🤖 A.D.A - Pocket AI

A.D.A (Advanced Digital Assistant) is a fully local, privacy-focused AI assistant for Windows. It combines a beautiful modern GUI with powerful voice control capabilities—all running entirely on YOUR computer with no cloud dependency.

🔒 Your data stays on your machine. No API keys required for core functionality. No subscriptions. No data collection.

✨ Key Features

Feature	Description
🎤 Voice Control	Wake word detection ("Jarvis") with natural language commands
💬 AI Chat	Interactive chat with local LLMs via Ollama with streaming responses
🏠 Smart Home	Control TP-Link Kasa smart lights and plugs from the app
📅 Planner	Manage calendar events, alarms, and timers
📰 Daily Briefing	AI-curated news from Technology, Science, and Top Stories
🌤️ Weather	Current weather and hourly forecast on your dashboard
🔍 Web Search	Search the web through voice or chat commands
🖥️ System Monitor	Real-time CPU and memory usage in the title bar

📸 Screenshots

The application features a sleek Windows 11 Fluent Design aesthetic with dark mode support.

📋 Prerequisites

Before you begin, make sure you have:

Required Software

Software	Purpose	Download
Miniconda	Python environment manager	miniconda.io
Ollama	Local AI model server	ollama.com
NVIDIA GPU (Recommended)	Faster AI inference	GPU with 4GB+ VRAM

Hardware Recommendations

Minimum: 8GB RAM, any modern CPU
Recommended: 16GB RAM, NVIDIA GPU with 6GB+ VRAM
Storage: ~5GB for models and voice data

🚀 Quick Start Guide

Follow these steps to get A.D.A running on your system.

Step 1: Install Miniconda

Download from miniconda.io
Run the installer (use default options)
Open Anaconda Prompt (Windows) or your terminal (macOS/Linux)

Step 2: Install Ollama

Download and install from ollama.com/download
Run the installer (Ollama will start automatically as a background service)

✅ Ollama runs in the background - no need to start it manually after installation.

Step 3: Download an AI Model

Open a terminal and pull your preferred model. You can choose from:

🔹 Option A: Qwen3 (Recommended for most users)

# Fast and efficient - great balance of speed and quality
ollama pull qwen3:1.7b

🔹 Option B: DeepSeek R1 (Better reasoning)

# Stronger reasoning capabilities - slightly slower
ollama pull deepseek-r1:1.5b

💡 Tip: You can switch models anytime in config.py by changing RESPONDER_MODEL.

Verify your model is installed:

ollama list

Step 4: Clone & Set Up the Project

# Clone the repository
git clone https://github.com/your-username/pocket_ai.git
cd pocket_ai

# Create a conda environment
conda create -n ada python=3.11 -y

# Activate the environment
conda activate ada

# Install dependencies
pip install -r requirements.txt

⏱️ Note: First installation may take 5-10 minutes as PyTorch and other large packages are downloaded.

Step 5: GPU Setup (NVIDIA Users)

For significantly faster AI inference, install PyTorch with CUDA support:

# Install PyTorch with CUDA 12.4 support
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124

Verify CUDA is working:

python -c "import torch; print(f'CUDA available: {torch.cuda.is_available()}')"

💡 CPU-only users: Skip this step—PyTorch will use CPU by default. It's slower but works fine.

Step 6: Run the Application

python main.py

🎉 That's it! A.D.A will launch with a beautiful GUI.

🎮 GPU Acceleration

A.D.A benefits greatly from GPU acceleration. Here's what runs on your GPU:

Component	GPU Benefit	Without GPU
Router Model	~50ms inference	~200ms inference
Ollama LLM	Fast streaming responses	Slower, but functional
Whisper STT	Real-time transcription	Slight delay

CUDA Requirements

NVIDIA GPU with CUDA Compute Capability 5.0+ (GTX 900 series or newer)
CUDA Toolkit: Bundled with PyTorch—no separate install needed
VRAM: 4GB minimum, 6GB+ recommended

Check Your GPU

# View GPU info and VRAM
nvidia-smi

🤖 Automatic Model Downloads

The following models are downloaded automatically on first run—no manual setup required:

Model	Purpose	Size	Downloaded From
Router Model	Intent classification	~500MB	Hugging Face
TTS Voice	Text-to-speech	~50MB	Piper Voices
STT Model	Speech-to-text (Whisper)	~150MB	OpenAI Whisper

📦 First launch will take a few minutes while these models download. Subsequent launches are instant.

