In a Dev category discussion titled Possible to use Perplexity.ai?, denvergeeks asks if the Perplexity Pro API can be integrated with a self-hosted Discourse instance, noting they’ve built a plugin via Claude 3.5 but it’s flagged as unofficial. Falco reassures them that custom plugins can do anything, and suggests using the built-in Discourse AI feature — which should support Perplexity out of the box via LLM configuration. denvergeeks replies that Perplexity doesn’t appear as a listed provider in settings, only a blank Custom template, and asks for next steps. Falco clarifies that since Perplexity follows the OpenAI API standard, users should select “OpenAI” in the settings and manually input the Perplexity API URL, model name, and key. This workaround allows integration without needing a dedicated provider entry.
If you are authoring a new plugin you can do anything.
That said, if all you want covered by the existing Discourse AI feature set, all you need is to install Discourse AI and go to the LLM configuration setting and configure it there, as we should support Perplexity API out of the box.