framejs

An anywidget for embedding framejs frames (and metapage URLs) in Jupyter and marimo notebooks.

Renamed in 0.4.0. This package was published as metaframe-widget, and the class as MetaframeWidget. Both names still work — metaframe-widget now installs this package, and MetaframeWidget is an alias of Frame — but new code should use framejs and Frame.

Install

pip install framejs

With environment extras:

pip install "framejs[jupyter]"   # includes jupyterlab
pip install "framejs[marimo]"    # includes marimo

Quick start

Jupyter

from framejs import Frame

w = Frame(url="https://framejs.app/j/f14d583125634c23851453c8038ddb7c", height="300px")
w

marimo

import marimo as mo
from framejs import Frame

w = Frame(url="https://framejs.app/j/f14d583125634c23851453c8038ddb7c", height="300px")
mo.ui.anywidget(w)

A widget is always created from a URL, editable at the URL.

URL forms

Any of these work as url=:

# Expiring snapshot — content-addressed, kept ~30 days, then garbage-collected
# (editor: "Create expiring snapshot")
Frame(url="https://framejs.io/j/<sha256>")

# Durable, editable frame — permanent, tied to your account (editor: "Save")
# Paste either host: a framejs.app/j/<uuid> URL is loaded from framejs.io, the
# runtime that serves frames with CORS headers (a notebook cannot load the
# framejs.app one directly). `w.url` still reads back what you passed.
Frame(url="https://framejs.app/j/<uuid>")

Prefer the durable /j/<uuid> form in notebooks you keep. See Short URLs for the full comparison.

API reference

Methods

• set_inputs(d) — merge a dict into current inputs
• set_input(key, value) — set a single input key
• on_outputs_change(callback) — register a callback for output changes
• on_saved_url_change(callback) — register a callback for when the user creates an expiring snapshot (/j/<sha256>) inside the widget
• pipe_to(target, output_key, input_key=None) — connect an output to another widget's input

Piping widgets

source = Frame(url="...")
sink = Frame(url="...")
source.pipe_to(sink, output_key="result", input_key="data")

No anywidget? Use a plain iframe

To send data into a frame and nothing more, you need no package at all — encode the inputs into the URL and show it with IPython.display.IFrame. See the Jupyter guide.

Links

• GitHub
• Jupyter examples
• marimo examples