📄 DocuChat

DocuChat is a high-performance Document & Web Analyzer built with LangChain, Gemini 1.5 Flash, and Pinecone. It enables users to upload PDF documents or provide website URLs to instantly chat with the content using state-of-the-art RAG (Retrieval-Augmented Generation).

🚀 Features

• PDF Knowledge Extraction: Upload and vectorize complex PDF documents for instant querying.
• Web URL Scraper: Paste any article or website URL to chat with its real-time content.
• Intelligent Conversations: Powered by Google Gemini 1.5 Flash for low-latency, context-aware responses.
• Advanced Vector Search: High-precision retrieval using Cohere English v3.0 Embeddings (1024-dim).
• Rich Markdown UI: Beautifully formatted AI responses including lists, tables, and code snippets.
• DB Management: One-click "Reset DB" to wipe vectors from Pinecone and start a fresh session.
• Premium Design: Sleek, dark-mode ready interface built with Tailwind CSS 4.

---

✨ Latest Improvements (May 2026)

Precision & Performance

• Smaller Chunks (500 chars) for more targeted, relevant responses
• Concise Answers - System prompt enforces 5-sentence limit
• Smart Context Filtering - Limits context to 2000 chars max
• Better Separation - Chunks separated with "---" for clarity

Reliability & Safety

• File Validation - PDF type/size checks (25MB limit)
• URL Validation - Validates format before processing
• Message Validation - 3-500 character range enforcement
• Request Timeout - 120-second limit prevents hanging requests
• Better Error Messages - Clear, actionable error feedback

Configuration

• Environment Variables - API URL now configurable (.env.local)
• Production Ready - Easy switch between dev/production endpoints

See SETUP_GUIDE.md for detailed configuration and troubleshooting.

---

🛠️ Tech Stack

Frontend

• Framework: Next.js 15 (App Router)
• Styling: Tailwind CSS 4
• Markdown Rendering: react-markdown
• Icons: Lucide React

Backend

• Runtime: Node.js + Express 5
• Orchestration: LangChain
• LLM: Google Gemini 1.5 Flash
• Embeddings: Cohere AI (embed-english-v3.0)
• Vector DB: Pinecone (v5 SDK)
• Storage: ImageKit

---

⚙️ Setup & Installation

1. Clone the Repository

git clone <your-repo-url>
cd DocuChat

2. Backend Configuration

Navigate to server and create a .env file:

PORT=3001
GEMINI_API_KEY=your_gemini_key
COHERE_API_KEY=your_cohere_key
PINECONE_API_KEY=your_pinecone_key
PINECONE_INDEX=your_index_name
IMAGEKIT_PUBLIC_KEY=your_key
IMAGEKIT_PRIVATE_KEY=your_key
IMAGEKIT_URL_ENDPOINT=your_endpoint

Install and start:

cd server
npm install
npm start

3. Frontend Configuration

cd ../client
npm install
npm run dev

---

📁 Project Structure

DocuChat/
├── client/              # Next.js Frontend
│   ├── app/             # App Router logic & UI
│   └── public/          # Favicon & Assets
├── server/              # Express Backend
│   ├── src/
│   │   ├── controllers/ # Business logic (Chat, Upload, Clear)
│   │   ├── routes/      # Express API endpoints
│   │   └── utils/       # RAG logic & Scrapers
│   └── uploads/         # Temporary processing
└── README.md

---

🛡️ Key Functions

• Hybrid Loader: Dynamically handles local PDFs, Cloud PDFs (ImageKit), and Web URLs.
• MRL Embeddings: Optimized 1024-dimension vector generation for high semantic accuracy.
• Auto-Cleanup: Temporary file handling with automated purging post-vectorization.

📝 License

This project is licensed under the ISC License.

---

Built with ❤️ for intelligent document interaction.