🍂 Autumn AI Assistant

A sophisticated AI assistant with advanced features including streaming responses, G2P-enhanced text-to-speech, and emotional intelligence.

✨ Latest Features (June 2025)

🚀 NEW: Streaming Responses

• Real-time token generation like ChatGPT/Claude
• 2.5x faster perceived speed with immediate response start
• Sentence-level TTS integration for natural conversation flow
• Comprehensive error handling with fallback to non-streaming

🎵 Enhanced TTS with G2P

• Misaki G2P engine for improved pronunciation
• Custom pronunciation dictionary for technical terms
• Multi-engine TTS support (Kokoro, Edge-TTS, pyttsx3 fallbacks)
• Emotion-aware voice selection with different speaking styles

🔧 Technical Improvements

• Unicode-safe logging - all emoji/special character issues resolved
• Optimized AI model: Now using gemma3:1b-it-q4_K_M for better performance
• Advanced error monitoring with automatic recovery
• Memory optimizations and cleanup management

Features

🎭 Sophisticated Personality

• Dual Modes: Serious (business) and Free (casual) personality modes
• Intelligent Switching: Automatically detects context and switches modes
• Sarcastic & Witty: Light humor and contextual sarcasm in casual mode
• Professional: Efficient and formal communication in serious mode

🧠 Hybrid Intelligence

• Local AI: Ollama + Phi-3 Mini for privacy and speed
• Cloud Fallback: Google Gemini API for complex reasoning
• Smart Routing: Automatically chooses best AI service for each task

💾 Incredible Memory

• Conversation History: Remembers your chats and context
• User Preferences: Learns and stores your preferences
• Personal Facts: Remembers important information about you
• Smart Retention: Automatic cleanup of old, unimportant memories

🎤 Voice Interaction

• Speech-to-Text: OpenAI Whisper (local processing)
• Text-to-Speech: pyttsx3 with personality-matched voice tones
• Wake Words: "Hey Autumn" voice activation
• Mode-Aware: Voice changes between serious and casual modes

🖥️ Beautiful GUI

• Draggable Widget: Always-on-top, movable interface
• Autumn Ball: Compact autumn leaf icon when minimized
• Expandable Chat: Full chat interface when expanded
• Modern Design: Autumn-themed colors and styling

🔗 Smart Integrations

• Calendar: Schedule management and conflict detection
• Web Search: Controlled access to weather, news, and information
• Task Management: Proactive reminders and follow-ups

Hardware Requirements

Optimized for lightweight systems:

• RAM: 4GB minimum, 8GB+ recommended
• Storage: 5GB for models and dependencies
• GPU: Optional (RTX 3050 4GB works perfectly)
• CPU: Any modern processor (tested on Ryzen 5 5500H)

Resource Usage:

• Memory: <200MB typical usage
• CPU: <50% on dual-core systems
• Storage: ~3GB for AI models

🚀 Quick Start

Prerequisites

• Python 3.8+
• Ollama installed and running
• GPU recommended (CUDA/ROCm) for best performance

Installation

1. Clone the repository

   git clone https://github.com/shhreyuuFTW/Autumn_AI_Assistant.git
   cd Autumn_AI_Assistant

2. Install dependencies

   pip install -r requirements.txt

3. Set up Ollama model

   ollama pull gemma3:1b-it-q4_K_M

4. Configure API keys (optional)

   # Copy example config and edit with your API keys
   cp config/settings.py.example config/settings.py
   # Edit config/settings.py with your Gemini API key if using cloud features

5. Run Autumn

   python app.py

🌊 Streaming Demo

Experience the real-time streaming responses:

python test_streaming_simple.py

This demonstrates:

• Token-by-token generation (like ChatGPT)
• Sentence detection for TTS integration
• Performance comparison with traditional non-streaming

🎵 G2P Enhanced TTS Demo

Test the improved pronunciation:

python demo_g2p_fixed.py

Compare traditional vs G2P-enhanced pronunciation with difficult words.

