[SC 9267] Expose log text interface to pass text for qualitative text sections

[AnilSorathiya]

Internal Notes for Reviewers

Expose interface to pass through text content in the VM Library that can populate tests blocks in the model documentation template.

Example:
load markdown file with model limitations

file = open("model_limitations.md", "r")
model_limitations = file.read()
file.close()

vm.log_text(
    content_id="model_limitations",
    text=model_limitations
)

• This will log data in the db and can be accessible on platform UI for any types of documents.

UX:

readme = open("../README.md", "r").read()
vm.log_text(content_id="my_readme", text=readme)

External Release Notes

Expose interface to pass through text content in the VM Library that can populate tests blocks in the model documentation template.

Example:
load markdown file with model limitations

file = open("model_limitations.md", "r")
model_limitations = file.read()
file.close()

vm.log_text(
    content_id="model_limitations",
    text=model_limitations
)

• This will log data in the db and can be accessible on platform UI for any types of documents.

UX:

readme = open("../README.md", "r").read()
vm.log_text(content_id="my_readme", text=readme)

[cachafla]

Looks good 👏 working well for me.

I'd suggest one final change to improve the UX. Currently a user sees the following since we return the metadata object:

I'd suggest we take advantage of the existing components we use to preview results to show the user a preview of the logged HTML. Since the text could be big, we could probably render just an collapsed-by-default accordion component like we do in places like this:

When the user successfully calls log_text() it would be good to see a success message and a collapsed accordion where the user can view the preview of the logged contents and check how it's formatted, etc.

[AnilSorathiya]

Looks good 👏 working well for me.

I'd suggest one final change to improve the UX. Currently a user sees the following since we return the metadata object:

I'd suggest we take advantage of the existing components we use to preview results to show the user a preview of the logged HTML. Since the text could be big, we could probably render just an collapsed-by-default accordion component like we do in places like this: When the user successfully calls `log_text()` it would be good to see a success message and a collapsed accordion where the user can view the preview of the logged contents and check how it's formatted, etc.

UX:

readme = open("../README.md", "r").read()
vm.log_text(content_id="my_readme", text=readme)

[github-actions]

PR Summary

This pull request introduces a new function log_text in the validmind/api_client.py module, which allows logging of free-form text to the ValidMind API. The text is converted to HTML with MathML support if it is not already in HTML format. The function returns an ipywidgets.Accordion widget containing the logged text as HTML. This enhancement provides a more flexible way to log textual data with associated metadata.

Additionally, a utility function is_html is added to validmind/utils.py to determine if a given string contains HTML content. This function uses a combination of heuristics, regular expressions, and BeautifulSoup parsing to make a robust determination.

The __init__.py file is updated to include the new log_text function in the module's public API.

Test Suggestions

• Test log_text with valid content_id and text to ensure it returns an Accordion widget with the correct HTML content.
• Test log_text with empty content_id and text to verify that it raises a ValueError.
• Test log_text with non-string content_id and text to ensure it raises a ValueError.
• Test is_html with a variety of HTML and non-HTML strings to ensure it correctly identifies HTML content.
• Test is_html with malformed HTML to verify it still returns True if the content is likely HTML.
• Test is_html with plain text to ensure it returns False.

[cachafla]

Awesome 🫡

[johnwalz97]

very nice 👏... this is going to be sick!