CogniTriage ⚡

Moving from passive analytics to active intervention.

CogniTriage is an AI-powered cognitive triage system that analyzes student learning patterns in real-time and generates personalized micro-interventions to unblock learning friction.

Quick links

• Repository: https://github.com/roboticspro1/DeepLearningWeek-ByteMe
• Run (local): streamlit run app.py

Quick note on API keys

• You can paste your OpenAI API key in the Streamlit sidebar at runtime.
• Or set the environment variable OPENAI_API_KEY before running:

export OPENAI_API_KEY="sk-..."
streamlit run app.py

Streamlit also supports a .streamlit/secrets.toml file with:

[openai]
api_key = "sk-..."

Overview

Instead of overwhelming dashboards full of metrics, CogniTriage shows only what matters: What is breaking down, why, and exactly how to fix it.

Key Features

• Priority Action Center: Displays only urgent topics where students are struggling (urgency > 30)
• Contradiction-Free Diagnostics: Intelligent classification of learning issues:
  • Active Decay (Forgetting) - Used to know it, but recently failed
  • Careless Errors/Fatigue - Fast but inaccurate responses
  • Deep Conceptual Gap - Slow and inaccurate
  • High Cognitive Load - Accurate but too slow
  • Retained Mastery - Performing well
• Cognitive Profile Radar: Visual summary of mastery across top 5 topics
• AI-Generated Micro-Tasks: 5-minute targeted interventions powered by GPT-4o-mini
• Root Cause Telemetry: Deep analytics showing trends, decay detection, and performance metrics

Project Structure

.
├── app.py              # Streamlit UI and main application logic
├── utils.py            # Core analysis functions and AI integration
├── data.csv            # Student performance data (timestamp, topic, correct, time_taken)
├── requirements.txt    # Python dependencies
├── lib/                # Frontend libraries
│   ├── tom-select/     # Dropdown UI component
│   ├── vis-9.1.2/      # Network visualization library
│   └── bindings/       # Custom JS bindings
└── graph.html          # Generated network visualization

Getting Started

Prerequisites

• Python 3.8+
• OpenAI API key (for AI-generated micro-tasks)

Installation

1. Clone the repository:

   git clone https://github.com/roboticspro1/DeepLearningWeek-ByteMe.git
   cd DeepLearningWeek-ByteMe

2. Install dependencies:

   pip install -r requirements.txt

3. Run the application:

   streamlit run app.py

4. Open your browser to http://localhost:8501

Usage

1. Enter OpenAI API Key: Paste your API key in the sidebar under "OpenAI API Key (For Unblocker)"
2. View Your Triage Queue: The Priority Action Center displays topics requiring intervention, ordered by urgency
3. Generate Micro-Tasks: Click "⚡ Generate 5-Min Micro-Task" for AI-generated, targeted learning activities
4. Analyze Cognitive Profile: View your mastery radar and deep telemetry in the tabs below

Data Format

The data.csv file should contain the following columns:

Column	Description
timestamp	When the activity occurred (ISO format)
topic	Topic being studied
correct	1 if answer was correct, 0 if incorrect
time_taken	Time taken to answer (seconds)

Example:

timestamp,topic,correct,time_taken
2024-01-15T10:30:00,Calculus,1,45
2024-01-15T10:35:00,Calculus,0,120
2024-01-15T11:00:00,Linear Algebra,1,30

Diagnostic Logic

The system uses contradiction-free logic to diagnose learning issues:

1. Active Decay: past_accuracy - recent_accuracy > 20% → Urgency: 95 🔴
2. Careless/Fatigue: accuracy < 50% AND time < 40s → Urgency: 80 🟠
3. Deep Gap: accuracy < 50% AND time ≥ 40s → Urgency: 90 🔴
4. High Load: accuracy ≥ 50% AND time > 120s → Urgency: 60 🔵
5. Retained: Everything else → Urgency: 10 🟢

Architecture

• Frontend: Streamlit (Python-based reactive UI)
• Analytics Engine: Pandas for data processing and trend detection
• AI Backend: OpenAI GPT-4o-mini for generating micro-tasks
• Visualization: Plotly (radar charts, graphs)

Technologies Used

• streamlit - Interactive web UI
• pandas & numpy - Data analysis
• openai - AI-powered interventions
• plotly - Data visualization

Contributing

This is a ByteMe hackathon project. Contributions welcome!

License

MIT

---

Built with ❤️ at Deep Learning NTU Week