Output appended to Output Widget via append_display_data not displayed

[basnijholt]

Environment data

• VS Code version: Version: 1.101.0
• Jupyter Extension version (available under the Extensions sidebar): 2025.5.0
• Python Extension version (available under the Extensions sidebar): 2025.8.0
• OS MacOS 15.5 (24F74)
• Python and/or Anaconda version: 3.13.3
• Type of virtual environment used (N/A | venv | virtualenv | conda | ...): uv venv && uv pip install ipywidgets notebook
• Jupyter server running: Local

Expected behaviour

When adding content to Output widgets via append_display_data() asynchronously, the content should appear in the Output widgets.

Actual behaviour

• Output widgets appear correctly
• BUT Output widgets render as empty (no content is displayed)
• Console shows errors: "Model not found in Kernel state to render output" and "vscode.fallbackToNextRenderer"
• The same code works perfectly in regular Jupyter Lab/Notebook

Steps to reproduce:

Minimal Example (core issue)

The issue occurs with any asynchronous content addition to Output widgets:

import asyncio
from IPython.display import display
from ipywidgets import HTML, Output

outputs = [Output() for _ in range(10)]
display(*outputs)

async def reproduce_bug():
    for i in range(10):
        await asyncio.sleep(0.01)
        outputs[i].append_display_data(HTML(f"Content {i}"))

asyncio.create_task(reproduce_bug())

Root cause: VS Code widget renderer loses track of Output widget state when content is added asynchronously via append_display_data(), causing the renderer to fall back instead of properly rendering the widget content. This affects any async workflow with Output widgets.

Note: This breaks the official recommended pattern for ipywidgets async usage as documented in:

• Output Widget documentation
• GitHub issue #3993

Using append_display_data() in async contexts is the standard way to update widgets from background threads and show progress in long-running operations.

Logs

Output for Jupyter in the Output panel (View→Output, change the drop-down the upper-right of the Output panel to Jupyter)

``` nothing relevant ```

Browser Console Errors (from Help > Toggle Developer Tools):

Model not found in Kernel state to render output [id], rendering a fallback mime type
Uncaught (in promise) vscode.fallbackToNextRenderer
    at renderOutputItem (ipywidgetsRenderer.js)

Visual comparison:
(Recorded with a slightly larger MRE, but same problem)

• In VS Code:

Screen.Recording.2025-06-17.at.14.29.17.mov

• In Jupyter Notebook:

Screen.Recording.2025-06-17.at.14.30.05.mov

[basnijholt]

Note that this is the recommended way of using ipywidgets in a thread/async context, see https://ipywidgets.readthedocs.io/en/stable/examples/Output%20Widget.html#interacting-with-output-widgets-from-background-threads

In pipefunc/pipefunc#824 I also found a workaround:

import os

def _register_widget(widget: Widget) -> None:
    """Register widget in VS Code to work around widget rendering bug.

    This is a workaround for VS Code Jupyter notebook environment where
    widgets created and immediately used in append_display_data() without
    being displayed first don't get properly registered in the widget state.

    See: https://github.com/microsoft/vscode-jupyter/issues/16739
    """
    if os.environ.get("VSCODE_PID") is None:
        return

    from IPython.display import display
    from ipywidgets import Output

    with Output():
        display(widget)

import asyncio
from IPython.display import display
from ipywidgets import HTML, Output

outputs = [Output() for _ in range(10)]
display(*outputs)

async def reproduce_bug():
    for i in range(10):
        await asyncio.sleep(0.01)
        w = HTML(f"Content {i}")
        _register_widget(w)  # THIS
        outputs[i].append_display_data(w)

asyncio.create_task(reproduce_bug())

@DonJayamanne, I guess this might suggest a solution to you?

[DonJayamanne]

This doesn't work even with synchronous code

• Cell 1

import asyncio
from IPython.display import display
from ipywidgets import HTML, Output

outputs = [Output() for _ in range(2)]
display(*outputs)

• Cell 2

outputs[0].append_display_data(HTML("Content 0"))

@basnijholt thanks for the details, however given this doesn't work even with sync code, the problem could be different (or not).
Either way, first we'd need to resolve the issue with sync code, then move to async issue (if it still persists after resolving the sync issue).

[basnijholt]

Oh, I didn't realize that! Yeah, likely the async problem will be fixed if the sync problem is fixed.

[basnijholt]

Not sure if helpful, but in Cursor the problem doesn't exist, even though it is using the same extension.