I have a question regarding the AI plugin. Currently, I only see OpenAI as an option. Is it possible for us to use a local model, such as Ollama, or another program that lets us host our own models? This would help us save on costs since the tokens would be free.
AI-generated summary
In this Support thread titled Inquiry About AI Plugin Options, beitmenotyou asks if local LLMs or alternative providers like Ollama can be used instead of only OpenAI. Falco confirms the plugin is provider-agnostic and supports local LLMs. sam suggests OpenRouter for free, decent models like Grok 4.1, noting that local models may be weaker but still viable, and mentions full support for AI in Discourse. NateDhaliwal adds that Gemini is also a free and very capable option. beitmenotyou follows up asking how to use local LLMs, then again asks how to use Gemini. NateDhaliwal provides helpful links: Discourse AI LLM Settings Page, Open Source LLM Hosting for Discourse, and Discourse Meta. The thread highlights flexibility in AI provider choice, including local, free, and cloud-based models, with clear documentation links for setup.
Yes, the plugin is completely provider agnostic, you may use it with a local LLM out of the box.
Note if you are looking for free openrouter always have a few models that are free and decent, for example grok 4.1 fast a super competent model is free today (and this changes every few days).
This is fully supported in Discourse AI.
Local models may work but generally are pretty weak so your results will be quite uneven.
Not to mention Gemini… IMO the free plan is very generous.
if I can usen local LLM how do I
how do I use gemini on my fourm please