Skip to content

shubhd556/Documind

Repository files navigation

🧠 DocuMind AI

Chat with your PDFs, Fully Offline. Privacy-First Document Intelligence.

DocuMind AI is a premium, browser-native document intelligence platform. Unlike traditional "Chat with PDF" apps that send your sensitive data to the cloud, DocuMind processes everything locally on your machine using WebGPU and Web Workers.

No data leaves your browser. No subscriptions. Just pure, local power.

License React Next.js Supabase Tailwind

image

✨ Key Features

  • 🔒 Local-First Privacy: AI processing happens entirely in the browser via Transformers.js. Your documents never touch a server.
  • ⛓️ AI Chain of Thought: A transparent "Thinking" UI that shows the AI's internal reasoning steps in real-time.
  • 📂 Multi-Document Sessions: Seamlessly switch between multiple PDFs. Session history and context are preserved locally.
  • 🌓 Split-View Preview: A professional-grade, edge-to-edge PDF viewer toggled directly from the sidebar.
  • ⚡ WebGPU Accelerated: Utilizes modern hardware acceleration for lightning-fast inference on Llama 3.2 1B models.
  • 🎨 Premium UI: A modern, dark-mode SaaS aesthetic built with Tailwind CSS, Framer Motion, and Shadcn/UI.

🛠️ Tech Stack


🚀 Getting Started

1. Prerequisites

  • Node.js 18.x or higher
  • A Supabase project (for Authentication)

2. Installation

3. Environment Variables

Create a .env.local file in the root directory and add your Supabase credentials:

  • VITE_SUPABASE_URL=your_supabase_project_url
  • VITE_SUPABASE_ANON_KEY=your_supabase_anon_key

🏗️ Architecture

DocuMind uses a Multi-Threaded Architecture to ensure smooth performance:

🧩 Main Thread

  • React UI rendering
  • Animations & interactions
  • File handling
  • User input

⚙️ Web Worker

  • Loads and runs the AI model
  • Handles inference (text generation)
  • Prevents UI blocking

💾 Local Storage

  • Stores chat history
  • Maintains session persistence

⚡ Performance Strategy

  • Quantized models for low memory usage
  • WebGPU / WASM fallback
  • Context trimming for faster inference
  • Chunk-based PDF processing (optional upgrade)
  • Worker-based parallel execution

🌐 Deployment

🚀 Vercel (Recommended)

  1. Push your project to GitHub
  2. Connect your repo to Vercel
  3. Add environment variables in dashboard
  4. Deploy

⚠️ Important

  • Ensure your Supabase redirect URLs include your production domain
  • Use HTTPS for WebGPU support in production

🔮 Future Improvements

  • 🔍 Semantic search (embeddings-based retrieval)
  • 📚 Multi-PDF knowledge base
  • ☁️ Cloud sync (Firestore / Supabase)
  • 🧠 Conversation memory
  • 📊 Answer citations & sources
  • 🔐 Full authentication system
  • 📱 Mobile optimization

📜 License

Distributed under the MIT License.
See LICENSE for more information.


🙌 Acknowledgments

  • Hugging Face team for Transformers.js
  • Open-source AI community
  • Shadcn/UI inspiration for clean design

💙 Philosophy

Built for Privacy, Speed, and Local Intelligence

No cloud. No tracking. Just your data — on your device.


⭐ Support

If you like this project:

  • ⭐ Star the repo
  • 🍴 Fork it
  • 🧠 Build something amazing

👨‍💻 Author

Built with ❤️ for Privacy and Speed.

About

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors