Run Python in the Browser

How to run Python in the browser

1. Open the Python editor.
2. Under the Run button, select Switch and choose Browser Python instead of Execution Server.
3. Select Run. The first time, the browser downloads Python and the data packages (about 35 MB in total) and caches them, so later runs start much faster.
4. Import NumPy, pandas, SciPy or scikit-learn as usual. The example below prints emscripten as the platform when it runs in the browser, and linux when it runs on the server.

import sys
import numpy as np
import pandas as pd

print("Python", sys.version.split()[0], "on", sys.platform)

df = pd.DataFrame({"city": ["Austin", "Boston", "Chicago"], "temp_f": [91, 77, 84]})
df["temp_c"] = ((df["temp_f"] - 32) * 5 / 9).round(1)
print(df)
print("Mean (C):", np.mean(df["temp_c"]).round(1))

Browser Python or Execution Server

How Python runs in a web page

Browsers run JavaScript and WebAssembly, not Python. The Pyodide project compiles CPython and scientific packages to WebAssembly, so the browser can run the same interpreter you would install on a computer. Other projects take different routes: PyScript runs Python inside HTML pages on top of Pyodide or MicroPython, JupyterLite runs notebooks in the browser, and Brython translates Python to JavaScript.

To run Python in your own page, load Pyodide and call runPython:

<script src="https://cdn.jsdelivr.net/pyodide/v0.25.1/full/pyodide.js"></script>
<script>
  async function main() {
    const pyodide = await loadPyodide();
    console.log(pyodide.runPython("sum(i * i for i in range(10))"));  // 285
  }
  main();
</script>

To put a runnable example on a blog or in course notes without writing any code, use the Run on CodeUtility button.

Limits of Python in the browser

• No direct network socketsBrowser security does not allow raw sockets, so libraries such as requests do not work as they are. Use the Execution Server for programs that call web APIs.
• No desktop windowsTkinter and other GUI toolkits cannot open windows. For charts, use the Matplotlib editor, which shows the figure next to the code; turtle drawings work on the Python turtle page.
• MemoryA browser tab has less memory than a computer, so very large datasets are better on the Execution Server.
• First loadPython and its packages are downloaded once, about 35 MB; on a slow connection the first run takes a while.

Frequently asked questions