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.
- 🔒 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.
- Frontend: Next.js 15 / Vite
- AI Inference: Transformers.js (Llama 3.2 1B quantized)
- Database & Auth: Supabase
- Animations: Framer Motion
- Icons: Lucide React
- Styling: Tailwind CSS
- Node.js 18.x or higher
- A Supabase project (for Authentication)
- git clone https://github.com/your-username/Documind.git
- cd Documind
- npm install
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
DocuMind uses a Multi-Threaded Architecture to ensure smooth performance:
- React UI rendering
- Animations & interactions
- File handling
- User input
- Loads and runs the AI model
- Handles inference (text generation)
- Prevents UI blocking
- Stores chat history
- Maintains session persistence
- Quantized models for low memory usage
- WebGPU / WASM fallback
- Context trimming for faster inference
- Chunk-based PDF processing (optional upgrade)
- Worker-based parallel execution
- Push your project to GitHub
- Connect your repo to Vercel
- Add environment variables in dashboard
- Deploy
- Ensure your Supabase redirect URLs include your production domain
- Use HTTPS for WebGPU support in production
- 🔍 Semantic search (embeddings-based retrieval)
- 📚 Multi-PDF knowledge base
- ☁️ Cloud sync (Firestore / Supabase)
- 🧠 Conversation memory
- 📊 Answer citations & sources
- 🔐 Full authentication system
- 📱 Mobile optimization
Distributed under the MIT License.
See LICENSE for more information.
- Hugging Face team for Transformers.js
- Open-source AI community
- Shadcn/UI inspiration for clean design
Built for Privacy, Speed, and Local Intelligence
No cloud. No tracking. Just your data — on your device.
If you like this project:
- ⭐ Star the repo
- 🍴 Fork it
- 🧠 Build something amazing
Built with ❤️ for Privacy and Speed.