In a notebook

Every object worth looking at renders itself as HTML in Jupyter, JupyterLab, VS Code notebooks, Marimo and anything else that speaks the IPython display protocol. There is nothing to import and nothing to enable.

import xsdkit

schemas = xsdkit.SchemaSet.from_file("report.xsd")
schemas                     # renders as a table, not as <SchemaSet object at 0x…>

Put this in a cell	And you get
schemas	Its documents and globals at a glance
element	The element's shape, a couple of levels deep
element.tree()	The full tree, colour-coded, as deep as you asked
type	Name, variety or content kind, base
type.facets	The constraints in force, as a table
schemas.validate(xml)	A summary line and a table of what was found
diagnostic	The message, coloured by severity, with its span and help

tree() returns something you can look at

import xsdkit

schemas = xsdkit.SchemaSet.from_file("report.xsd")
report = schemas["{urn:example}report"]

report.tree()

A plain str would have been the wrong type here, and the reason is worth knowing because it bites everyone once. A notebook displays repr() of the last expression, and repr of a string escapes every newline — so a method that returns rendered text as a str shows you

'report: {urn:example}Report\n  @id\n  title: xs:string\n  issued: xs:date\n…'

which is exactly the thing you were trying to read, made unreadable.

tree() returns a Tree: it renders as itself in a REPL, as HTML in a notebook, and as plain text through print, while still behaving as the text it is — len, in, ==, splitlines() and count() all work.

t = report.tree()
"price" in t             # True
len(t.splitlines())      # 10
print(t)                 # plain text, for a terminal or a log

Pass a depth when a schema is deep and you only want the top:

report.tree(depth=2)

Recursion stops where the shape starts repeating, so a self-referential schema prints instead of hanging.

The renderings follow your theme

Colours come from JupyterLab's own CSS variables — --jp-content-font-color1, --jp-code-font-family, --jp-mirror-editor-def-color and friends — with literal fallbacks for hosts that do not define them. Switch JupyterLab to dark mode and the trees, tables and diagnostics follow, because they are not carrying a hardcoded palette that assumes a white page.

A session that reads well

import xsdkit

schemas = xsdkit.SchemaSet.from_file("report.xsd")
schemas

report = schemas["{urn:example}report"]
report.tree()

report["item"]["price"].type.facets

schemas.validate(open("report.xml").read())

Each cell displays something readable, so exploring a schema you have never seen is a matter of subscripting and looking, rather than printing dictionaries of ids.

Outside a notebook

The same objects print sensibly in a terminal. str() and print() give the plain-text form of everything above, and diagnostics print in the compiler-style block shown in Diagnostics.

There is also an inspector for the command line:

cargo run --example inspect -- schemas/report.xsd --lax