Understanding algorithms requires building spatial and temporal intuition that textbooks and static diagrams cannot provide. Existing tools are domain-specific — one for sorting, another for attention mechanisms, yet another for quantum circuits. None offer a unified visual language across these domains, and none bridge web-based exploration with programmatic workflows.

Eigenvue solves this with:

• One platform, four domains — classical algorithms, deep learning, generative AI, and quantum computing under a single interaction model.
• Dual distribution — explore in the browser at eigenvue.web.app, or pip install eigenvue for Jupyter notebooks and research scripts.
• Real code, not config — algorithms are defined as TypeScript and Python functions, making them testable, debuggable, and contributor-friendly.

22 algorithms with step-by-step animated visualizations, synchronized code highlighting (pseudocode, Python, JavaScript), key concepts, common pitfalls, quizzes, and further resources.

Browse the full catalog: Algorithm Reference →

Explore all visualizations instantly — no installation required:

👉 eigenvue.web.app

pip install eigenvue

import eigenvue

# List all available algorithms
eigenvue.list()

# Open an interactive visualization in your browser
eigenvue.show("binary-search")

# Get step data programmatically
steps = eigenvue.steps("self-attention")

# Use in a Jupyter notebook  (pip install eigenvue[jupyter])
eigenvue.jupyter("transformer-block")

npm install eigenvue

import { list, steps, show } from "eigenvue";

// List all available algorithms
list();

// Get step data programmatically
const result = steps("binary-search");

// Open an interactive visualization in your browser
show("self-attention");

Eigenvue follows a clean three-layer architecture:

┌─────────────────────────────────────────────────────────┐
│  Generator Layer       TypeScript & Python functions     │
│                        that produce Step sequences       │
├─────────────────────────────────────────────────────────┤
│  Step Format           Universal JSON contract (v1.0.0)  │
│                        between generators & renderers    │
├─────────────────────────────────────────────────────────┤
│  Rendering Layer       Canvas 2D engine with layouts,    │
│                        animation, and playback controls  │
└─────────────────────────────────────────────────────────┘

Adding a new algorithm never requires modifying the rendering engine. Improving the renderer never requires touching algorithm logic. Cross-language parity is CI-enforced to ±1e-9 tolerance.

📐 Architecture docs →

eigenvue/
├── algorithms/          # Shared algorithm definitions (meta.json per algorithm)
│   ├── classical/       # Binary Search, Bubble Sort, QuickSort, …
│   ├── deep-learning/   # Perceptron, Feedforward Network, …
│   ├── generative-ai/   # Tokenization, Self-Attention, …
│   └── quantum/         # Grover's Search, Quantum Gates, …
├── web/                 # Next.js web application (Canvas 2D, Tailwind CSS)
├── python/              # Python package (pip install eigenvue)
├── node/                # npm package (zero runtime dependencies)
├── shared/              # Shared types, schemas, and constants
├── docs/                # Documentation site (Astro / Starlight)
├── scripts/             # Build, codegen, and maintenance scripts
├── tests/               # Cross-platform integration tests
└── .github/             # CI/CD workflows

Prerequisites: Node.js 22+, Python 3.10+, Git

cd web
npm install
npm run dev          # → http://localhost:3000

cd python
pip install -e ".[dev]"
pytest

cd node
npm install
python ../scripts/bundle-node-data.py
npm run build
npm test

cd docs
npm install
npm run build

Full documentation is available at eigenvue.web.app/docs.

Contributions are welcome! Please read the Contributing Guide for development setup, branching strategy, and commit conventions.

Adding a new algorithm? Follow the step-by-step guide: Adding an Algorithm →

If you use Eigenvue in research or teaching, please cite:

@article{eigenvue2026,
  title   = {Eigenvue: A Visual Learning Platform for Algorithms,
             AI Architectures, and Quantum Computing},
  author  = {Mishra, Ashutosh},
  year    = {2026},
  url     = {https://eigenvue.web.app}
}

See the full Citation guide →

MIT © 2026 Ashutosh Mishra