💎 Experience the Future of Learning

Transform your video library into an interactive, intelligent knowledge base with our stunning Glassmorphism Aurora interface.

🚀 Quick Start • ✨ Features • 🎨 Live Demo • ⚙️ Architecture • 📖 Documentation

---

🎨 Live Interface Preview

🌟 Main Dashboard - Aurora Welcome Screen

Beautiful gradient-based UI with glassmorphism effects and pre-built example prompts

💬 Interactive Chat Interface

Real-time streaming responses with context-aware answers and video timestamps

---

🌟 What is Aurora RAG?

Aurora RAG transforms your video course library into an intelligent, AI-powered knowledge base Production-ready conversational assistant with stunning Glassmorphism UI

🏗️ Built With FastAPI Backend ⚡ High-performance async Vanilla JS/CSS Frontend 🎨 Zero dependencies Ollama AI Models 🤖 Local & private

🎯 Core Technology Hybrid RAG Pipeline 🔍 BM25 + Embeddings Cross-Encoder Reranking 🎯 Precision results Streaming Responses ⚡ Real-time feedback

💎 Design Philosophy Glassmorphism UI ✨ Premium aesthetics Aurora Theme 🌌 Purple-cyan gradients Mobile-First 📱 Fully responsive

---

🎭 Why Choose Aurora?

💡 "It's not just about finding answers; it's about the experience."

🧠 Context-Aware Intelligence Remembers your entire conversation history for coherent, multi-turn dialogues. Ask follow-up questions naturally without repeating context. ✅ Multi-turn conversations ✅ Context retention ✅ Smart reference resolution ✅ Conversation branching ⚡ Real-Time Streaming ChatGPT-style token-by-token streaming for instant feedback. See responses as they're generated, not after completion. ⚡ Token streaming ⚡ Progress indicators ⚡ Stop generation control ⚡ Sub-second first token 🎯 Precision Retrieval Hybrid search combining BM25 keyword matching with semantic embeddings, followed by cross-encoder reranking for maximum accuracy. 🔍 Hybrid search (BM25 + Vector) 🎯 Cross-encoder reranking 📊 Relevance scoring 📑 Source attribution

📊 How Aurora Works graph TD Q[🎓 User Question] -->|Hybrid Search| S[🔍 Retrieval] S -->|BM25 + Vector| R[🎯 Reranking] R -->|Cross-Encoder| C[📚 Context] C -->|Prompt| L[🤖 LLM Generation] L -->|Stream| A[💬 Final Answer] style Q fill:#7F00FF,color:#fff style A fill:#00D9FF,color:#fff Loading 🎬 Smart Deep Linking Jump directly to exact timestamps in source videos. Every answer includes clickable timestamps that take you to the relevant moment. ⏱️ Precise timestamp extraction 🔗 Clickable video links 📽️ Multiple source references 💯 Confidence scoring 🎨 Premium UI/UX Beautiful Aurora theme with purple-cyan gradients, glassmorphism effects, smooth animations, and dark mode optimization. ✨ Glassmorphism design 🌊 Smooth transitions ⌨️ Typing animations 📋 Copy to clipboard

⚡ Performance Highlights

~200ms Hybrid Search

~150ms Reranking

~500ms First Token

100% Local & Private

---

✨ Core Features

Everything you need for a premium RAG experience, out of the box

🤖 Intelligent Backend Architecture

🏗️ FastAPI Core High-Performance Server • async/await for high concurrency • Type-safe with Pydantic models • Auto-generated OpenAPI docs • CORS-enabled REST API • Health check endpoints

🧠 RAG Pipeline Advanced Retrieval System • Multi-stage retrieval pipeline • Context-aware generation • Smart chunk selection • Prompt engineering • Source attribution

📡 Streaming Real-Time Responses • Token-by-token streaming • Newline-delimited JSON • Graceful error handling • Stop generation control • Progress feedback

💾 Sessions Persistent History • SQLite-based storage • Multi-session support • Conversation history • Context retention • Export capabilities