🎙️ Voice Assistant Setup

A.D.A includes Alexa-like voice control with wake word detection.

How It Works

Say "Jarvis" to wake the assistant
Speak your command naturally
A.D.A processes your request and responds

Example Voice Commands

Command	What It Does
"Jarvis, turn on the office lights"	Controls smart lights
"Jarvis, set a timer for 10 minutes"	Creates a countdown timer
"Jarvis, what's on my schedule today?"	Reads your calendar
"Jarvis, search the web for Python tutorials"	Performs web search
"Jarvis, add buy groceries to my to-do list"	Creates a task

Voice Configuration

Edit config.py to customize:

# Change wake word (default: "jarvis")
WAKE_WORD = "jarvis"

# Adjust sensitivity (0.0-1.0, lower = less false positives)
WAKE_WORD_SENSITIVITY = 0.4

# Enable/disable voice assistant
VOICE_ASSISTANT_ENABLED = True

⚙️ Configuration

All configuration is centralized in config.py:

AI Models

Change the chat model in config.py:

# The main chat model (runs on Ollama)
# Options: "qwen3:1.7b" (fast) or "deepseek-r1:1.5b" (better reasoning)
RESPONDER_MODEL = "qwen3:1.7b"

# Ollama server URL (usually no need to change)
OLLAMA_URL = "http://localhost:11434/api"

Model Comparison:

Model	Speed	Reasoning	Best For
`qwen3:1.7b`	⚡ Fast	Good	Daily use, quick responses
`deepseek-r1:1.5b`	Moderate	Excellent	Math, coding, complex questions

Text-to-Speech

# Voice model (downloads automatically on first run)
TTS_VOICE_MODEL = "en_GB-northern_english_male-medium"

Weather Location

The default location is New York City. To change it:

Open the app
Go to Settings tab
Enter your latitude and longitude

🏗️ Project Architecture

pocket_ai/
├── main.py                 # Application entry point
├── config.py               # Centralized configuration
├── requirements.txt        # Python dependencies
│
├── core/                   # Backend logic
│   ├── router.py           # FunctionGemma intent classifier
│   ├── function_executor.py # Executes routed functions
│   ├── voice_assistant.py  # Voice pipeline orchestrator
│   ├── stt.py              # Speech-to-text with wake word
│   ├── tts.py              # Piper text-to-speech
│   ├── kasa_control.py     # Smart home device control
│   ├── weather.py          # Open-Meteo weather API
│   ├── news.py             # DuckDuckGo news + AI curation
│   ├── tasks.py            # To-do list management
│   ├── calendar_manager.py # Local calendar/events
│   ├── history.py          # SQLite chat history
│   └── llm.py              # Ollama LLM interface
│
├── gui/                    # PySide6 GUI
│   ├── app.py              # Main window setup
│   ├── handlers.py         # Chat message handling
│   ├── tabs/               # Individual tab screens
│   │   ├── dashboard.py    # Weather + status overview
│   │   ├── chat.py         # AI chat interface
│   │   ├── planner.py      # Calendar + tasks
│   │   ├── briefing.py     # AI news curation
│   │   ├── home_automation.py  # Smart device control
│   │   └── settings.py     # App configuration
│   └── components/         # Reusable UI widgets
│
├── merged_model/           # Fine-tuned FunctionGemma router
└── demo.py                 # Standalone voice assistant demo

How It Works

┌─────────────┐     ┌──────────────────┐     ┌─────────────┐
│  User Input │────▶│ FunctionGemma   │────▶│  Function   │
│  (Voice/Text)  │     │   Router         │     │  Executor   │
└─────────────┘     └──────────────────┘     └──────┬──────┘
                                                    │
       ┌────────────────────────────────────────────┼────────────────┐
       │                                            │                │
       ▼                                            ▼                ▼
┌──────────────┐                          ┌──────────────┐   ┌──────────────┐
│  Kasa Lights │                          │   Calendar   │   │  Web Search  │
└──────────────┘                          └──────────────┘   └──────────────┘
                                                    │
                                                    ▼
                                          ┌──────────────┐
                                          │ Qwen LLM     │
                                          │ (via Ollama) │
                                          └──────────────┘
                                                    │
                                                    ▼
                                          ┌──────────────┐
                                          │ Piper TTS    │
                                          │ (Voice Out)  │
                                          └──────────────┘

