AI Agent Desktop

A clean, modular monolith desktop AI assistant with voice input, chat interface, and powerful tools.

🚀 Quick Start

1. Install dependencies:

   setup.bat        # Windows
   pip install -r requirements.txt  # Or manually

2. Run the agent:

   run.bat          # Windows
   python run.py    # Or directly

Architecture

This is a modular monolith - a single-process application with clean module boundaries:

src/
├── app.py              # Main application entry point
├── config/             # Configuration management
│   ├── settings.py     # YAML-based settings with Pydantic
│   ├── agent_config.py # Agent API parameters
│   └── prompts.py      # System prompts
├── core/               # Core agent logic
│   └── agent.py        # AI agent with streaming & tools
├── storage/            # Data persistence
│   ├── secure.py       # Encrypted storage (keyring)
│   ├── chat_history.py # Conversation persistence
│   └── memory.py       # User memories
├── tools/              # Agent capabilities
│   ├── memory.py       # Memory management
│   ├── todos.py        # Task management
│   ├── filesystem.py   # File operations
│   ├── terminal.py     # Command execution
│   └── ...             # More tools
├── services/           # In-process services
│   ├── transcribe.py   # Voice-to-text
│   └── tts.py          # Text-to-speech
└── ui/                 # PyQt6 interface
    ├── widget.py       # Floating widget
    └── components/     # Reusable UI parts
        ├── chat_window.py       # Chat interface
        ├── settings_window.py   # Settings dialog
        ├── multiline_input.py   # Text input widget
        ├── screenshot_selector.py # Screenshot tool
        └── chat_history_json_window.py # History viewer

Features

🎤 Voice Input

• Long-press (1s) to record
• Auto-transcribe using OpenAI Whisper
• Multi-language support (en, ro, ru, de, fr, es)

💬 Chat Interface

• Type or speak your messages
• Real-time streaming responses
• File drag-and-drop attachment
• Screenshot sharing (up to 5)
• Syntax-highlighted code blocks with Pygments
• Encrypted, persistent history
• Token usage tracking
• Stop generation at any time
• Copy, edit, or delete user messages

🛠️ Agent Tools

• Memory: Remember user preferences
• Todos: Task management
• Files: Read, write, search, edit
• Terminal: Run commands
• Documents: Create Word files
• Charts: Generate visualizations
• Web: Search and browse
• Images: AI image generation

Configuration

Quick Start

1. Set your API key:

   • Right-click widget → Settings
   • Enter API token and base URL
   • Saved securely in OS keyring

2. Customize: Edit config.yaml and prompts/system_prompt.md

Settings Structure

# config.yaml
agent_name: Djasha

api:
  base_url: 'https://api.openai.com/v1'

agent:
  model_name: gpt-5.1
  system_prompt_path: prompts/system_prompt.md

ui:
  theme: dark

tools:
  enabled_tools: [memory, todos, filesystem, ...]

Development

Project Structure

• src/ - All application code
• config.yaml - User configuration
• requirements.txt - Python dependencies

Adding New Tools

1. Create a new file in src/tools/
2. Define a class with schema property and run() method
3. Register in src/tools/__init__.py
4. Add to get_default_tools() function

Customizing the UI

• Components are in src/ui/components/
• Main widget in src/ui/widget.py
• Styles are inline using PyQt6 stylesheets

Storage and Security

• Chat history and memories are stored encrypted at:
  • %APPDATA%/ai-agent-desktop/chat_history.enc
  • %APPDATA%/ai-agent-desktop/memories.enc
• Encryption key (data_key) is stored in Windows Credential Manager:
  • Service: ai-agent-desktop/data_key, Username: data_key
• API Token saved from Settings is stored under:
  • Service: ai-agent-desktop/api_token, Username: api_token
• See secure_storage/README.md for details.

Contributing

Contributions are welcome! Feel free to:

• Report bugs or issues
• Suggest new features
• Submit pull requests
• Improve documentation

License

This project is licensed under the MIT License - see the LICENSE file for details.

Author

destorted93

• GitHub: @destorted93
• Repository: ai-agent-desktop