Coffee Server + Ollama proxy — coffee.serverReachability finds the API;
chat goes browser → POST /api/sidecar/ollama/chat → Ollama.
Open via npm start and /sidecar2-poc. Slots: GET /api/sidecar/poc/containers, then ../sidecar-internal.json if that fails. The Ollama card runs ensure when needed (skips if API is already up). First chat may pause briefly while Ollama loads the model into memory (normal even if the model is already pulled). Commands work with or without a leading /.
Containers
Coffee …
Ollama …
Starting…
Quick Brew Registry
ollama-gemma3
Same as card — ollama run gemma3 via ensure + chat.