Playground

The Playground is a built-in chat interface for testing models loaded in Salut. Open it from the menu bar icon → Playground… or with the global hotkey.

Hotkey

A global keyboard shortcut that opens the Playground window from anywhere. Default: ⌥⌘S.

Click the recorder button to set a new shortcut, press ⌫ to clear it. Changes take effect after restarting Salut.

The hotkey requires Accessibility permission. If not already granted, click Grant Access… in the Playground settings pane and approve the system prompt.

Float Above Other Windows

When enabled, the Playground window stays above all other windows — useful for keeping it visible while working in another app. Disable this if you prefer normal window layering.

Temperature

Controls the randomness of model output. Range: 0.0–2.0, default: 0.7.

  • Lower values (0.0–0.3) produce more focused, deterministic responses.

  • Higher values (0.8–2.0) produce more creative, varied responses.

Max Tokens

The maximum number of tokens the model can generate per response. Default: 2048, maximum: 131072.

Higher values allow longer responses but take more time and memory. Most conversational replies fit well within 2048 tokens.

System Prompt

Optional instructions prepended to every conversation. Use this to set the model’s behavior, persona, or constraints.

The system prompt is sent as the first message in the conversation and persists across the “New Chat” button — change it in settings to update future conversations.

Using the Chat

The chat interface streams responses from whichever model is selected in the dropdown. If no models are loaded, it shows a prompt to load one in Settings → Models.

  • Send: press Enter or click Send.

  • New line: press Shift+Enter.

  • Stop generation: click the Stop button during streaming.

  • New conversation: click New Chat to clear history and start fresh.

  • Thinking mode: toggle the Thinking checkbox to enable extended reasoning (if supported by the model).