Skip to main content

​
Environment Variables

Cluely supports multiple AI providers through environment variables. Create a .env file in the root directory to configure your preferred AI provider.

​
AI Provider Options

Google Gemini provides the latest AI technology with vision capabilities and fastest responses.
GEMINI_API_KEY=your_primary_api_key_here
GEMINI_FALLBACK_API_KEY=your_backup_api_key_here
The fallback API key is optional but recommended. If the primary key hits rate limits, Cluely automatically switches to the fallback key.
Get your API key: Google AI StudioConfiguration details:
  • Default model: gemini-2.5-flash
  • Automatic rate limit handling with exponential backoff
  • Supports both image and audio analysis

​
Application Settings

​
Window Configuration

Cluely’s main window is configured in electron/main.ts with these settings: 
// Window dimensions
width: 800
height: 600

// Window behavior
alwaysOnTop: true        // Stays above other windows
focusable: true          // Can receive keyboard focus
resizable: true          // User can resize
frame: false             // Frameless window
transparent: true        // Transparent background

​
Screenshot Settings

Screenshot behavior is controlled in electron/ScreenshotHelper.ts:
  • Maximum screenshots: 5 per queue
  • Storage locations:
    • Primary queue: ~/Library/Application Support/interview-coder/screenshots/
    • Extra queue: ~/Library/Application Support/interview-coder/extra_screenshots/
  • Format: PNG
  • Auto-cleanup: Old screenshots are automatically deleted when queue exceeds limit
When you take a screenshot:
  1. Window hides automatically (100ms delay)
  2. Screenshot is saved to appropriate queue directory
  3. If queue exceeds 5 screenshots, oldest is deleted
  4. Window reappears after capture
  5. Screenshots are deleted from disk when removed from queue

​
Model Configuration

​
LLM Parameters

All AI providers use these default parameters (defined in electron/LLMHelper.ts): 
temperature: 0.7    // Creativity vs consistency
top_p: 0.9          // Nucleus sampling (Ollama)
max_tokens: 4096    // Maximum response length (OpenRouter)

​
System Prompt

Cluely uses a specialized system prompt to provide helpful, structured responses: 
You are Wingman AI, a helpful, proactive assistant for any kind of problem or situation.

CRITICAL: You MUST use Markdown for all responses.
1. Use headers (#, ##), lists (* or 1.), and bold text
2. Use LaTeX for ALL mathematical formulas and equations
   - Block equations: $$x = \frac{-b \pm \sqrt{b^2-4ac}}{2a}$$
   - Inline math: $E=mc^2$
3. Use code blocks with language specification
4. Analyze the situation and suggest several possible responses
5. Always explain your reasoning

​
Development Configuration

​
Running in Development Mode

The package.json defines these development scripts: 
# Start development server (port 5180)
npm run dev

# Start Electron in development mode
npm run electron:dev

# Start both concurrently
npm start
The npm start command automatically starts Vite dev server on port 5180, waits for it to be ready, then launches Electron.

​
Build Configuration

# Build for production
npm run build

# Create distributable
npm run dist
Build outputs:
  • macOS: DMG installer for x64 and arm64
  • Windows: NSIS installer and portable executable
  • Linux: AppImage and DEB package

​
Port Configuration

Cluely requires port 5180 for the Vite development server. If another application is using this port, the app will fail to start.
To check and free port 5180: 
# Find processes using port 5180
lsof -i :5180

# Kill the process (replace [PID] with actual process ID)
kill [PID]

​
Advanced Configuration

​
Switching AI Providers at Runtime

Cluely supports switching between AI providers without restarting: 
// Switch to Ollama
await llmHelper.switchToOllama('llama3.2', 'http://localhost:11434')

// Switch to Gemini
await llmHelper.switchToGemini('your-api-key', 'models/gemini-2.5-flash')

// Switch to OpenRouter
await llmHelper.switchToOpenRouter('your-api-key', 'google/gemini-2.5-flash')

// Switch to K2 Think
await llmHelper.switchToK2Think('your-api-key')

​
Connection Testing

Test your AI provider connection: 
const result = await llmHelper.testConnection()
if (result.success) {
  console.log('Connected successfully')
} else {
  console.error('Connection failed:', result.error)
}

​
Installation Workarounds

​
Sharp Build Errors

If you encounter Python or Sharp build errors during npm install: 
# Use prebuilt binaries (recommended)
rm -rf node_modules package-lock.json
SHARP_IGNORE_GLOBAL_LIBVIPS=1 npm install --ignore-scripts
npm rebuild sharp
The postinstall script automatically handles this: 
"postinstall": "cross-env SHARP_IGNORE_GLOBAL_LIBVIPS=1 npm rebuild sharp"
Sharp is an image processing library that requires native binaries. The SHARP_IGNORE_GLOBAL_LIBVIPS=1 flag tells Sharp to use prebuilt binaries instead of compiling from source, avoiding the need for Python and build tools.

Build docs developers (and LLMs) love

Get started for freeTalk to us