Minify HTML code and measure the size reduction

It also checks that ordinary elements close in the correct order, while recognizing standard HTML void elements and explicit self-closing syntax. Every successful result includes compact HTML, original and minified UTF-8 byte counts, the percentage saved, and the number of comments removed. The operation is deterministic, requires no network connection, and provides a conservative alternative to aggressive markup rewriting.

Remove formatting without erasing page content

Readable HTML commonly contains indentation, blank lines, and explanatory comments that help authors maintain a document but add bytes to the delivered asset. This capability removes ordinary HTML comments and deletes whitespace-only text nodes when they occur between two markup tags. It deliberately preserves text containing any non-whitespace character, because a space inside visible prose can separate words and may be meaningful to layout or accessibility. Attribute values remain unchanged, including quoted values that contain angle brackets, spaces, comment-like sequences, or embedded data. The scanner reads each complete tag while respecting quotes instead of stopping at the first greater-than character it sees. Script and style bodies receive special treatment: their content is copied exactly until the corresponding closing tag, so comparison operators, strings, CSS selectors, and comment syntax inside those bodies are not mistaken for HTML markup. The returned html field is suitable for writing to a deployment artifact, while the readable source can remain in version control. This conservative scope avoids attribute reordering, text normalization, entity decoding, tag omission, or other transformations that might change browser behavior.

Catch incomplete or incorrectly nested markup

Minification should never disguise a truncated template as a successful production file. As it scans the document, the capability records every ordinary opening element and requires closing elements to appear in reverse nesting order. An opening div followed by an opening span must therefore close the span before the div. An unexpected closing tag, an unfinished tag, an unfinished comment, or an element still open at the end produces an invalid-input error instead of partial output. Standard void elements such as img, input, meta, link, br, hr, source, and wbr do not require closing tags. Explicit self-closing syntax is also accepted for any named element, which supports XML-shaped fragments and component output that uses a trailing slash. This is intentional structural validation rather than a complete conformance checker for the HTML specification. In particular, the algorithm does not infer optional end tags or repair browser-tolerated markup. If a source relies on omitted closing tags for elements such as paragraphs or list items, close them explicitly before submitting it. The strict rule makes malformed template output easier to detect and keeps results predictable across automation environments.

Measure savings in a repeatable publishing workflow

The response measures both the submitted source and compact result as UTF-8 bytes, which reflects the storage and transfer size of ordinary HTML assets more accurately than a JavaScript character count. This matters when markup contains accented text, non-Latin scripts, emoji, or other characters encoded with multiple bytes. reduction_percent is calculated from those exact byte totals and rounded to two decimal places, while comments_removed reports how many complete HTML comments were deleted. A static-site generator can pass each rendered page through this capability, store the returned html value, and record the measurements in its build report. A template pipeline can use structural errors to stop publication when generated markup is cut off or incorrectly nested. Because processing is pure and deterministic, identical input always yields identical output and metrics, without network calls, external dependencies, randomness, or time-sensitive behavior. Each automated API request costs $0.002. Keep in mind that this operation does not compress transport data, bundle linked assets, optimize inline JavaScript or CSS, or rewrite document semantics. Server compression such as Brotli can still be applied afterward for additional transfer savings.