HMIF 1.1 specification

A minimal, versioned machine interchange format for Hieropedia registry records, citations, claim maps, evidence dossiers, and preservation pointers.

Stable entry points

Scope

HMIF represents Hieropedia editorial records. It does not assert machine consciousness, belief, legal authorship, autonomous agency, or independent religious authority.

Current publication scope

The current site publishes English and its AI-readable edition. Publication scope identifies the available editions. Retained translation-status 1.0 data records source translation history, not current public availability. Other language sources remain in Git history.

Policy directories now use format 1.1 and list only published human editions. The frozen 1.0 policy schema required six links and remains a historical contract; consumers must use the 1.1 schema for current policy directories. Article records and the dataset also use HMIF 1.1: human-edition lists now reflect publication scope rather than requiring six languages. The frozen 1.0 record and policy schemas remain available; strict 1.0 consumers must use the current 1.1 schemas for this release. The JSON-LD context, source relationships and citation 1.2 format are unchanged.

Versioning

The current HMIF record format version is 1.1; citation, claim-map and aggregate exports use 1.2. Published article records identify their article version, review date, human-edition translation versions, citation metadata, and claim map.

Held records: 3 October 2026 correction

Records with status: stub remain held and outside the published-article dataset. Their version and review fields preserve canonical unknowns as null or absence; unrecorded publication and revision dates and unavailable citation and claim-map links are omitted. Human-edition rows contain only language and links to the held pages, not article translation receipts.

This corrects earlier default values and unavailable links under HMIF 1.0. Consumers must distinguish held records from published articles; the corrected stub shape is not fully backward-compatible. The published-article schema remains strict.

Policy and site-information directories: 3 October 2026 clarification

HieropediaPolicyRecord describes a directory of currently published human-edition links, not the actual policy or site-information text. Its eight fields identify the representation format, local page key, directory name and description, editorial source language, and human links. The separate policy-directory schema documents this HMIF 1.1 policy-directory shape; registry-record validators remain strict and must not be used for these descriptors.

The human pages own their actual text. A descriptor has a retrievable resource URL but no named record identity, policy version, body or review date. A link does not imply a translation review receipt. A release snapshots the directory and human pages without establishing a policy effective date. The published-article dataset excludes these descriptors and held records.

The context is unchanged: name maps to schema:name, while policy description uses the project description predicate rather than the registry summary mapping to schema:description. Nested language and URL values remain project-predicate literals, not named RDF links. The definitions below explain existing mappings without claiming complete vocabulary coverage.

Registry status records publication state; subjectState concerns the documented subject. evidenceStatus carries editorial evidence text or null, distinct from the sidecar-derived evidence-strength model. These fields do not share an interchangeable scale or change the values recorded for a subject.

External sources and internal editorial reviews

In sources, external source records include a url. Records with kind: "internal-review" instead carry access: "non-public", retain their sourceId and label, and have no url. They identify internal editorial review, not independent external corroboration. These current records omit private audit paths and contents.

The canonical registry uses schema 0.3. The same source shapes are carried in the published edition registries and HMIF. Strict registry 0.2 consumers reject the URL-less internal rows. The current HMIF 1.1 retains the same JSON-LD context and source-row definitions. Clients that assume every source has a URL must adapt.

Consumers must check the record kind before treating a source as a link. Stable identifiers do not establish public availability, independent support or local custody of external content. This distinction preserves existing source identifiers and evidence assessments.

Claim-pointer target URLs

Existing machine claim pointers use absolute public URLs. Editorial-record targets retain the article’s human edition; evidence-dossier targets remain at the public root, and external-source URLs are unchanged. Alias downloads retain their canonical record and edition targets.

This corrects target strings without changing pointer relationships or claim identifiers. Consumers must accept absolute targets rather than prepend another base URL.

Bibliography references introduced in format 1.1

Citation and claim-map format 1.1 introduced optional sourceReferences rows for bibliography entries explicitly cited by a claim block. Each row gives the actual reference ID, its owning human-edition fragment URL, and the entry’s explicit HTTP(S) source links; a reference without such a link has an empty source-URL array. Existing source pointers are preserved, including editorial notes and dossiers.

These rows describe citation syntax, not verified support for every claim, source independence, or source availability. An absent field in an older export does not establish that its prose had no citations. Strict 1.0 validators reject those added fields. The frozen 1.1 schema documents that migration; current 1.2 preserves the same claim and source-reference definitions and adds the release context described below. HMIF record, article and translation versions are unchanged.

