MiniMax - Raycast Extension

A "Bring Your Own Key" Raycast extension for AI chat. Currently supports MiniMax M2.1 with streaming responses.

Features

• Conversational chat with persistent history
• Streaming responses in real-time
• Quick question (Ask AI) for simple queries
• Conversation history integrated in the main view
• Automatic filtering of model "thinking" content
• Concise Mode for brief, focused responses

Installation

# Clone the repository
git clone https://github.com/MonforteGG/raycast-minimax
cd raycast-minimax

# Install dependencies
npm install

# Run in development mode
npm run dev

Configuration

Open Raycast → Search for "AI Chat" → Cmd + , to open preferences:

Preference	Type	Description
MiniMax API Key	password	Your MiniMax API key (required)
Model	dropdown	MiniMax-M2.1 (recommended) or MiniMax-M2
System Prompt	text	Custom system prompt (optional)
Temperature	dropdown	0.3 / 0.7 / 1.0 / 1.5
Max Tokens	dropdown	1024 / 2048 / 4096 / 8192
Stream Responses	checkbox	Enable streaming (default: true)
Concise Mode	checkbox	Brief 2-3 sentence answers unless more detail requested (default: true)

Getting a MiniMax API Key

1. Visit MiniMax Platform
2. Create an account or sign in
3. Navigate to API Keys section
4. Generate a new API key
5. Copy and paste it into the extension preferences

Commands

AI Chat

Conversational chat with integrated history.

• Top bar: Type your message
• Enter: Send the message
• Left panel: New Chat + conversation history
• Right panel: Current conversation
• Cmd + N: New conversation
• Cmd + Backspace: Delete conversation
• Cmd + C: Copy conversation

Ask AI

Quick question with streaming response.

• Type your question in the form
• See the response in real-time
• Actions: Copy, Paste, Continue in Chat

Project Structure

raycast-minimax/
├── package.json              # Raycast manifest
├── tsconfig.json
├── assets/
│   └── icon.png
├── src/
│   ├── ask-ai.tsx            # Command: quick question
│   ├── ai-chat.tsx           # Command: chat with history
│   ├── providers/
│   │   ├── base.ts           # Provider interface
│   │   └── minimax.ts        # MiniMax M2.1 implementation
│   ├── hooks/
│   │   ├── useChat.ts        # Main chat hook
│   │   └── useChatStore.ts   # Chat state management
│   ├── components/
│   │   ├── ChatView.tsx      # Chat view
│   │   └── QuickAIResult.tsx # Quick response view
│   └── utils/
│       ├── storage.ts        # LocalStorage persistence
│       └── errors.ts         # Error handling

MiniMax M2.1 API

Endpoint: https://api.minimax.io/v1/chat/completions

Models:

• MiniMax-M2.1: Recommended model, optimized for chat
• MiniMax-M2: 200k context window for longer conversations

The provider automatically filters <think>...</think> content generated during the model's internal reasoning.

Error Handling

• 401: Invalid API key → Opens preferences automatically
• 429: Rate limit → Wait message
• 500+: Server error → Retry message

Development

# Development with hot-reload
npm run dev

# Build
npm run build

# Lint
npm run lint

# Fix lint
npm run fix-lint

Dependencies

• @raycast/api: ^1.93.0
• @raycast/utils: ^1.19.0

License

MIT - see LICENSE for details.