🔍 Smart RAG Based Q&A Assistant

AI-powered assistant that answers natural language questions from uploaded PDFs using semantic retrieval, and LLM-based re-ranking. Delivers accurate, explainable responses with relevance scores and direct PDF highlighting for traceability.

---

📸 Demo

---

🧠 Features

Feature	Description
📄 PDF Upload	Upload and parse PDF with page-level metadata
🔍 Semantic Search	Embed + retrieve most relevant chunks using similarity scoring
🧠 Re-Ranking	Use LLM to sort top chunks before answering
💬 Chat Support	Ask questions via text or voice
📊 Similarity Scores	View how relevant each chunk is to your query
🔦 PDF Highlighting	See exactly which paragraph the answer came from

---

🛠️ Tech Stack

Layer	Tools
Frontend	React.js, TailwindCSS, React-PDF, Web Speech API
Backend (Node.js)	Node.js, Express.js, OAuth, JWT
Backend (AI-Engine)	Python, FastAPI , PyMuPDF
LLM	Ollama
Embeddings	all-MiniLM-L6-v2
Vector DB	ChromaDB

---

⚙️ Project Workflow

A clear separation of responsibilities ensures maintainability and scalability. The project is divided into two major flows:

---

🛠️ Admin Workflow (Document Management & Embedding)

Step	Description
1️⃣	PDF Upload: Admin uploads PDF files via the dashboard.
2️⃣	PDF Parsing: The system extracts text from each page using PyMuPDF .
3️⃣	Text Chunking: Extracted text is split into smaller chunks (with token limits).
4️⃣	Metadata Addition: Each chunk is enriched with metadata: chunk_index, page_number, text_start, text_end, source, etc.
5️⃣	Embedding Generation: Each chunk is passed through an embedding model (e.g., all-MiniLM-L6-v2) and stored in a vector database like ChromaDB.
✅	Ready for Querying: Admin-processed files are now available for user interaction.

---

🙋 User Workflow (Querying via Text or Voice)

Step	Description
1️⃣	File Selection: User selects a specific PDF file to query from the list of uploaded documents.
2️⃣	Input Method: User types a question or uses voice input (handled via react-speech-recognition).
3️⃣	Embedding Query: User query is converted into an embedding and searched in the vector database (ChromaDB).
4️⃣	Top-k Retrieval: Most similar chunks are retrieved based on cosine similarity.
5️⃣	Contextual Prompt Construction: Retrieved chunks and metadata are appended to the query for contextual understanding.
6️⃣	Answer Generation: Query is sent to a language model (e.g., GPT-4) along with relevant context for accurate response generation.
7️⃣	Answer Display: Answer is rendered on the UI. Additional metadata like page number and similarity score is also shown.
8️⃣	PDF Viewer Sync (Bonus): The highlighted chunk is shown in the PDF viewer with react-pdf.

---

✅ Bonus Features Implemented

• 🔎 Similarity Score Display
• 📄 PDF Viewer with Highlighted Chunks
• 🎙️ Voice Input Support
• 🔐 Admin-only Access for Upload & Embedding

---

📷 Screenshots

Chat Interface	PDF Highlight

---

Setup Instructions

1. Clone the Repository

   git clone https://github.com/shivamworld0608/Ollabot.git
   cd Ollabot
   

2. Create a .env file in backend

   #OAuth Credentials
   GOOGLE_CLIENT_ID=87796226935-nr26lcqfgqfsoepgr3h30qc4nn224lqt.apps.googleusercontent.com
   GOOGLE_CLIENT_SECRET=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
   
   
   #JWT Credentials
   JWT_SECRET="b1a26c4a14718e4244721cc7db67f6e42befce460c32f7e08f2040cb07ae4ed3"
   JWT_EXPIRES_IN="30d"
   JWT_COOKIE_EXPIRES_IN=30
   
   #basic server credentials
   MONGO_URI="mongodb+srv://pandeyashishivam:[email protected]/"
   CLIENT_URL="http://localhost:5173"
   AI_ENGINE_URL="http://localhost:8000"
   SERVER_URL="http://localhost:5000"
   PORT=5000
   
   
   

3. Create a .env file in frontend

   VITE_APP_BASE_URL='http://localhost:5000'
   
   

4. AI-Engine Setup (FastAPI)

   cd ai-engine
   pip install -r requirements.txt
   python main.py
   

5. Backend Setup (Nodejs,Express)

    cd backend
    npm i
    npx nodemon server.js
   

6. Frontend Setup (React)

   cd frontend
   npm install
   npm run dev
   

Make sure to set the correct backend URL in your frontend env and also correct ai-engine url in backend env

🤝 Contributing

We welcome contributions from the community! Here’s how you can help:

📌 Guidelines

• 📝 Open an Issue
  For major features or changes, please open an issue first to discuss your ideas.

• 📂 Follow Standards
  Stick to the existing project structure and naming conventions for consistency.

• ✅ Test Before Push
  Ensure all features are tested and stable before submitting a pull request.

---

📬 Contact

Have questions, feedback, or just want to connect? Feel free to reach out!

Platform	Link
GitHub	@shivamworld0608
Email	[email protected]
LinkedIn	linkedin.com/in/pandey-shivam-