Release-scoped block identities: citation format 1.2

Every current citation and claim map contains a release object with exactly releaseId, manifestUrl, citationsUrl and publicRecordBundleUrl. The aggregate and all its citation rows carry the same object. Its URLs identify the selected immutable release using that release’s delivery policy. Metadata without a delivery field uses https://hieropedia.org/data/releases/. A release may explicitly select the public artifact repository misterspin/hieropedia-releases, with download URLs under https://github.com/misterspin/hieropedia-releases/releases/download/, followed by the exact release identifier and filename. Temporary redirect URLs are not identifiers. The 1.2 schema and its four-field release object remain unchanged. Only the current release is provided by this site; a historical locator does not guarantee a retrievable resource. Standalone maps also carry the citation’s existing article and translation versions.

A released block is identified by (release.citationsUrl, articleId, locale, segmentId). Registry article identifiers remain stable, but segmentId is a positional label and claimId remains articleId + ':' + segmentId. Insertion, removal or reordering can change a bare block ID’s referent. Separate editions have separate identities; matching IDs or text hashes do not establish translation equivalence. Rows can represent headings and reference lists as well as prose, so they are not necessarily atomic factual propositions.

To resolve a block, verify the selected manifest’s release ID and the immutable citation aggregate’s exact URL, digest and byte count. Require unique article/locale and claim/segment keys, consistent release objects and the exact claim-ID concatenation before selecting a row. Preserve the exact release spelling: .02 and .2 are distinct. A canonical live article URL and an article version alone do not identify a historical block.

For rendered context in a downloaded release, read its sibling bundle.json, verify the bundle URL, digest and size, and inspect the unique canonical map and HTML members and existing anchor. Reject duplicate, ambiguous, traversal or linked archive members. Article files are inside the bundle; this contract does not introduce individually served per-release article URLs. A retained personal copy may provide historical context, but the site does not provide earlier full-site snapshots.

Format 1.2 keeps claim rows, source relationships, human text, anchors and article/translation versions unchanged. Frozen 1.0 and 1.1 schemas remain unchanged; the current site also publishes the frozen 1.1 schema. Strict older schemas reject the added envelope fields. Release identity records a publication snapshot, not an editorial review, translation receipt, source-availability finding or proof of factual support, independence, authorship or authority.

