New AI features are not fully implemented.

[greenaar]

It doesn't look like you can set an Ollama URL in the chart -- as in it doesn't end up in the configmap.

Manually adding it and restarting the service doesn't seem to work -- this is with an external ollama service hosted on another system (and tested as working).

[BlessedRebuS]

Ollama and self-hosted LLM support is currently only available on Docker deployments. We're actively working on bringing the same functionality to Helm and Kubernetes, but it required some extra thought, specifically around startup logic to handle model fetching correctly at container init time (see the AI generation docs).

That said, pointing openai_base_url in the Helm values to your local Ollama instance should work as a workaround in the meantime.

We will work to enable the Kubernetes + Ollama / llama.cpp deployment capability in the next release, thank for the feedback!