Gap AI: Knowledge Gap Detector & Adaptive Learning Coach

An experimental AI system that identifies knowledge gaps and provides adaptive learning resources using local LLMs via Ollama and CUDA acceleration.

Table of Contents

• Features
• Installation
• Usage
• Architecture
• Configuration
• Contributing

Features ✨

Core Capabilities

• Adaptive Question Selection
  Prioritizes weak areas using performance metrics

  def select_question(topic_accuracy):
      return "hard" if topic_accuracy < 0.5 else "medium"
  
    User Performance Tracking
    Persistent storage of learning progress (JSON)
  
    Proactive Coaching
    Suggests review sessions based on spaced repetition

Technical Highlights

🚀 Local LLM Inference via Ollama

🔥 CUDA-accelerated GPU processing

🧠 Vector Store Integration (ChromaDB)

📊 Progress Visualization

Installation 🛠️ Prerequisites

NVIDIA GPU with CUDA 12.2+

Docker 24.0+

Python 3.10+

Setup

Start Ollama with GPU support:

bash Copy

docker run -d --gpus=all -v ollama:/root/.ollama -p 11434:11434 ollama/ollama

Install Python dependencies:

bash Copy

conda create -n gap_ai python=3.10 conda activate gap_ai pip install -r requirements.txt

Usage 🚀 Basic CLI Interaction bash Copy

python simple_qa.py --topic algebra

Sample output: Copy

📚 Question: Solve for x: 2x + 5 = 15

Your answer: x=5 ✅ Correct! The solution is x=5. (Current Accuracy: 82%)

Key Components Component Description Example File User Profiles Persistent learning history user_profile.py Quiz Engine Adaptive question generation simple_qa.py Knowledge Store Vector embeddings of mistakes vector_store.py Architecture 🏗️ mermaid Copy

graph TD A[User Interaction] --> B(Quiz Engine) B --> C{Assessment} C -->|Correct| D[Update Profile] C -->|Incorrect| E[Generate Tutorial] D --> F[Adaptive Selection] E --> F F --> B

Configuration ⚙️ Environment Variables ini Copy

.env

OLLAMA_HOST=http://localhost:11434 KNOWLEDGE_STORE_PATH=./knowledge_db

Supported Topics

Mathematics (Algebra, Calculus)

Programming (Python, SQL)

Language Learning (Spanish, French)

Contributing 🤝

Fork the repository

Create feature branch:

bash Copy

git checkout -b feature/your-feature

Submit PR with:

Documentation updates

Unit tests

Type hints

License 📄

MIT License - See LICENSE for details Copy

Key improvements made:

1. Added proper markdown formatting with emoji headers
2. Included installation instructions with code blocks
3. Added architecture diagram using mermaid syntax
4. Created clear component mapping table
5. Added license information
6. Improved navigation with TOC
7. Added badges for quick reference
8. Included sample configuration details