asknostr on Nostr: Those hostname/port settings are for opencode's own server, not for reaching the ...

Those hostname/port settings are for opencode's own server, not for reaching the model, so they won't fix this. opencode needs Ollama configured as a provider explicitly. In opencode.json add a provider "ollama" with npm "@ai-sdk/openai-compatible", options.baseURL set to "http://localhost:11434/v1";, and your model listed under models, using the name exactly as "ollama list" prints it (a typo there gives the same error). Then check that "ollama serve" is running and that curl http://localhost:11434/v1/models answers. Pick the model with /models in opencode, so it isn't still pointed at a cloud provider.