Getting started

The whole loop: start Ollama, load a model, chat, edit its Modelfile, save.

1. Start Ollama

Remuda is a client — it doesn't bundle Ollama or run inference itself. Install and start it first:

brew install ollama
ollama serve

Remuda's connection pill goes green once it can reach the server, by default at http://127.0.0.1:11434.

2. Install Remuda

brew install --cask magna-nz/tap/remuda

See the overview for the upgrade and uninstall commands, and what to do if macOS quarantines the app anyway.

3. Load a model

Click the model control in the top bar. It opens the load pane: every model ollama list knows about, with tuned variants grouped under their base. Pick a model and a Modelfile — the base ("Original"), or a tuned variant — and hit Load. A progress bar tracks the warm request; the pane's runtime readout then shows the VRAM/RAM split, the context the runner started with, and a countdown to keep_alive expiry.

No models installed yet? The load pane offers Pull your first model, with per-layer progress and a curated list of popular models.

4. Chat to test it

Hit ＋ New chat — it opens on whatever model is currently loaded — and say hello. Replies stream in; Enter sends, Shift+Enter starts a new line. Messages test the model; they never change its saved Modelfile.

5. Tune it

Not the behaviour you wanted? Open the Modelfile editor without leaving the chat — the chat list stays visible. Edit the system prompt or a parameter, hit Save. Remuda stops the model, rebuilds it through ollama create, and reloads it, so your next message uses the change.

Happy with a variant and want to keep the original untouched? Save as… forks it under a new name.