CogniOS

CogniOS is a lightweight, self-contained system telemetry monitor and workload optimization platform written in Python. It provides high-concurrency performance metrics logging, Isolation Forest anomaly detection, neural network-based workload classification, and process priority adjustments with safety fallbacks for sandboxed environments.

The project is built using Test-Driven Development (TDD) and features a dynamic dark-themed dashboard.

---

📂 Project Structure

/cognios
  ├── /src
  │    ├── __init__.py
  │    ├── database.py       # Thread-safe SQLite connection manager (WAL mode)
  │    ├── telemetry.py      # Real-time telemetry loop & stateful OS simulation
  │    ├── os_doctor.py      # Isolation Forest anomaly detection classifier
  │    ├── focus_os.py       # PyTorch CNN workload classifier & process prioritizer
  │    ├── blackbox.py       # Historical logs database cleanup & CSV exporter
  │    ├── research.py       # FCFS vs SJF waiting time and RL scheduler simulator
  │    └── ui.py             # Streamlit graphical dashboard
  ├── /tests
  │    └── test_all.py       # Pytest suite verifying all core modules
  ├── requirements.txt       # Project dependencies
  └── main.py                # Platform orchestrator (starts daemons and UI)

---

⚡ Core Features

• Non-Blocking Telemetry DB: Uses SQLite configured with Write-Ahead Logging (WAL mode) and a NORMAL synchronous profile. This setup supports concurrent, thread-safe writes from the telemetry collector while Streamlit and the ML diagnostic modules continuously read the data.
• Stateful System Simulator: When real OS metrics are restricted, the telemetry engine runs a state machine mimicking transitions between 'idle' operations and heavy 'spike' workloads (e.g. active code compilation) with realistic random-walk momentum and slow-drifting disk allocation.
• ML Diagnostics (OS Doctor): An unsupervised IsolationForest continuously trains on active telemetry to distinguish between healthy baselines and high-intensity resource spikes.
• Workload Classification & Nice Control (FocusOS):
  • Features a PyTorch Convolutional Neural Network (WorkloadCNN) that maps multi-channel temporal metric windows (channels=3, time_steps=10, features=5) to active execution states.
  • Contains a nice value scheduler with automated psutil.AccessDenied exception handling to safely mock optimizations in sandboxed or non-sudo setups.
• Scheduling Simulator: Evaluates queue waiting times across First Come First Serve (FCFS), mathematically optimal Shortest Job First (SJF), and simulated Reinforcement Learning (RL) scheduling policies.

---

🚀 Getting Started

Prerequisites

• Python 3.10+ (Tested on Python 3.14)
• Virtual environment tools (venv)

Installation & Setup

1. Clone the repository and navigate to the project directory:

   cd CogniOS

2. Initialize and activate a virtual environment:

   python3 -m venv .venv
   source .venv/bin/activate

3. Install dependencies:

   pip install -r cognios/requirements.txt

Running the Platform

To start the unified platform (starts the telemetry daemon, anomaly monitor, and the graphical Streamlit dashboard):

python cognios/main.py

Once running, access the interactive dashboard at: 👉 http://localhost:8501

---

🧪 Testing

The platform includes a comprehensive pytest suite covering telemetry capture, ML diagnostics, neural net tensor passes, SQLite concurrent writes, database pruning, and UI integration.

To run the complete test suite:

pytest cognios/tests/test_all.py -v

---

🛠️ Subsystem Architecture

graph TD
    A[main.py: Platform Orchestrator] --> B[telemetry.py: Telemetry Daemon]
    A --> C[os_doctor.py: ML Anomaly Detector]
    A --> D[ui.py: Streamlit Dashboard Subprocess]
    
    B -->|Batch Insert (psutil/mock)| E[(cognios.db: SQLite WAL)]
    C -->|Train / Read| E
    D -->|Plot / Read| E
    
    subgraph "Workload Optimization"
        F[focus_os.py: WorkloadCNN]
        G[focus_os.py: Process Prioritizer]
    end
    
    subgraph "Logs & Cleanup"
        H[blackbox.py: DB Purge & CSV Exporter]
    end
    
    subgraph "Scheduler Research"
        I[research.py: FCFS vs SJF vs RL Simulator]
    end

Loading