kCompanion: AI-Driven System Companion

Basic installation sudo chmod +x setup.sh sudo ./setup.sh This project implements an AI-driven system companion for Linux with a hybrid architecture:

1. Kernel Module: Collects system metrics efficiently and provides real-time monitoring
2. PyTorch AI Daemon: Uses deep learning with CUDA for advanced anomaly detection

Components

Kernel Module (kcompanion.c)

• Collects system metrics (CPU, memory, I/O, syscalls)
• Provides a character device interface (/dev/kcompanion)
• Performs basic statistical analysis
• Sends data to userspace for advanced AI processing

AI Daemon (kcompanion_ai.py)

• PyTorch-based deep learning with CUDA acceleration
• LSTM model for time-series prediction
• Autoencoder for anomaly detection
• Reinforcement learning for adapting to user feedback

Requirements

• Linux kernel 6.8+
• NVIDIA GPU with CUDA support
• PyTorch with CUDA
• Python 3.8+

Installation

Kernel Module

# Compile and install the kernel module
make
sudo insmod kcompanion.ko

# Verify it's loaded
lsmod | grep kcompanion
ls -l /dev/kcompanion

PyTorch AI Daemon

# Install Python dependencies
pip install -r requirements.txt

# Run the AI daemon
sudo python3 kcompanion_ai.py

Usage

The system works automatically once both components are running. The kernel module collects data and the AI daemon analyzes it.

To read current suggestions:

cat /dev/kcompanion

To provide feedback (helps the AI learn):

# Positive feedback
echo "good:This suggestion was helpful" > /dev/kcompanion

# Negative feedback
echo "bad:This suggestion was not useful" > /dev/kcompanion

Architecture

┌────────────────────┐     ┌─────────────────────────┐
│   Kernel Module    │     │   PyTorch AI Daemon     │
│   (kcompanion.c)   │     │   (kcompanion_ai.py)    │
│                    │     │                         │
│  ┌──────────────┐  │     │  ┌─────────────────┐    │
│  │System Metrics│  │     │  │LSTM Predictor   │    │
│  │Collection    │  │     │  │(Time Series)    │    │
│  └──────┬───────┘  │     │  └────────┬────────┘    │
│         │          │     │           │             │
│  ┌──────┴───────┐  │     │  ┌────────┴────────┐    │
│  │Basic         │  │     │  │Anomaly          │    │
│  │Statistics    │◄─┼─────┼──┤Autoencoder      │    │
│  └──────┬───────┘  │     │  └────────┬────────┘    │
│         │          │     │           │             │
│  ┌──────┴───────┐  │     │  ┌────────┴────────┐    │
│  │Character     │◄─┼────►│  │Suggestion       │    │
│  │Device        │  │     │  │Generator        │    │
│  └──────────────┘  │     │  └─────────────────┘    │
└────────────────────┘     └─────────────────────────┘

Performance Notes

• The AI daemon requires significant GPU resources for CUDA-accelerated deep learning
• The kernel module has minimal performance impact (< 1% CPU overhead)

Security Considerations

• The kernel module runs with kernel privileges - use caution
• The character device has appropriate permission restrictions

License

GPL v2 (required for kernel modules)

How It Works

The AI-driven anomaly detection system works through:

1. Data Collection: The module periodically collects system metrics using kernel APIs
2. Deep Learning Analysis:
   • The userspace AI component uses LSTM networks for time-series prediction
   • Autoencoders detect anomalies through reconstruction error
   • Neural networks classify system states to generate appropriate suggestions
3. Adaptive Learning:
   • The AI system continuously learns from historical data
   • User feedback is incorporated to improve future suggestions

Uninstallation

1. Unload the module and stop the AI daemon:

sudo ./setup.sh stop

2. Completely uninstall:

sudo ./setup.sh uninstall

Security Considerations

• The module is designed to be memory-safe and avoid kernel crashes
• It uses appropriate locking mechanisms (spinlocks, mutexes) to ensure thread safety
• Permission checks are in place to prevent unauthorized access

Limitations

• GPU performance depends on your NVIDIA hardware capabilities
• Some metrics are currently simulated (placeholders) rather than actual system values
• Memory usage is fixed to avoid allocation issues in kernel space

License

This software is licensed under the GPL license, as required for Linux kernel modules.