LLM Quick Actions

Performs one-shot actions (polish writing, summarize, translate, etc.) with multiple LLM providers using Raycast.

Supported Providers

• OpenAI - GPT-5.x, GPT-4.1 series
• Anthropic - Claude Sonnet, Haiku, Opus 4.5
• Google - Gemini 2.5, 3.x series

Features

• Real-time streaming responses
• Per-command model selection or use global default
• Token counting and cost estimation
• Keyboard shortcuts for quick regeneration

Commands

Command	Description
Summarize	Summarize selected text
Rewrite	Rewrite with academic tone
Refine	Fix grammar and improve writing
Custom Action	Use custom prompt
Execute	Replace selection with LLM output
Preview	Preview LLM output without replacing
Transform	Replace with prompt entered at runtime
Transform Preview	Same as Transform but preview first
Translate	Translate to target language
Explain	Explain in simple terms
Reply	Draft a reply to a message

Setup

1. Get API keys from your preferred provider(s):

   • OpenAI: platform.openai.com
   • Anthropic: console.anthropic.com
   • Google: aistudio.google.com

2. Configure API key(s) in extension preferences

3. Select your default model

Tips

• Use Cmd + R to regenerate results
• Use Cmd + Shift + R to retry with GPT-5.2
• Use Cmd + Option + R to retry with Claude Sonnet 4.5
• Use Cmd + Option + Shift + R to retry with Gemini 3 Pro
• Each command can override the default model in its preferences

Development

# Generate package.json from template (after editing models)
npm run build-package

# Build extension
npm run build

# Development mode
npm run dev

Model definitions are centralized in scripts/models.js.