TimeBack Platform Content Alpha Data Dictionary

public read view · Content. Belongs here because this is the blank question, test, test spec, test bank, article, video, audio, image, interactive, external link, media asset, or script a student interacts with.

content_item — Content Item

The catalog row for one student-touchable thing, across versions.

Purpose

Use content_item to find and classify things students can open, read, watch, answer, take, launch, or follow. The public object is a read view: QTI-backed identity and descriptive fields come from qti.artifact, while Alpha-only facts come from extension sidecars. GET /items and item detail flatten the assessment_test fields onto test/test_spec/test_bank rows and decodable fields onto subtype=decodable article rows; the fields array below documents the base slice, while publicApiProjection is the exact union returned by publicItem.

Relationships

• Belongs to exactly one platform workspace/tenant.
• Has zero or more immutable content_version view rows over qti.artifact_version; TimeBack adopts a deployed legacy :v1 identity only when its complete immutable projection matches, while every new exact version UUIDv5 is bound to canonical source bytes and receives the next version_number.
• May join to media_asset views and extension tables: canonical content_kc_tag, frozen content_curriculum_link archive, content_effectiveness, and content_trust_evidence.
• Does not contain student answers, attempts, events, mastery, or gradebook results.

Constraints

• content_item is not a physical copied table. Downstream migrations create alpha.content_item_view and alpha.content_item_extension, not alpha.content_item with sourced QTI columns.
• The public item response conditionally flattens assessment_role, test_type, is_mastery_gate, test_family_id, phonics_constraint, constraint_set_id, and policy_version. The assessment_test and decodable dictionary objects document those conditional slices; they are not separate decodable or assessment-test endpoints.
• content_kind is exactly one of question, test, test_spec, test_bank, article, video, audio, image, interactive, media_asset, external_link, or script and must not be inferred from title, URL, package path, MIME type, or XML text at read time.
• subtype is nullable and governed: diagram only when content_kind='image'; decodable only when content_kind='article'; no subtype value can create a new content_kind.
• dok is nullable and governed by Webb DOK. It is required for question items that count toward a test_spec's kc_coverage and for each kc_coverage cell; it is never used as difficulty.
• expected_xp is required for every Content row and is student-independent. The accepted authoring policy uses speed-running or an approved calibration process, but current runtime enforces only a finite nonnegative number and import/seed paths may publish policy-derived or default values without calibration evidence. Awarded XP lives only in the Results XP ledger.
• usage_scope storage is non-null and defaults to practice_eligible. Public reads from deployments before this field existed may omit it; clients and routers must interpret omission as practice_eligible. Routine instruction, practice, next-task, and next-form routing may select only practice_eligible items; held_out_canary and secure_form remain explicitly readable but are not routine-routing candidates.
• subject_id is exactly one of math, reading, language, science, computer_science, vocabulary, writing, social_studies after write-time alias folding; target_grade_id is a canonical id. Economics, history, geography, and civics remain course-grain disciplines rather than subject ids. Neither field is a free-form display label.
• latest_version_id is advance-only and resolves default student-view reads to the highest content_version.version_number for the item. GET student-view with an exact version_id instead projects that immutable version and its matching answer-slot/component rows; test/form renderers use this exact read so a later successor cannot change a pinned launch. An already-stored historical source hash is a no-op and must not move latest_version_id backward.
• status='retired' or archived rows remain retained in storage when Events or Results still reference their versions. Current public Content item/version reads return not-found for retired items; a historical-read surface is pending rather than implied here.
• updated_at is the item-row change-polling input for modifiedSince; consumers must not use created_at as an activity feed or assume that independent facet, link, reference, trust, or effectiveness changes advance this timestamp.
• Common Cartridge bundles and imscp_v1p1 packages are generated transport output from Curriculum collections plus Content references; they are never content_kind values or stored Content rows.

Invalid examples

• A mastery value stored on content_item instead of Results.
• A student response JSON blob stored on content_item metadata.
• A KC stored as the tag 'fractions', a kc_refs facet, a CASE CFItem, or a Curriculum component instead of a weighted content_kc_tag registry relationship.
• content_kind parsed from a filename such as spring-placement-test.xml.
• A lesson, unit, chapter, playlist, question bank, library, Common Cartridge bundle, or imscp_v1p1 package stored as a Content item.

