SAGE XR is an AI-powered chatbot that uses Retrieval-Augmented Generation (RAG) to answer questions based on uploaded documents. It uses embeddings, vector database (Pinecone), and local Llama model to generate context-aware responses.
- Upload document (PDF, DOCX, TEXT)
- Automatic text chunking
- Embedding generation (convert texts to vectors)
- Vector search (Pinecone)
- Context-aware, AI responses (Llama)
- Source citations that allow users to view the docuemnt chunks used to generate response
- AI quiz generation from uploaded documents (Multiple-choice question)
- Response-time monitoring
- Chat-based interface
User question -> Convert to embedding -> Search vector database -> Retrieve relevant documents -> Build context -> send to Llama model -> Generate response
- Frontend: React.js
- Backend: Node.js, Express.js
- Database: MongoDB
- Vector Database: Pinecone
- AI Model: Local Llama (LM Studio)
- Embeddings: Local embedding API
cd backend npm install npm start
cd frontend npm install npm start
This project was originally developed as a group project. I later made improvements including:
- Integrated Pinecone vector search
- Improved embedding and retrieval pipeline
- Fixed memory/ context filtering issues
- Improved backend query flow
- Added source citation that allow users to view the document chunks used to generate AI responses
- Developed a AI quiz generator that creates multiple-choice questions from uploaded documents
- Built AI performance monitoring using InfluxDB to track response time, token usage, GPU and Memory usage
