Enso - AI-Powered Learning Assistant

Enso is a modern PyQt6 application that implements an Local LLM-powered learning companion using the Feynman Technique. The application processes documents using natural language processing and engages users in a conversational learning experience guided by an AI model.

Features

• 🎨 Modern gradient UI with smooth transitions and ChatGPT-like interface
• 📚 Support for PDF, DOCX, and TXT documents
• 📑 Page range selection for focused learning
• 🤖 Local model integration for intelligent responses
• 💡 RAG-powered context retrieval using FAISS
• 🎯 Socratic teaching method with progressive hints
• 💬 Rich text formatting with bold, italic, and bullet points
• ⚡ Real-time token streaming for dynamic responses

Setup Instructions For Electron

1. Clone the repository:

git clone https://github.com/hazikchaudhry/ENSO.git
cd ENSO

2. Install dependencies:

cd electron-app
npm install

3. Install Python dependencies:

cd backend
pip install -r requirements.txt

4. Run the application:

# Terminal 1 - Start backend
cd backend
python server.py

# Terminal 2 - Start Electron app
cd electron-app
npm start

Setup For AI

1. Install dependencies:

pip install -r requirements.txt

2. Install Ollama on Windows at https://ollama.com/ and download the Mistral model OR Gemma model:

# Pull required model
ollama pull mistral:7b-instruct
ollama run gemma3:4b

3. Run the application:

python main.py

Usage

1. Launch the application
2. Click "Upload Document" to select your learning material (PDF/DOCX/TXT)
3. Select the page range you want to focus on
4. Verify Ollama is running locally
5. Start your learning journey with the AI companion

Technical Architecture

The application leverages several key technologies:

• Document Processing: PyMuPDF (fitz) and python-docx for text extraction
• AI/ML Components:
  • LangChain for orchestrating AI workflows
  • Ollama for local LLM inference
  • HuggingFace embeddings for text vectorization
  • FAISS for efficient vector similarity search
• UI Framework: PyQt6 with modern gradient styling

Learning Approach

Enso implements the Feynman Technique through:

1. Engaging users in natural dialogue about concepts
2. Identifying knowledge gaps through Socratic questioning
3. Providing progressive hints rather than direct answers
4. Using analogies and real-world examples
5. Encouraging users to explain concepts in their own words

Features in Detail

• Smart Context Retrieval: Uses FAISS similarity search to find relevant context
• Real-time Responses: Streams tokens for dynamic response generation
• Rich Text Support: Format your messages with bold, italic, and bullet points
• Conversation Management: Tracks conversation state and maintains context
• Error Handling: Robust error management with user-friendly messages