Frequently Asked Questions

Is Eigenvue free?

Yes. Eigenvue is free and open-source under the MIT licence. Both the web application at eigenvue.web.app and the Python package (pip install eigenvue) are available at no cost. You can inspect, modify, and redistribute the source code on GitHub.

Can I use Eigenvue offline?

The web application requires an internet connection on the first visit. After the initial load, most assets are cached by the service worker, so basic browsing may work offline. For a fully offline experience, use the Python package: once installed, eigenvue.show() and eigenvue.jupyter() run a local server and do not need network access.

Can I use Eigenvue in the classroom?

Absolutely. Educators are encouraged to use Eigenvue for lectures, assignments, and labs. You can:

• Project the web app and step through an algorithm live during a lecture.
• Share a URL with a specific step and input set so students can explore on their own (see Sharing & URLs).
• Assign notebook-based exercises that use eigenvue.jupyter() to visualize solutions.

No licence fees apply for educational or commercial use.

Which browsers are supported?

Eigenvue supports the latest two major versions of:

• Google Chrome
• Mozilla Firefox
• Microsoft Edge
• Apple Safari (macOS and iOS)

JavaScript must be enabled. The visualizer relies on the Canvas API and modern ES modules, so very old browsers (Internet Explorer, legacy Edge) are not supported.

Can I embed a visualization in my own website?

Not yet as a first-class feature, but you can embed the visualizer using an <iframe>:

<iframe
  src="https://eigenvue.web.app/algo/merge-sort?data=[9,4,7,2,1]&step=0"
  width="100%"
  height="600"
  frameborder="0"
  allow="fullscreen"
></iframe>

The embedded page includes the full three-panel layout and playback controls. A dedicated embed mode with a slimmer UI is on the roadmap.

How do I report a bug?

Open an issue on the GitHub issue tracker. Please include:

1. A description of the unexpected behaviour.
2. The algorithm name and the inputs you used.
3. Your browser (or Python version) and operating system.
4. Screenshots or console output if available.

The maintainers triage new issues regularly and aim to respond within a few days.

How do I request a new algorithm?

File a feature request on the GitHub issue tracker with the label enhancement. Describe the algorithm, link to a reference (paper, textbook chapter, Wikipedia article), and explain why it would be a good fit for the platform. Community upvotes help prioritise which algorithms are built next.

If you want to implement it yourself, see the contributing guide for instructions on adding a new algorithm generator.

I get an error when I run pip install eigenvue. What should I do?

Try the following steps in order:

1. Check your Python version. Eigenvue requires Python 3.10 or later. Run python --version to confirm.
2. Upgrade pip.

   python -m pip install --upgrade pip

3. Use a virtual environment to avoid conflicts with system packages:

   python -m venv .venv
   source .venv/bin/activate   # macOS / Linux
   .venv\Scripts\activate      # Windows
   pip install eigenvue

4. Check your network. The package is hosted on PyPI, so you need internet access for the initial install.

If the issue persists, open a GitHub issue with the full error output.

Can I upload my own model for visualization?

Not in the current release. The visualizations are based on built-in algorithm generators that ship with the package. Support for user-supplied model weights and custom architectures is planned for a future version. In the meantime, you can contribute a new generator if your algorithm fits the existing generator interface.

How does Eigenvue handle floating-point numbers?

All numeric computations in the web app use JavaScript’s IEEE 754 double-precision floating point. In the Python package, standard Python float (also IEEE 754 double precision) is used. This means:

• Very small rounding differences may appear in deep-learning and generative-AI visualizations (for example, softmax outputs that sum to 0.9999999 instead of 1.0).
• The state inspector displays values rounded to a sensible number of decimal places for readability, but the underlying computation uses full precision.
• If you need exact comparisons in your own code, use a tolerance (for example, math.isclose(a, b, rel_tol=1e-9)) rather than strict equality.

Floating-point behaviour is consistent across browsers and matches the Python package output for identical inputs.