Project Structure

Autumn_AI_Assistant/
├── app.py                 # Main application entry point
├── requirements.txt       # Python dependencies
├── setup.py              # Automated setup script
├── .env                  # Environment configuration
├── config/
│   ├── personality.py    # Autumn's personality system
│   └── settings.py       # Application settings
├── core/
│   ├── brain.py         # AI intelligence and routing
│   ├── memory.py        # Memory and storage system
│   └── voice.py         # Voice input/output
├── services/
│   ├── gemini.py        # Gemini API integration
│   ├── web_search.py    # Web information access
│   └── calendar.py      # Calendar integration
├── gui/
│   └── widget.py        # PyQt6 GUI interface
└── data/
    └── autumn_memory.db # SQLite memory database

Configuration

Environment Variables

# Required
GEMINI_API_KEY=your_key_here

# Optional Performance
MEMORY_LIMIT_MB=200
CPU_LIMIT_PERCENT=50

# Optional Features
AUTO_START_VOICE=true
DEBUG=false
LOG_LEVEL=INFO

Voice Settings

• STT Model: Whisper Tiny (39MB, fast)
• TTS Engine: pyttsx3 (local, customizable)
• Wake Words: "autumn", "hey autumn"

AI Models

• Primary: Ollama Phi-3 Mini (2.3GB, local)
• Fallback: Google Gemini Flash (cloud)

Usage Examples

Text Interaction

You: "Schedule a meeting with John tomorrow at 2 PM"
Autumn: "I'll schedule that meeting for you! Let me check for any conflicts... Done! Meeting with John set for tomorrow at 2 PM. Anything else you need, sir? 😊"

Voice Interaction

You: "Hey Autumn"
Autumn: "Yes sir? How can I help you today?"
You: "What's my schedule like?"
Autumn: "Let me check your calendar... You have three meetings today: 9 AM with the team, 1 PM lunch with Sarah, and 4 PM project review. Looks like a busy but manageable day!"

Mode Switching

You: "Urgent: I need the quarterly report ASAP"
Autumn: [Serious Mode] "Understood. I'll locate the quarterly report immediately and have it ready for you. Processing now."

You: "Thanks, you're awesome!"
Autumn: [Free Mode] "Aww, thank you! Just doing my job... though I do it with exceptional style! 😏"

Technical Architecture

Local-First Design

• Privacy: Voice and personal data stay on your machine
• Speed: Local processing for instant responses
• Reliability: Works offline for basic interactions
• Efficiency: Optimized for minimal resource usage

Intelligent Routing

User Input → Mode Detection → Task Classification
                ↓
        ┌─── Simple Chat ─── Local AI (Phi-3)
        ├─── Complex Task ─── Cloud AI (Gemini)
        ├─── Web Info ─────── Web APIs
        └─── Calendar ─────── Calendar Service
                ↓
        Response + Personality → Voice/Text Output

Memory System

• SQLite Database: Persistent storage
• Smart Indexing: Fast retrieval of relevant memories
• Automatic Cleanup: Configurable retention policies
• Context Awareness: Uses memory to inform responses

Development

Adding New Features

1. New Services: Add to services/ directory
2. Personality Traits: Modify config/personality.py
3. Memory Types: Extend core/memory.py
4. GUI Components: Update gui/widget.py

Testing

# Test voice system
python -c "from core.voice import AutumnVoice; import asyncio; asyncio.run(AutumnVoice().initialize())"

# Test AI brain
python -c "from core.brain import AutumnBrain; print('Brain module OK')"

# Test memory
python -c "from core.memory import AutumnMemory; print('Memory module OK')"

Debugging

# Enable debug mode
export DEBUG=true
export LOG_LEVEL=DEBUG

python app.py

Troubleshooting

Common Issues

PyAudio Installation (Windows)

# If PyAudio fails to install
pip install pipwin
pipwin install pyaudio

Ollama Connection

# Check if Ollama is running
ollama list
curl http://localhost:11434/api/tags

Voice Issues

# Test microphone access
python -c "import pyaudio; print('Microphone OK')"

Memory Issues

# Check database
sqlite3 data/autumn_memory.db ".tables"

Performance Optimization

• Use Whisper "tiny" model for fastest STT
• Limit conversation history to 50 turns
• Enable automatic memory cleanup
• Use local AI for simple interactions

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Test thoroughly
5. Submit a pull request

License

MIT License - See LICENSE file for details

Acknowledgments

• OpenAI Whisper for excellent local STT
• Ollama for making local LLMs accessible
• PyQt6 for the beautiful GUI framework
• Google Gemini for powerful cloud AI capabilities

---

Built with ❤️ for productivity and personality

Autumn AI Assistant - Your sophisticated digital companion 🍂