Example

{
  "content_id": "58612cab-9c46-426c-8c20-0e9f19c807c5",
  "workspace_id": "0d4ce2f4-1c42-4f3c-9f0d-03fb7f5271d3",
  "content_kind": "question",
  "subtype": null,
  "dok": "dok2",
  "title": "Linear equations checkpoint",
  "subject_id": "math",
  "target_grade_id": "7",
  "expected_xp": 12,
  "source_system": "qti",
  "source_identifier": "RESPONSE-001",
  "latest_version_id": "efcf3561-3a66-4825-9588-e792ef20c312",
  "status": "published",
  "usage_scope": "practice_eligible"
}

Queries

GET /tenants/{tenantId}/alpha/content/items?content_kind=question&subject_id=math&target_grade_id=7&limit=50

GET /tenants/{tenantId}/alpha/content/items?usage_scope=held_out_canary&content_kind=test&limit=50

GET /tenants/{tenantId}/alpha/content/items?curriculum_node_id=case:math:7.ee.b.4&modifiedSince=2026-06-01T00:00:00Z

GET /tenants/{tenantId}/alpha/content/items/{contentId}

Fields (21)

public API projection over the canonical Alpha weighted relationship table · Content. Belongs here because it states which governed units of learned knowledge one student-touchable Content item addresses and with what strength.

content_kc_tag — Content Knowledge Component Tag

The canonical weighted many-to-many Q-matrix row from Content to a governed registry KC.

Purpose

Use content_kc_tag and GET /kc-tags as the authoritative governed alignment read to select, compare, assess, teach, and analyze Content by exact registry KC identity. kc_refs is only a non-authoritative discovery hint; content_curriculum_link is only historical reconciliation evidence. Never infer alignment from Curriculum components, CASE CFItems, facet strings, or app-local Q-matrices.

Relationships

• Belongs to exactly one Content item identity in the same tenant and survives ordinary Content version changes.
• References exactly one active alpha.knowledge_component registry identity in the same tenant; it never references case.cf_item or alpha.curriculum_components.
• The native content_kc_tag_active_kc_fk makes the active-tag/active-KC rule a commit-time database invariant across concurrent Content revival and Curriculum retirement; application reads and row locks are friendly prechecks, not the sole race guarantee.
• Many Content items may reference one KC and one Content item may reference many KCs.
• GET /kc-tags is authoritative for alignment reads. kc_refs may aid catalog discovery but cannot prove alignment; the frozen curriculum-link archive cannot be merged into current answers.

Constraints

• weight is an independent relevance strength greater than 0 and at most 1. It is not a probability, does not represent learner mastery, and active weights are not required to sum to 1.
• is_primary identifies the principal KC addressed by this Content item. Zero or one active row per Content item may be primary.
• confidence is confidence in the correctness of the mapping itself from 0 through 1, not confidence that a learner has mastered the KC.
• provenance is required non-blank, well-formed Unicode text without U+0000, names the source, approval, import, or authoring process that asserted the mapping, and is at most 256 UTF-8 bytes. It is rejected rather than truncated, trimmed, coerced, or Unicode-repaired.
• The same tenant_id scopes Content, KC, and tag; unknown, cross-tenant, or retired KC identities are rejected.
• Active reads require both tag.retired_at IS NULL and knowledge_component.retired_at IS NULL, even though pending migration 020 also prevents a committed active-tag/retired-KC state.
• The relationship is item-scoped and version-invariant. It applies across ordinary Content versions; version-specific alignment requires a distinct Content identity.

Invalid examples

• kc_id points to case.cf_item or alpha.curriculum_components instead of alpha.knowledge_component.
• weight=70 to mean 70 percent; the stored value is 0.70.
• Three active rows for one Content item all claim is_primary=true.
• confidence stores a learner mastery probability rather than confidence in the tag assertion.
• A kc_refs facet string is treated as a substitute for this governed relationship.

Example

{
  "tenant_id": "0d4ce2f4-1c42-4f3c-9f0d-03fb7f5271d3",
  "content_id": "58612cab-9c46-426c-8c20-0e9f19c807c5",
  "kc_id": "001c6d5d-b370-559a-93d4-3483845ad2f2",
  "weight": 0.8,
  "is_primary": true,
  "provenance": "incept:course-approval:linear-equations-v3",
  "confidence": 0.95,
  "created_at": "2026-07-16T14:00:00Z",
  "updated_at": "2026-07-16T14:00:00Z",
  "retired_at": null
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/kc-tags

POST /tenants/{tenantId}/alpha/content/items/{contentId}/kc-tags with Idempotency-Key

DELETE /tenants/{tenantId}/alpha/content/items/{contentId}/kc-tags/{kcId} with Idempotency-Key

Public API fields (10)

public read view · Content. Belongs here because immutable authoring and TimeBack-import versions identify the exact blank version a student saw or an author edited; answers and scores still live in Results.

content_version — Content Version

One immutable exact edition of a content item.

