AI Bots 3.0.2

[021]

021 submitted a new resource:

[021] AI Bots - Multifunctional artificial intelligence forum bots.

Artificial intelligence on the forum? Ooh yeah :cool:​

A powerful tool that will blow your forum activity to the skies 🌠 thanks to two built-in handlers: ChatGPT and GPT Dall-E. Bots on these handlers are capable of:

• Reply in threads to quotes, mentions and posts posted in specified nodes.
• Reply to quotes, mentions and private messages in conversations.
• Generate images in three sizes up to 4 pieces, understanding messages in any language...

Read more about this resource...

 

[scriptop]

Legendary stuff

 

[ProWeb]

Brilliant! I'm looking forward to setting this up and testing it!!!

 

[colcar]

Brilliant! I'm looking forward to setting this up and testing it!!!

Tell me how you get on and what you use it for (y)

 

[021]

021 updated [021] AI Bots with a new update entry:

1.0.1

Now when permissions associated with reply limits are set to 0 (default value), they are not ignored.
Fixed a error that occurred when triggering a GPT Dall-E based bot in private messages

Read the rest of this update entry...

 

[021]

021 updated [021] AI Bots with a new update entry:

1.0.2

Fixed: Some template modifications might not apply to new versions of XenForo

Read the rest of this update entry...

 

[akok]

Interesting mod, although it draws curves in the images and it's a bit expensive compared to the prices of a text-based model. There is no translator yet, but will it be possible to set it up for a specific language, not just English?

 

[021]

Interesting mod, although it draws curves in the images and it's a bit expensive compared to the prices of a text-based model. There is no translator yet, but will it be possible to set it up for a specific language, not just English?

A translator is just a properly composed prompt for a bot, you can compose it in such a way that it translates into any language :)

 

[akok]

I was expecting a separate bot))) As for the hints, I'm looking into it. Is it possible to set up the bot to search the forum before giving an answer (for example, through Google)? If I ask to search through hints, old links are displayed.

 

[021]

I was expecting a separate bot))) As for the hints, I'm looking into it. Is it possible to set up the bot to search the forum before giving an answer (for example, through Google)? If I ask to search through hints, old links are displayed.

No, unfortunately it's not possible

 

[akok]

It's a pity, although the bot (according to its words) can search on Google. I managed to make the bot search for information on a sitemap based on a request, and then extract the context of the found topic or provide an answer based on the found instruction.

 

[021]

It's a pity, although the bot (according to its words) can search on Google. I managed to make the bot search for information on a sitemap based on a request, and then extract the context of the found topic or provide an answer based on the found instruction.

ChatGPT may provide incorrect information.
ChatGPT does not have access to the internet, so it cannot make any requests to the network

 

[akok]

I agree, it (ChatGPT) can't. Although it has a sitemap of the forum in its database, which surprised me. I've come across Android apps that "give access" to ChatGPT for current information. Is it possible to implement this as an extension?

Smart ignore in threads - не робить

 

[sevenmix]

whatt are the limits in the prompt input?

i will like to have my rules on it, so the users could use it as an assistant

 

[sevenmix]

Another thing, the regex isnt working with conversations right? it will be great

 

[021]

I agree, it (ChatGPT) can't. Although it has a sitemap of the forum in its database, which surprised me. I've come across Android apps that "give access" to ChatGPT for current information. Is it possible to implement this as an extension?

I think yes.

Smart ignore in threads - не робить

The function works, but, as stated in the explanation to it, it requires fine tuning of the bot – reducing the context, specifying the bot name in the prompt, etc.
Also, please note that the function is experimental and does not guarantee a 100% result and can indeed skip messages that are not related to the bot.

---

whatt are the limits in the prompt input?

You have 4096 tokens available for input and output in the API. That is, theoretically, you can upload information up to 4000 tokens, but then this can create problems with a response. The recommended limit will be approximately 2000-2500 tokens so that you have some margin for context and response.

Another thing, the regex isnt working with conversations right? it will be great

Regexes work in conversations, but the bot does not respond in those conversations in which it is not a recipient.

 

[akok]

The function works, but, as stated in the explanation to it, it requires fine tuning of the bot – reducing the context, specifying the bot name in the prompt, etc.
Also, please note that the function is experimental and does not guarantee a 100% result and can indeed skip messages that are not related to the bot.

Correctly, when I set up the bot, it started ignoring user mentions. But I can't get it to ignore other people's quotes, the bot responds to them

 

[akok]

Some users complain that they cannot ask the bot a question because their IP addresses are being blocked on the OpenAI side. Is there any plan to add the ability to use proxies in the mode?

 

[021]

Some users complain that they cannot ask the bot a question because their IP addresses are being blocked on the OpenAI side. Is there any plan to add the ability to use proxies in the mode?

Requests to the OpenAI API come from the forum server, so users' IPs have no effect on whether users can use the bot or not

 

[ProWeb]

ChatGPT 3.5 turbo has a token limit on the input/output of 4k.

... so I have a question about this setting: "Context limit in threads"


Does this contribute to each request's token amount? For example, if the last 10 messages have 8k tokens, will some of that be truncated, and will that impact or limit the reply / output from the bot?

If I were to put "100" in that field, and the text in a thread's replies total 50k tokens, does the bot only use the most recent posts, does it reject the request, or something else?

Similarly, does the bot "prompt in threads" text also contribute/count towards the tokens input and limit?

Hopefully that makes sense.