User speaks or types a command
FunctionGemma Router (fine-tuned local AI) classifies intent
Function Executor runs the appropriate action
Qwen LLM generates a natural language response
Piper TTS speaks the response (if voice enabled)

🏠 Smart Home Integration

A.D.A supports TP-Link Kasa smart devices:

Supported Devices

✅ Smart bulbs (on/off, brightness, color)
✅ Smart plugs (on/off)
✅ Smart light strips

Setup

Ensure your Kasa devices are on the same network as your computer
Open A.D.A and go to the Home Automation tab
Click Refresh to scan for devices
Control devices through the GUI or voice commands

Voice Commands

"Turn on the living room lights"
"Set the bedroom lights to 50%"
"Turn off all lights"
"Change the office light to blue"

🔧 Troubleshooting

Common Issues

❌ Ollama connection refused

Problem: The app can't connect to Ollama.

Solution:

Make sure Ollama is running: ollama serve
Check if the model is downloaded: ollama list
Verify the URL in config.py matches your setup

❌ CUDA/GPU not detected

Problem: PyTorch is running on CPU instead of GPU.

Solution:

Install CUDA-compatible PyTorch:

pip install torch --index-url https://download.pytorch.org/whl/cu121

Verify CUDA: python -c "import torch; print(torch.cuda.is_available())"

❌ Voice assistant not working

Problem: Wake word isn't being detected.

Solution:

Check your microphone permissions
Ensure realtimestt is installed: pip install realtimestt
Try lowering WAKE_WORD_SENSITIVITY in config.py

❌ Smart devices not found

Problem: Kasa devices don't appear in the app.

Solution:

Ensure devices are on the same WiFi network
Try the Kasa app first to verify devices work
Check firewall isn't blocking device discovery (UDP port 9999)

🤝 Contributing

Contributions are welcome! Here's how to get started:

Fork the repository
Create a feature branch: git checkout -b feature/amazing-feature
Make your changes
Run tests: pytest tests/
Submit a pull request

📜 License

This project is open source. See LICENSE for details.

🙏 Acknowledgments

Ollama - Local LLM inference
QFluentWidgets - Beautiful UI components
Piper TTS - Lightweight text-to-speech
python-kasa - Kasa device control
RealTimeSTT - Speech recognition

Made with ❤️ for local AI enthusiasts

Name		Name	Last commit message	Last commit date
Latest commit History 56 Commits
core		core
data		data
gui		gui
tests		tests
.DS_Store		.DS_Store
.gitignore		.gitignore
README.md		README.md
chat_history.db		chat_history.db
check_icons.py		check_icons.py
config.py		config.py
debug_router.py		debug_router.py
demo.py		demo.py
generate_training_data.py		generate_training_data.py
icons.txt		icons.txt
import_error.log		import_error.log
import_error.txt		import_error.txt
install_log.txt		install_log.txt
main.py		main.py
requirements.txt		requirements.txt
speed_test.py		speed_test.py
test_date_parsing.py		test_date_parsing.py
test_drawer.py		test_drawer.py
test_settings.py		test_settings.py
train_function_gemma.py		train_function_gemma.py
training_dataset.jsonl		training_dataset.jsonl
training_dataset_functions.jsonl		training_dataset_functions.jsonl
upload_model.py		upload_model.py
verify_unload.py		verify_unload.py

Folders and files

Latest commit

History

Repository files navigation

🤖 A.D.A - Pocket AI

✨ Key Features

📸 Screenshots

📋 Prerequisites

Required Software

Hardware Recommendations

🚀 Quick Start Guide

Step 1: Install Miniconda

Step 2: Install Ollama

Step 3: Download an AI Model

Step 4: Clone & Set Up the Project

Step 5: GPU Setup (NVIDIA Users)

Step 6: Run the Application

🎮 GPU Acceleration

CUDA Requirements

Check Your GPU

🤖 Automatic Model Downloads

🎙️ Voice Assistant Setup

How It Works

Example Voice Commands

Voice Configuration

⚙️ Configuration

AI Models

Text-to-Speech

Weather Location

🏗️ Project Architecture

How It Works

🏠 Smart Home Integration

Supported Devices

Setup

Voice Commands

🔧 Troubleshooting

Common Issues

🤝 Contributing

📜 License

🙏 Acknowledgments

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages