improvement: inline audio and video in HTML export by mscolnick · Pull Request #8931 · marimo-team/marimo

mscolnick · 2026-03-30T18:51:27Z

When exporting a marimo notebook to HTML, <audio> and <video> elements with virtual file URLs (./@file/...) were not embedded as data URIs — only <img> tags were. This meant audio clips broke in exported HTML files.

Added "audio" and "video" to VIRTUAL_FILE_ALLOWED_TAGS so they are inlined just like images. Also added a 10MB size cap (MAX_VIRTUAL_FILE_INLINE_BYTES) — files exceeding this limit are skipped and should be served from the public/ folder instead.

Closes #8801 ## Summary When exporting a marimo notebook to HTML, `<audio>` and `<video>` elements with virtual file URLs (`./@file/...`) were not embedded as data URIs — only `<img>` tags were. This meant audio clips broke in exported HTML files. Added `"audio"` and `"video"` to `VIRTUAL_FILE_ALLOWED_TAGS` so they are inlined just like images. Also added a 10MB size cap (`MAX_VIRTUAL_FILE_INLINE_BYTES`) — files exceeding this limit are skipped and should be served from the `public/` folder instead.

vercel · 2026-03-30T18:52:06Z

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
marimo-docs	Ready	Preview, Comment	Mar 31, 2026 8:42pm

Copilot

Pull request overview

This PR updates the HTML export pipeline to inline virtual-file media for <audio> and <video> elements (in addition to <img>), and introduces a maximum inline size threshold intended to prevent oversized assets from being embedded.

Changes:

Expand HTML virtual-file inlining to include audio and video tags (previously img only).
Add a max_inline_bytes limit to virtual-file → data-URI replacement during DOM traversal and HTML export.
Add/extend tests covering audio/video replacement and size-limit skipping behavior.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 4 comments.

File	Description
marimo/_server/export/exporter.py	Allows `audio`/`video` inlining; adds a 10MB inline cap and threads it through export inlining + virtual-files map generation.
marimo/_convert/common/dom_traversal.py	Adds `max_inline_bytes` support to `replace_virtual_files_with_data_uris()` to skip oversized virtual files.
tests/_server/export/test_exporter.py	Adds export-level tests for audio inlining and oversized virtual-file behavior.
tests/_convert/common/test_dom_traversal.py	Adds DOM traversal tests for audio/video replacement and max-inline-bytes skipping.

Copilot · 2026-03-30T18:57:43Z

marimo/_server/export/exporter.py

+VIRTUAL_FILE_ALLOWED_TAGS = {"img", "audio", "video"}
+# Maximum file size to inline as a data URI in exported HTML (10 MB).
+# Files exceeding this are skipped and should be served from the public/ folder.
+MAX_VIRTUAL_FILE_INLINE_BYTES = 10 * 1024 * 1024


The comment says files over the limit "should be served from the public/ folder", but the exporter currently just skips inlining and leaves the original ./@file/... URL in the HTML. In a static/offline HTML export there’s no handler for ./@file/ URLs, so skipped media will still be broken unless the URL is rewritten (e.g., to ./public/) or the export surfaces a user-visible warning/error when skipping.

Copilot · 2026-03-30T18:57:44Z

marimo/_server/export/exporter.py

+            byte_length = int(byte_length_str)
+        except Exception as e:
+            LOGGER.warning(
+                "File not found in export: %s. Error: %s", file_url, e


The warning logged in the URL parsing block uses "File not found in export" even when the failure is actually due to an invalid virtual-file URL format or non-integer byte length. Consider changing the message to reflect parsing/validation failure (and keep the "not found" wording for actual read_virtual_file errors) to make logs actionable.

Suggested change

"File not found in export: %s. Error: %s", file_url, e

"Invalid virtual file URL in export (parsing/validation failed): %s. Error: %s",

file_url,

e,

Copilot · 2026-03-30T18:57:44Z

tests/_server/export/test_exporter.py

+    file_manager = AppFileManager.from_app(InternalApp(app))
+    cell_ids = list(file_manager.app.cell_manager.cell_ids())
+
+    # 10_000_000 bytes exceeds the 5 MB limit


The inline comment here is inconsistent with the test data and the new 10MB cap: it mentions "10_000_000 bytes" and a "5 MB limit", but the URL uses 20_000_000 bytes and MAX_VIRTUAL_FILE_INLINE_BYTES is 10MB. Please update the comment (and/or the byte length used) so the test documents the real behavior.

