Is there perhaps any plan to implement group-based limits? For example “Group A” can send 100 messages or 10,000 tokens to AI Bot per day?
Agree, we would very much like to build a system for quotas
I just discovered one issue. My bot wasn’t writing back to me so I checked the logs and I saw that the temperature was out of range.
OpenAI models have a temperature range 1 - 2 whereas Anthropic models have a range of 0 - 1. So, if you set a temperature of, for example, 1.2 in your persona, then you can not use an Anthropic model in your LLM dropdown list when you go to chat with the bot (and vice-versa with OpenAI).
This could prove to be a limitation because it forces users to either:
- Use ONLY the default values if there are models from companies with different ranges
- Use ONLY models from one company so that you can actually set the temperature.
I wanted to provide a selection of models to my users, however currently I can’t let them choose between OpenAI or Anthropic models because I need to set a temperature
Coming back to this, and just throwing an idea up into the air…
Stripe subscriptions with metered usage looks like it might play well with this idea… Quite similar to how Mailgun operates: ‘x’ amount of tokens allowed per month. Additional ‘y’ amount of tokens used incurs an additional ‘z’ - or simply block the user’s AI access after ‘x’ is used up. The user would be able monitor their token usage via their profile of course
2 posts were split to a new topic: Help with Installing AI Plugin on Discourse v3.4.0.beta1
FYI to all the folks
Hey everyone!
Looking for setting for the AI editor found in the composer editor to enable include image generation. I have OpenAI setup and purring nicely and had accidently had that enabled when I first installed but having trouble finding. There are so many settings that it takes time to master them. Been at it about 2 weeks now and loving it!
Solved my own issue . I needed to rebuild Discourse after setting options and then the settings to select Illustrate Post in the AI Modal appeared. Works great!
Hi everyone,
I’m currently using GPT-4o Mini with the Discourse AI plugin and wondering if it’s possible for my AI bot to read and process the contents of uploaded documents directly, similar to what can be done on ChatGPT. So far, I haven’t found a way to enable this functionality.
If this feature is already available and I’ve missed it, could someone point me in the right direction? Alternatively, if it’s not yet possible, is this something that might be planned for a future update?
I’d love to have my bot read documents uploaded in the composer for more dynamic and interactive discussions. Any guidance or insight would be much appreciated!
Thanks in advance!
This should be possible through the use of custom AI bot - Personas. When creating a persona at the very bottom, you should see a way to upload documentation so the bot can refer to it when responding to requests.
@Saif thanks for the feedback. I am aware of that feature and already use that as part of the persona. Works great! I’m referencing allowing a user who is using the chatbot to upload a document and to ask questions about it. My chatbot has been unable to do this. If I take a snapshot with the information on it and upload as an image to the chat bot it is able to read and respond to it without issue. Does not seem to be able to read anything from a document though?
Ahh gotcha, thank you for clarifying. Currently, we don’t support this use case as we are more focused on the admins setting up the persona with the supported documents.
Perhaps something for us to think about in the future.
Thank you @Saif and might I say you and the entire team are doing a fantastic job. This is a very nice experience working with your product!
I do like the idea of allowing people to upload giant PDFs and TXT files and ask them questions in a persona direct, I think it makes sense to make a feature request for this.
We have a lot of the infrastructure to enable this already.
Take my breath away @sam You guys are the best.
Can the Ai bot interact in topics if it’s mentioned?
Yes, if an admin has allowed that.
You’re threatening us with a good time my friend. This is definitely the cherry on top of the delicious dessert which the team has already given us. But it’s also poetic in a sense; the idea of an open source technology enabling easy access to intelligent document chatbots using (possibly) open source models is just lovely, and I really support that as a pro open source AI researcher.
Otherwise people have to rely on running small models locally with ollama, or trust all of the closed source solutions such as ChatGPT. What I mean is that although Discourse is (and should always be) a community discussion software, it has the potential to do a lot of good for a lot of people the further the team goes in this direction.
I personally work with many people who can’t really use any AI in their work (ChatGPT, Claude, Perplexity, etc) because of privacy and other concerns. They also don’t have and can’t afford dedicated AI engineers to build a custom solution, nor are they technical individuals themselves. So I see a very large market gap here for small and medium businesses/organizations, where Discourse AI is the perfect solution. Actually, my newly founded company wouldn’t even exist without Discourse AI.
Y’all have opened up a gold mine here.
Can you specify which setting is to be activated? I know the merefield chatbot can be invoked to join public chats but for the Ai plug-in bot I’m unaware.