A browser-based tool that transforms academic research papers into runnable Jupyter notebooks. Upload a PDF and get an educational implementation that demonstrates the paper's core algorithms using simplified, CPU-friendly code.
- Multi-Provider Support: Works with Gemini, OpenAI, Anthropic, Groq, Ollama (local), and HuggingFace
- Multi-Step Generation: Optional 3-step pipeline (Analyze → Design → Generate) for higher quality output
- Domain Detection: Automatically identifies paper domain (ML, NLP, Vision, RL) and applies appropriate simplifications
- Browser-Only: No backend required - runs entirely in your browser
- Educational Output: Generated notebooks include explanations, visualizations, and step-by-step comments
Prerequisites: Node.js 18+
# Install dependencies
npm install
# Start development server
npm run devOpen http://localhost:5173 in your browser.
- Select an AI provider from the dropdown
- Enter your API key (not required for Ollama)
- Toggle "Multi-Step Generation" for better quality (takes longer)
- Upload a research paper PDF
- Download the generated
.ipynbnotebook
| Provider | API Key Required | Notes |
|---|---|---|
| Google Gemini | Yes | Best PDF support (native) |
| OpenAI | Yes | GPT-4o has vision support |
| Anthropic | Yes | Claude 3 models supported |
| Groq | Yes | Fast inference |
| Ollama | No | Local models, requires Ollama running |
| HuggingFace | Yes | Inference API |
├── App.tsx # Main application component
├── components/
│ ├── ConfigPanel.tsx # Provider and model selection
│ ├── FileUpload.tsx # PDF upload handler
│ ├── ProcessingState.tsx # Progress indicator
│ └── ResultView.tsx # Notebook preview and download
├── services/
│ ├── geminiService.ts # Core generation logic for all providers
│ ├── orchestratorService.ts # Multi-step generation pipeline
│ ├── pdfService.ts # PDF text extraction
│ └── modelService.ts # Model fetching utilities
├── constants.ts # System prompts and templates
└── types.ts # TypeScript type definitions
For Ollama users, ensure the server is running with CORS enabled:
OLLAMA_ORIGINS='*' ollama serve- Generated notebooks are simplified implementations for educational purposes
- Performance will not match original paper results
- Scanned PDFs may not extract text properly (Gemini handles these best)
- Some providers may have rate limits or CORS restrictions
MIT