Suggested change

# 10_000_000 bytes exceeds the 5 MB limit

# 20_000_000 bytes (20 MB) exceeds the 10 MB inline limit

Copilot · 2026-03-30T18:57:44Z

tests/_server/export/test_exporter.py

+
+    # The large file should NOT be inlined in the HTML output
+    assert "data:audio/x-wav;base64," not in html
+    # It should fall through to the virtual_files dict instead


This assertion comment says the oversized virtual file should "fall through to the virtual_files dict", but exporter._build_virtual_files_dict now receives the same max_inline_bytes cap and will skip adding oversized entries there as well. Either adjust the test expectation/docs, or change the implementation if the dict fallback is still intended for some asset types.

Suggested change

# It should fall through to the virtual_files dict instead

# The HTML should keep a reference to the original virtual file path

Copilot

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

marimo/_convert/common/dom_traversal.py

- Fix misleading "File not found" log message for URL parsing errors - Fix incorrect test comments about size limits and fallback behavior

…port Instead of leaving a broken ./@file/ URL in static HTML when a file exceeds the inline size limit, emit a text/plain data URI with a human-readable message explaining why the file was not inlined.

… files The test expected oversized virtual files to keep their original ./@file/ URL, but commit d39e7d3 changed the behavior to emit a text/plain placeholder data URI instead. Update assertions accordingly.

mimetypes.guess_type returns audio/x-wav on macOS/Linux but audio/wav on Windows, causing test failures on Windows CI.

… tests The test_proxy_static_file test was flaky on macOS CI because time.sleep(1) was insufficient for the subprocess server to be ready. Poll the socket instead to wait for actual readiness.

mscolnick requested a review from dmadisetti as a code owner March 30, 2026 18:51

Copilot AI review requested due to automatic review settings March 30, 2026 18:51

Copilot started reviewing on behalf of mscolnick March 30, 2026 18:52 View session

mscolnick added the enhancement New feature or request label Mar 30, 2026

mscolnick changed the title ~~fix: inline audio and video in HTML export~~ improvement: inline audio and video in HTML export Mar 30, 2026

vercel bot deployed to Preview March 30, 2026 18:53 View deployment

Copilot AI reviewed Mar 30, 2026

View reviewed changes

mscolnick requested a review from Copilot March 30, 2026 19:37

Copilot started reviewing on behalf of mscolnick March 30, 2026 19:38 View session

Copilot AI reviewed Mar 30, 2026

View reviewed changes

marimo/_convert/common/dom_traversal.py Outdated Show resolved Hide resolved

mscolnick added 2 commits March 30, 2026 15:54

review: address code-review feedback

eda8ed2

- Fix misleading "File not found" log message for URL parsing errors - Fix incorrect test comments about size limits and fallback behavior

fix: emit placeholder data URI for oversized virtual files in HTML ex…

d39e7d3

…port Instead of leaving a broken ./@file/ URL in static HTML when a file exceeds the inline size limit, emit a text/plain data URI with a human-readable message explaining why the file was not inlined.

vercel bot deployed to Preview March 30, 2026 20:20 View deployment

fix: update test to match placeholder data URI behavior for oversized…

7ed9b92

… files The test expected oversized virtual files to keep their original ./@file/ URL, but commit d39e7d3 changed the behavior to emit a text/plain placeholder data URI instead. Update assertions accordingly.

vercel bot deployed to Preview March 30, 2026 21:35 View deployment

fix: accept both audio/wav and audio/x-wav MIME types in tests

58b535c

mimetypes.guess_type returns audio/x-wav on macOS/Linux but audio/wav on Windows, causing test failures on Windows CI.

vercel bot deployed to Preview March 31, 2026 15:10 View deployment

fix: replace fixed sleep with TCP readiness check in proxy middleware…

636c328

… tests The test_proxy_static_file test was flaky on macOS CI because time.sleep(1) was insufficient for the subprocess server to be ready. Poll the socket instead to wait for actual readiness.

vercel bot deployed to Preview March 31, 2026 20:42 View deployment

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

improvement: inline audio and video in HTML export#8931

improvement: inline audio and video in HTML export#8931
mscolnick wants to merge 6 commits intomainfrom
ms/embed-audio

mscolnick commented Mar 30, 2026 •

edited

Loading

Uh oh!

vercel bot commented Mar 30, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 30, 2026

Uh oh!

Copilot AI Mar 30, 2026

Uh oh!

Copilot AI Mar 30, 2026

Uh oh!

Copilot AI Mar 30, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

	# 10_000_000 bytes exceeds the 5 MB limit
	# 20_000_000 bytes (20 MB) exceeds the 10 MB inline limit

	# It should fall through to the virtual_files dict instead
	# The HTML should keep a reference to the original virtual file path

Conversation

mscolnick commented Mar 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

vercel bot commented Mar 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 30, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 30, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 30, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 30, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

mscolnick commented Mar 30, 2026 •

edited

Loading

vercel bot commented Mar 30, 2026 •

edited

Loading