📚 RAG-based Chatbot with Gemini + Pinecone

An interactive Retrieval-Augmented Generation (RAG) chatbot built with Node.js, Express, LangChain, Google Gemini API, and Pinecone.
It allows you to chat with your documents (PDFs) through a clean, modern web interface.

---

🚀 Features

• 📄 PDF Ingestion & Indexing – Load a PDF, split into chunks, embed with Gemini, and store in Pinecone
• 🔍 RAG-powered QA – Retrieves relevant chunks and generates answers using Gemini
• 💬 Multi-turn Conversations – Context-aware chat with history memory
• 🎨 Modern Frontend – TailwindCSS-based UI with dark mode
• 🖥️ Chat Controls
  • Clear Chat → clears UI only (bot still remembers)
  • New Chat → resets both UI and backend context
• ⏳ Typing Indicator – Bot shows “🤖 Bot is typing...”
• 💾 Persistent Chat – Messages survive page reloads (localStorage)

---

🛠️ Tech Stack

• Backend: Node.js, Express, LangChain, Pinecone, Google Generative AI
• Frontend: HTML, TailwindCSS, Vanilla JS
• Database: Pinecone (Vector DB for embeddings)
• Embedding Model: text-embedding-004 (Gemini)
• LLM: gemini-2.0-flash

---

📂 Project Structure

RAG_ChatBot/
├── public/           # Frontend files
│   ├── script.js     # Chat logic (frontend)
│   ├── style.css     # Custom styling (scrollbar, overrides)
|
│── views/ 
|   ├── index.html    # Chat UI
|
│── index.js          # Index PDF into Pinecone
│── query.js          # Query logic with RAG + Gemini
│── server.js         # Express server (API endpoints)
│── package.json      # Dependencies & scripts
│── .env              # API keys (Gemini, Pinecone)
│
└── node_modules/     # Installed dependencies


## ⚡ Setup Instructions

### Set up Pinecone Index
Go to Pinecone Console
Create a new Index with:
Name: same as PINECONE_INDEX_NAME in .env
Dimension: 768 (for Gemini embeddings text-embedding-004)
Metric: cosine

### Clone the Repository  
cd RAG_ChatBot open terminal there
Install Dependencies
Create a .env file in the project root
and put your API keys for
GEMINI_API_KEY, PINECONE_API_KEY, PINECONE_ENVIRONMENT, PINECONE_INDEX_NAME

### Index your PDF
Keep your PDF in the root folder with the name doc.pdf
node index.js

### Start the Server
npx nodemon server.js

Screenshots