🎨 Premium Frontend Experience

🌌 Aurora Theme Stunning Visuals • Purple-cyan gradient palette • Smooth color transitions • Dark mode optimized • Custom CSS variables • Consistent branding

💎 Glassmorphism Modern Design Language • Translucent card effects • Backdrop blur filters • Elevated shadows • Frosted glass aesthetics • Layered depth

📱 Responsive Works Everywhere • Mobile-first architecture • Adaptive layouts • Touch-optimized controls • Tablet support • Cross-browser compatible

⚡ Interactive Rich User Experience • Typing animations • Stop generation button • Copy to clipboard • Markdown rendering • Syntax highlighting

🔍 Search & Retrieval Features

Hybrid Search 🔤 BM25 Keyword Search Traditional TF-IDF ranking ➕ 🧮 Semantic Vector Search Deep learning embeddings = ✨ Best of Both Worlds

Cross-Encoder Reranking 📊 Initial Retrieval Get top-100 candidates ⬇️ 🎯 Precise Scoring Rerank with transformer model ⬇️ 🏆 Top-K Results Return best 5-10 chunks

Source Attribution 📹 Video References Exact timestamps ➕ 📄 Chunk Metadata Course & module info = 🔗 Clickable Links

🛠️ Technology Stack

Frontend

Backend

LLM Engine

Vector DB

Database

Models: llama3.2:latest • bge-m3:latest • nomic-embed-text • gemma3:4b

---

🚀 Quick Start Guide

Get Aurora RAG running in under 5 minutes!

📋 Prerequisites

✅ Python 3.8 or higher
✅ Ollama installed with models: llama3.2, bge-m3
✅ Modern web browser (Chrome, Firefox, Edge)

⚡ Installation Steps

1️⃣	Clone the Repository git clone <repository-url> cd RAG-Based-AI
2️⃣	Install Dependencies cd project/backend pip install -r requirements.txt
3️⃣	Start the Backend Server # From project/backend directory uvicorn main:app --port 8000 --host 0.0.0.0 Or use the convenient startup script: # Windows .\start_aurora.bat # Linux/Mac ./start_aurora.sh ✨ API Documentation: http://localhost:8000/docs
4️⃣	Launch the Frontend Simply open project/frontend/index.html in your browser! # Or serve it locally cd project/frontend python -m http.server 3000 🌐 Open: http://localhost:3000

🎉 That's it! Start chatting with your video courses!

---

⚙️ System Architecture

graph TB
    User[👤 User] -->|HTTPS| Frontend[🎨 Aurora Frontend UI]
    Frontend -->|REST API| Backend[⚡ FastAPI Backend Server]
    
    Backend --> Pipeline[🧠 RAG Pipeline Orchestrator]
    
    Pipeline --> Search[🔍 Hybrid Search Engine]
    Search --> Vector[📊 Semantic Search<br/>ChromaDB + Embeddings]
    Search --> Keyword[📝 Keyword Search<br/>BM25 Index]
    
    Pipeline --> Rerank[🎯 Cross-Encoder Reranking<br/>Score & Sort Results]
    
    Pipeline --> Context[📚 Context Builder<br/>Prompt Engineering]
    
    Context --> LLM[🦙 Ollama LLM<br/>llama3.2:latest]
    
    LLM -->|Streaming Tokens| Backend
    Backend -->|JSON Stream| Frontend
    
    Backend --> DB[(💾 SQLite Database<br/>Chat History)]
    
    style Frontend fill:#7F00FF,color:#fff
    style Backend fill:#00D9FF,color:#fff
    style Pipeline fill:#FFD700,color:#000
    style LLM fill:#00FF88,color:#000

