Module site 

The site’s registry dataset: what GitHub Pages renders about the executable, derived from the registry, the index, the command line, the benchmark projection and the accepted baselines, and nothing else. The site has no content model of its own for any of it; its templates read site/data/registry/registry.json and lay it out, and the shell site generator reads docs/generated/registry.json for the ids it turns into routes.

Deterministic: same repository tree and same executable, same bytes. No timestamps of its own (a baseline’s finished_at is observed evidence, copied as committed), no absolute paths, no git state (the site’s source.json carries the commit; it is written at build time by the site generator and is not compared by generate --check).

Structs

BaselineView

One accepted baseline.

BenchmarksView

The benchmark system as the site shows it.

CapabilityView

One builtin capability: the canonical descriptor and where its source is.

Generator

The generator’s identity.

HttpView

The HTTP surface as the registry projects it.

IndexView

The index as the site shows it.

KindView

One kind.

McpView

The MCP surface as the registry projects it.

MeasurementView

One measurement of a baseline.

ModuleView

One module.

ObjectView

One object of the index, without its content.

ProjectionView

One declared provider projection.

ReferencePage

A page of the crate’s reference: the web surface that publishes it and the page within it, as the rustdoc check derives it from the crate’s own inventory.

RegistryView

The capability registry as the site shows it.

ResourceView

One MCP resource a builtin capability answers.

RouteView

One HTTP route.

SiteRegistry

The whole dataset.

SystemTargetView

One system target.

TargetView

One benchmark target.

ToolView

One MCP tool.

Constants

EVIDENCE_PRODUCER

Where the site’s evidence verdicts come from, named in the dataset beside them.

GRAPH_SOURCE

Where why-graph.json says it came from.

MCP_RESOURCE_TEMPLATE

The URI shape of a declarative object’s resource, as every surface that describes the MCP projection states it.

PRODUCT_SCHEMA

The dataset’s own format version.

PRODUCT_SOURCE

Where product.json says it came from.

PUBLIC_DOMAIN_FIELDS

The fields of one domain the public dataset carries: an allow-list, for the reason PUBLIC_FEATURE_FIELDS is one. The body stays in the file the record names in source, as a feature’s does.

PUBLIC_FEATURE_FIELDS

The fields of one feature the public dataset carries. An allow-list, not an exclusion: a field added to the resolved view reaches the published site only when it is named here, so a value that names a machine or an internal state cannot arrive by accident.

SCHEMA

The dataset’s own format version, distinct from the crate’s.

WHY_SCHEMA

The dataset’s own format version.

WHY_SOURCE

Where why.json says it came from.

Functions

allowed

The fields of v an allow-list names, and nothing else: what reaches the published site is what was named, not what was not excluded.

claim_evidence

The evidence verdict of every claim, as the site renders it beside the declared status: the evidence section of product.json.

dataset

Build the dataset.

product_artifacts

The product model as the site’s templates read it: site/data/registry/product.json.

render

The dataset as the committed file: pretty JSON, trailing newline.

why_artifacts

The Why catalogue and its graph, as the site’s templates read them: site/data/registry/why.json and site/data/registry/why-graph.json.