's really no difference from 4K model, it's same exact code.

text-davinci-003 only available with Completions API https://platform.openai.com/docs/api-re ... ompletions, it is not supported by chat API https://platform.openai.com/docs/api-reference/chat.
Not sure what advantages it has over gpt-3.5-turbo.
Have you tried gpt-4 model?
It supposedly best available model as of today.

[amorino]

Many thanks for all your answers
Gpt4 is not yet available for my account
Any advise to get it soonest in my list models ?

[privet.fun]

Gpt4 is not yet available for my account
Any advise to get it soonest in my list models ?

I do not recall doing anything special https://openai.com/blog/gpt-4-api-general-availability, you should get it by eom or perhaps email their support?

July 6, 2023...
Today all existing API developers with a history of successful payments can access the GPT-4 API with 8K context. We plan to open up access to new developers by the end of this month, and then start raising rate-limits after that depending on compute availability.

[amorino]

Today all existing API developers with a history of successful payments can access the GPT-4 API with 8K context. We plan to open up access to new developers by the end of this month, and then start raising rate-limits after that depending on compute availability.

[/quote]

Here are models I see (see image)
What setting did you put for the GPT 4?
Here are my settings :

Code: Select all

{
  "api_key": "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx",
  "url_chat": "https://api.openai.com/v1/chat/completions",
  "model": "gpt-3.5-turbo",
  "temperature": 0.2,
  "max_tokens": 4096,
  "message_tokens": 2048,
  "top_p": 1,
  "frequency_penalty": 0.8,
  "presence_penalty": 0.8,
  "prefix": "",
  "prefix_tokens": 0
}

mpdels.PNG

[privet.fun]

Here are models I see (see image)
What setting did you put for the GPT 4?

You can see list of available for your account models at Playground https://platform.openai.com/playground, select Chat and then select desired model.
You can switch to gpt-4 by replacing gpt-3.5-turbo with gpt-4 in your settings.
Check stats under your account https://platform.openai.com/account/usage to see what models you are using, how many tokens and what are the costs, just keep in mind that usage stats lagging some 10-15mins behind.
Cost of using gpt-4 may be unpleasant surprise, keep that in mind.

[amorino]

Here are models I see (see image)
What setting did you put for the GPT 4?

You can see list of available for your account models at Playground https://platform.openai.com/playground, select Chat and then select desired model.
You can switch to gpt-4 by replacing gpt-3.5-turbo with gpt-4 in your settings.
Check stats under your account https://platform.openai.com/account/usage to see what models you are using, how many tokens and what are the costs, just keep in mind that usage stats lagging some 10-15mins behind.
Cost of using gpt-4 may be unpleasant surprise, keep that in mind.

I found gpt-3.5-turbo and 3.5-turbo-16k

what settings do you advise for a better results 3.5-turbo-16k as I don't have the 4 for now ?

If I pay GPT plus will I be able to get GPT4 do you think ?

[privet.fun]

what settings do you advise for a better results 3.5-turbo-16k as I don't have the 4 for now ?

Changing setting very straightforward, you just need to update model name and mode maximum allowed tokens:
"model": "gpt-3.5-turbo-16",
"max_tokens": 16384,

If I pay GPT plus will I be able to get GPT4 do you think ?

Sorry, I have no idea.
I'd suggest asking OpenAI support to enable gpt-4 on your account.

[nikk106]

Hello privet.fun,
Tried many times with different usernames and different forums. Every time I get an error message. What am I doing wrong?

Pict-1.png

Pict-2.png

I'm using
php version 8.2.8
phpBB 3.3.10
mysql ver 8.0.33

[amorino]

Hello privet.fun,
Tried many times with different usernames and different forums. Every time I get an error message. What am I doing wrong?


Pict-1.png

Pict-2.png

I'm using
php version 8.2.8
phpBB 3.3.10
mysql ver 8.0.33

Could you try to change these settings :

Code: Select all

"prefix": "", "prefix_tokens": 0

[nikk106]

Hello privet.fun,
Tried many times with different usernames and different forums. Every time I get an error message. What am I doing wrong?

Pict-1.png

Pict-2.png

I'm using
php version 8.2.8
phpBB 3.3.10
mysql ver 8.0.33

Could you try to change these settings :

Code: Select all

"prefix": "", "prefix_tokens": 0

Tried it, same result. Still getting the error message "Please select at least one forum". The extension doesn't seem to see my forums at all.