Puck.uno HTML validator and tidying service GitHub issue

• • Purpose
• • Where it fits
• • Why a service rather than embedded
• • Possible API shape (sketch)
• • Open
• • Status

{"vibecode": {
    "doc": "html-validator-service",
    "role": "speculative puck.uno-hosted service offering HTML5 validation and tidying as a network service; positioned as the heavy-weight authority complementing Uma's local schema-driven first pass",
    "key_concepts": ["html5_validator_service", "html_tidy", "uma_complement",
        "service_not_embedded", "whatwg_spec_check"],
    "status": "brainstorm",
    "related": "markie may absorb this functionality — see [markie.md](../markie/index.md)"
}}

Possible consolidation: Markie is a related puck.uno-hosted service idea that could grow to include HTML validation and tidying. Worth deciding whether this remains a separate service or folds into Markie before either is built.

Purpose GitHub issue

A Puck.uno-hosted service that validates and tidies HTML documents. Two related operations:

• Validate — given an HTML document, check it against the WHATWG HTML spec. Report any structural violations (unclosed tags, invalid attribute values, content-model breaches, missing required attributes, etc.).
• Tidy — given a slightly-broken HTML document, return a cleaned-up version that's well-formed and structurally valid. The HTML version of "format my code, please."

Where it fits GitHub issue

• Uma (especially the puck.uno/uma/html5 subclass) is the local first-pass tool — it parses, manipulates, and serializes HTML. Uma's schema-driven approach catches many issues at construction time.
• The validator/tidy service is the heavier-weight authority. It runs against a full HTML5 implementation (probably gumbo or an equivalent server-side parser) and produces detailed reports. Useful for one-off checks on large documents, for CI pipelines that gate on validation, and for cleaning up ingested HTML from untrusted sources.

Why a service rather than embedded GitHub issue

• A full HTML5 validator is large (hundreds of K) and changes over time as the spec evolves.
• Most apps don't need it on the hot path.
• A network service is the right shape: validate occasionally, not in every request.
• Centralizes spec-version tracking — clients automatically get the current standard without per-app updates.

Possible API shape (sketch) GitHub issue

$result = %['puck.uno/html5-validator'].validate($html)
$result.valid?           # boolean
$result.violations       # array of structured violation reports
$result.tidied           # cleaned-up version (optional, configurable)

Or as a Sammy-style HTTP endpoint with a JSON request/response.

Open GitHub issue

• Hosting and access model — free tier, paid tier, on-premises?
• Spec version selection (validate against HTML5.x specifically? the living standard?).
• Tidying preferences (preserve comments? normalize whitespace? indent style?).
• Relationship to Bryton — could a bryton-html-valid xeme type wrap this for "did my page validate?" tests.

Status GitHub issue