Vocabulary

HieropediaRecord: A versioned machine-readable representation of one Hieropedia registry record.
hmifVersion: Version of the Hieropedia Machine Interchange Format.
recordId: Stable Hieropedia registry identifier.
humanEditions: Human-edition links for registry records and policy/site-information directory descriptors; published article records also include actual translation state and version metadata.
citationMetadata: Resolvable citation metadata for the canonical English article.
claimMap: Resolvable claim-block map with source or dossier pointers.
evidenceDossier: Resolvable public evidence dossier when one exists.
sourcePreservation: Resolvable source-preservation register when one exists.
publishedOnOrBefore: Recorded upper bound for publication of the Hieropedia editorial record, not its exact first publication date or a documented subject’s event date.
lastReviewed: Recorded editorial review date of the Hieropedia record, distinct from revision, retrieval, generation and deployment dates. A successful build does not establish a new review.
lastRevised: Latest dated canonical revision row for the current article version of the Hieropedia record; absent when no such revision is recorded.
HieropediaPolicyRecord: A directory descriptor linking to human-language policy or site-information pages; it is not their text and asserts no policy version or review date.
HieropediaDataset: The HMIF dataset of published Hieropedia registry articles, excluding held records and policy/site-information directory descriptors.
pageKey: Project-local discriminator for a policy or site-information directory page, not a registry identifier or named RDF identity.
description: A descriptive literal in the project namespace; for policy/site-information directories it is a short directory description, not the linked page text.
sourceLanguage: The recorded editorial source-language literal; for policy/site-information directories it is English and does not establish translation review receipts.
language: Language-code literal used in nested edition rows, including policy/site-information directory links.
url: URL-string literal used in nested edition rows, including policy/site-information directory links; this existing mapping does not coerce the value to an RDF named link.
objectType: Editorial object-type classification carried from the registry record's type field, such as concepts, institutions or people. It is distinct from the JSON-LD representation type in @type.
status: Retained bibliographic state: article or stub. When presentationState is present, this describes the historical publication and does not count a current standalone article or planned research entry.
attributes: Controlled topic tags carried from the record's tags array. These are distinct from attributes in an editorial sidecar.
relations: Related-entry rows with an edition-appropriate relation label and a target URL string. The current JSON-LD context retains target values as literals rather than named RDF links.
taxonomy: The structured taxonomy object carried by the record, using Hieropedia's existing classification fields and values. Its classifications and derivation rules remain distinct from the editorial sidecar's status and evidence-profile model. Machine-role labels describe documented computational systems and functions relevant to an entry. They may refer to an explicitly identified example or comparison in a concept article; they do not identify the concept or a human subject as a machine or apply to every case discussed. An AI theme, an imagined system, or ordinary online hosting alone does not justify a role label. An empty role list means no role is assigned on the available evidence, not that machine involvement has been disproved. These labels do not establish consciousness, belief or autonomous agency. For mediating-system, the documented function must materially shape an interaction, interpretation, preservation or transmission process; this role can coexist with more specific computational roles. Term-origin labels distinguish externally attested or descriptive terminology from an expressly identified Hieropedia analytical framework. The value external-or-descriptive includes descriptive editorial vocabulary, including a term formulated within Hieropedia; it does not by itself establish outside coinage or academic recognition. Editorial use of a descriptive concept does not by itself require the hieropedia-analytical-framework value.
subjectState: Recorded state of the documented subject, such as active, inactive, defunct, uncertain, historical or unassessed. It is distinct from the Hieropedia record's article-or-stub publication state.
asOf: Date as of which subjectState describes the documented subject. It is distinct from publication, editorial review, revision and build dates.
provenanceRelations: Provenance relation rows carried from the edition, with a controlled type and a target URL string. The current JSON-LD context does not coerce those target strings to named RDF links.
sources: Structured source records carried by the edition. External records include identifiers, kinds, labels and URLs; internal-review records retain identifiers and labels, mark access as non-public and contain no URL. Internal reviews record editorial provenance, not independent external corroboration. Listing a source does not establish its independence, public availability or local custody of its content. Source kinds are coarse groupings, separate from judgments of independence. The compound kind independent-institutional-or-scholarly-source covers independent reporting, institutional publications or scholarly work. It can include an institution’s or product publisher’s account of its own work; institutional or scholarly status does not by itself make that account independent confirmation. Independence is assessed for the claim being supported. The kind label does not determine the independent-source count or replace the source limitations.
evidenceStatus: Narrative evidence-status text carried by the registry edition, or null when not supplied. It is distinct from the evidenceStrength value derived from an editorial sidecar.
classificationConfidence: Recorded confidence category for the editorial classification: high, moderate, low or unassessed, or null when not supplied. It is not a numerical probability or the sidecar-derived evidence-strength rating.
classificationRationale: Narrative explanation carried by the edition for Hieropedia's classification of the subject, or null when not supplied.
sourceLimitations: Narrative description carried by the edition of the sources' evidentiary limits and unresolved gaps, or null when not supplied.
articleVersion: Version label for the Hieropedia article. Published machine editions use the canonical article version, with the existing 1.0 fallback; held stubs retain canonical null or absence. It is distinct from the HMIF format version and the subject's own versions.
machineEdition: URL string for this record's machine-readable ai.jsonld resource. The current JSON-LD context treats it as a literal rather than a named RDF link.
editorialDisclosures: Disclosure text carried from the edition when present and nonempty, for relevant editorial relationships or involvement. Omission does not establish that no relationship or conflict exists.
sourceProfileNotice: Conditional sourcing-transparency object with a kind, label and summary, derived for documentary articles from record type, editorial evidence-profile data, flags and source-limitation wording. Its absence is not a finding of source independence.
presentationState: English presentation state, separate from the retained bibliographic article/stub status. Consolidated identities retain historical publications with current exposition in another article section. Administrative reservations link to About and are outside the research catalogue. Neither state asserts identity with the destination. Historical-document marks an earlier publication at its original route after the continuing identity acquired a different current route.
narrativeLocation: Named link to the current English narrative or administrative About page. It does not identify the subject and is not a sameAs relation.
historicalPublication: Named link to the versioned manifest preserving the exact prior source, published resources and document-qualified citation meanings.
historicalAssessmentOn: Historical assessment date of the preserved record, or null for an unreviewed reserved identifier. Consolidation does not renew that review.
priorPublications: Named links to version-qualified prior publication manifests for this subject identity. Historical claims retain their original document and release meanings; their presence does not renew an assessment.