AudioMind

Note: This started as a personal/toy project I created to ask questions to podcasts I listen to. I've made it generic so others can use it with their own audio content. Feel free to adapt it for your needs!

An MP3 to AI Chat Assistant - A configurable AI chat assistant that can be customized for your content and use case. Transform audio content into interactive, searchable conversations.

⚠️ Important Security Notice

This is a demonstration project and should not be deployed to production without implementing additional security measures:

• The application currently has no user authentication
• Lack of authentication could lead to:
  • Unauthorized access to your API endpoints
  • Potential abuse of your API rate limits
  • Excessive costs from unrestricted API usage
  • Malicious use of your services
• Consider implementing proper authentication, rate limiting, and usage monitoring before any production deployment

Technologies Used

• Frontend Framework: SvelteKit 2.x with Svelte 5
• UI/Styling:
  • TailwindCSS for styling
  • DaisyUI for UI components
  • Typography plugin for content formatting
• AI/Machine Learning:
  • Anthropic Claude 3 for natural language processing
  • Deepgram for audio transcription
  • Voyage for vector database and similarity search
• Database: Turso (LibSQL) for data storage
• Development Tools:
  • TypeScript for type safety
  • Vite for development and building
  • ESLint and Prettier for code quality
  • Vitest and Playwright for testing
• Performance:
  • LRU Cache for response caching
  • Server-sent events for real-time progress updates

Setup

1. Clone the repository
2. Copy .env.example to .env:

cp .env.example .env

3. Fill in your environment variables in .env
4. Install dependencies:

npm install

5. Start the development server:

npm run dev

Configuration

Environment Variables

Required environment variables in your .env file:

# Required
ANTHROPIC_API_KEY=your_api_key_here
TURSO_URL=your_turso_url_here
TURSO_AUTH_TOKEN=your_turso_auth_token_here
DEEPGRAM_API_KEY=your_deepgram_key_here
VOYAGE_API_KEY=your_voyage_key_here

Application Configuration

The application can be customized by modifying src/lib/config/app-config.ts. This file contains non-sensitive configuration options:

{
    // Application name and description
    app_name: 'My Custom Assistant',
    app_description: 'Chat with your content using AI',

    // AI Configuration
    ai: {
        // Available models:
        // - claude-3-opus-20240229 (most capable)
        // - claude-3-sonnet-20240229 (balanced)
        // - claude-3-haiku-20240307 (fastest)
        model: 'claude-3-haiku-20240307',

        max_tokens: 1024,

        // Customize the AI's behaviour
        system_prompt: 'Your custom system prompt...',

        // Define different response styles
        style_instructions: {
            normal: {
                description: 'Balanced and clear responses',
                instruction: 'Your custom instruction...'
            },
            // Add more styles as needed
        }
    },

    // Set default response style
    default_response_style: 'normal'
}

Response Styles

The application supports multiple response styles that can be configured:

• normal: Balanced and clear responses
• concise: Brief and direct responses
• explanatory: Detailed explanations
• formal: Professional and structured responses

You can modify existing styles or add new ones in the configuration file.

Features

• Audio file upload and processing
• Real-time transcription progress updates
• Vector-based semantic search
• Configurable AI response styles
• Interactive chat interface
• Server-side streaming responses
• Response caching for performance

Security Notes

• Never commit your .env file
• Keep all sensitive information (API keys, credentials) in environment variables
• The app-config.ts file should only contain non-sensitive application settings
• Implement proper authentication before deploying to production
• Consider adding rate limiting to protect your API usage
• Monitor API usage to prevent excessive costs

Development

To start the development server:

npm run dev

Visit http://localhost:5173 to see your application.

License

MIT