Itās Discourse Chatbot š¤. The clue is in the name (although you can change its name!) Please post issues in the Chatbot Topic. But in any case the model will be set in the Chatbot settings :).
Suffice to say activate Chatbotās verbose logs in settings, diverting info to warn (so you can read them in prod) and check /logs
Yes I understand that AI can be of different vendors, not necessarily from Discourse.
By this I understood that āChatbotā has been developed/is handled by Discourse, but no any 3rd party AI Vendor. And so I found/paid attention to this admin setings page: ā¦plugins/discourse-chatbot/settings (which I earlier missed).
But even now (after this new knowledge), it always tries to reply/keeps showing āReplyingā¦ā. (Iām not able to decipher āLogsā or sidekiq msgs and find whatās the culprit).
If you can tell if the other bot (turned on from below named setting) is more exhaustive to use (and easy to configure)?
AI bot add to header >> Display a button in the header to start a PM with a AI Bot
If you would pls, Why the dropdown to control which models can it (the AI Bot Plugin) use only offers 'Chat Gpt' models only, and how can it use them, even wo asking my chatgpt Api key (does it has access to my ChatGpt api key from the AI settings that I configured elsewhere)?
And last, would you opine that this ChatBot is better than the other offered under AI settings?
Iād now rather go to this Plugin page to clear my above given doubts!
Today Iām releasing a small update to Chatbot but one that fixes a nasty bug and introduces RAG to reasoning models:
If you think Chatbot is awesome and I did a great job, you can buy me a coffee, and if Chatbot is important to your business or organisation, please consider sponsoring me on one of the higher tiers . This will all help to guarantee timely updates to Chatbot and its future.
Hello, when I select the default openai mpdeli o3 mini, the plugin works properly and responds, but it does not respond to posts with photos. The vision setting is directly selected, there is only the gpt 4o setting, is that why I get an error?
I set it this way and did tests, it cannot see the content of the images and does not respond. Could it be because I use all trust levels o3-mini? When I switch to all trust levels 4o, it responds.
If you enable the experimental setting chatbot_support_pdf it will now send the contents of any uploaded pdfs within the current conversation to OpenAI which will form part of the context of the conversation.
TIP: with every upload in the public area ask Chatbot for a document summary, then it will embed the resulting answer and it will be easy to ask Chatbot later to find it.
NB: this may only work with direct calls to OpenAI. If you are using other base URLs to invoke other services there is no guarantee they will support it.
Is it possible for a new forum to run this bot with a limited budget or no cost at the start? I think it seems interesting and I would like to try it out but I am afraid that the quota will be reached or it will cost me more than expected. I read through OpenAI:s pricing model but I thought it was a bit vague.
Hey @Olle11 itās not possible to run this plugin for āfreeā unless you set up a local LLM on your infrastructure but thatās ironically an expensive option because youāll need a very beefy server indeed.
However, you may get some free credits ($5?) when you set up your OpenAI account (link in settings)
You can set quotas for different levels of users or only make it available to admins if you prefer.
Using it with gpt-4o-mini should not break the bank and that model is powerful enough for some great interaction.