Loading

🔄 How It Works

1️⃣ 🎤 User Query Submit question via chat

2️⃣ 🔍 Search BM25 + Vector similarity

3️⃣ 🎯 Rerank Cross-encoder scoring

4️⃣ 📚 Context Build RAG prompt

5️⃣ 🤖 Generate LLM produces answer

6️⃣ 💬 Stream Display in real-time

---

📂 Project Structure

RAG-Based-AI/
│
├── 📁 project/
│   │
│   ├── 📁 backend/                    # ⚡ FastAPI Backend Application
│   │   ├── main.py                    # 🚀 API entry point & endpoints
│   │   ├── rag_pipeline.py            # 🧠 Core RAG orchestration logic
│   │   ├── search.py                  # 🔍 Hybrid search engine
│   │   ├── models.py                  # 📋 Pydantic request/response models
│   │   ├── config.py                  # ⚙️  Configuration management
│   │   ├── prompts.py                 # 💬 Prompt templates
│   │   ├── utils.py                   # 🛠️  Helper utilities
│   │   ├── database.py                # 💾 SQLite operations
│   │   ├── requirements.txt           # 📦 Python dependencies
│   │   ├── embeddings.joblib          # 📊 Pre-computed embeddings (7MB)
│   │   ├── bm25_index.joblib          # 📝 BM25 search index (8MB)
│   │   └── chat_history.db            # 💬 Chat session database
│   │
│   └── 📁 frontend/                   # 🎨 Aurora UI (Vanilla JS)
│       ├── index.html                 # 🌐 Single Page Application
│       ├── style.css                  # 💎 Glassmorphism styling
│       ├── script.js                  # ⚡ Chat logic & API calls
│       └── 📁 assets/                 # 🖼️  Images & resources
│
├── 📁 legacy/                         # 📦 Legacy/Archived Scripts
│   ├── app.py                         # Old Streamlit prototype
│   ├── video_to_mp3.py                # Video ingestion pipeline
│   ├── preprocess_json.py             # Data preprocessing
│   └── ...
│
├── 📁 jsons/                          # 📄 Transcription JSON files
│   ├── 01_Installing VS Code.json
│   ├── 02_Your First HTML.json
│   └── ... (18 course transcripts)
│
├── 📁 tests/                          # 🧪 Testing & Evaluation
│   ├── evaluate.py                    # RAG evaluation metrics
│   └── eval_dataset.json              # Test dataset
│
├── README.md                          # 📖 This file
├── .env                               # 🔐 Environment variables
├── start_aurora.bat                   # 🪟 Windows startup script
└── start_aurora.sh                    # 🐧 Linux/Mac startup script

---

🔮 Future Roadmap

🚧 In Development 🎙️ Voice Input/Output Speak to the assistant, hear responses 📊 Analytics Dashboard Query statistics and usage metrics 🔄 Auto-Refresh Index Watch folder for new videos 🌍 Multi-Language Support i18n for global users

💡 Planned Features 🤖 Multi-Model Switching DeepSeek, Mistral, GPT-4, Claude 📤 File Upload Drag & drop PDFs, DOCX, TXT 🐳 Docker Compose One-command deployment 🔐 Authentication User accounts and permissions

💬 Have a feature request? Open an issue!

---

📖 Documentation

🔌 API Endpoints

Method	Endpoint	Description
GET	/	API information and health status
POST	/chat	Send a query and receive streaming response
GET	/sessions	List all chat sessions
GET	/sessions/{id}	Get specific session history
GET	/models	List available Ollama models
GET	/health	Backend health check

📚 Full API Docs: Visit http://localhost:8000/docs after starting the backend

---

⚡ Performance Metrics

🔍 Search Speed ~200ms Hybrid retrieval time

🎯 Reranking ~150ms Cross-encoder scoring

🤖 First Token ~500ms Time to first response

📊 Index Size 720 chunks From 18 video courses

Tested on: Intel i7-12700K, 32GB RAM, RTX 3080

---

🤝 Contributing

We welcome contributions from the community!

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

Read CONTRIBUTING.md for detailed guidelines

Follow our CODE_OF_CONDUCT.md

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

Built with amazing open-source tools:

FastAPI • Ollama • ChromaDB • Sentence-Transformers • BM25

---

💜 Built with passion by

Piyush Ramteke

---

⭐ If you found this helpful, please consider giving it a star!

_{© 2026 Aurora RAG Assistant. All rights reserved.}