Purpose

Use content_version when you need the exact student view, authoring draft, or QTI XML/projection that existed at a point in time. Launches, trust-based release reads, references, Events, and Results bind to exact versions rather than mutable content_item alone.

Relationships

• Belongs to exactly one content_item.
• Any exact version, including a TimeBack-import version, may be referenced by trust-based release reads, Content relationships, Events, and Results to identify the reviewed or served bytes.
• Parent of answer_slot, answer_key, and scoring_plan for QTI-backed questions/tests.

Constraints

• Unique (content_id, version_number).
• Every stored version is immutable. A database trigger allows an identical no-op UPDATE but rejects any changed qti.artifact_version UPDATE; callers append instead.
• student_view must not include teacher-only answer keys.
• editing_draft requires author:content and may include source-trace needed to preserve QTI authoring fidelity. A governed PP100 QTI import writes the validated exact item_family_id and rigor_calibration under editing_draft.sourceTrace and writes its matching DOK on content_item; malformed, duplicate, DOK/tier-mismatched, or sub-standard hard authority is rejected before mutation.
• For content_kind=question, submitted canonical_xml must be a complete QTI assessment-item document; its choices, response declarations, correct responses, mappings, and response-processing projection are regenerated under the new version identity in the same transaction.
• A prompt-only question version that omits canonical_xml must preserve the most recent canonical grading structure while assigning the carried-forward projection to the new version identity.
• Version ETags derive from version_number plus immutable content_hash evidence, not from a mutable timestamp.

Invalid examples

• Updating any stored student_view, authoring body, canonical XML, hash, ordinal, or supersession link in place.
• Building editing_draft from student_view and losing authoring fields.
• Putting a student's submitted answers in student_view.

Example

{
  "version_id": "efcf3561-3a66-4825-9588-e792ef20c312",
  "content_id": "58612cab-9c46-426c-8c20-0e9f19c807c5",
  "version_number": 1,
  "student_view": {
    "kind": "question",
    "identifier": "RESPONSE-001"
  },
  "etag": "W/\"content:58612cab:v1:sha256-e8a86d\"",
  "created_by": "system:package-ingest"
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/versions

GET /tenants/{tenantId}/alpha/content/items/{contentId}/versions/{versionId}

POST /tenants/{tenantId}/alpha/content/items/{contentId}/versions with Idempotency-Key; for a question, canonical_xml atomically regenerates version-scoped grading projections

PATCH /tenants/{tenantId}/alpha/content/items/{contentId} with If-Match: W/"content:..."

Fields (11)

virtual value-space dictionary · Content. Belongs here because the tokens define what stored Content HTML bodies mean before any Curriculum member selects a renderer.

content_html_body_contract — Content HTML Body Markup Contract

Governed tb-* HTML vocabulary for student-safe Content bodies.

Purpose

Use this contract when a Content version stores instructional HTML that the registered instructional renderer will display. The contract mirrors the owner-provided Instructional content renderer vocabulary into the Content dictionary so renderer docs are consumer guides, not a competing source of truth.

Relationships

• content_version.student_view may carry HTML using these classes/attributes.
• Curriculum course_component_member.target_ref selects the Content row; Curriculum course_component_member.renderer_ref selects the registered renderer.
• tb-qti-assessment-item points to tb-qti-config entries, but QTI XML and scoring semantics remain in QTI-owned objects.
• Durable learner progress is reported as Caliper events through Events/Results, not stored in Content, host app local state, or renderer private state.

Constraints

• Only the listed tb-* classes and data-catalog-idref attribute are governed by this contract; ordinary presentational classes carry no Platform3 business semantics.
• Article blocks are renderable only when they carry nonempty text or exact type=html with nonempty HTML; HTML blocks use the same centralized first-party player sanitizer as top-level student_view.html.
• data-catalog-idref must resolve to an id in the same hidden tb-catalog body.
• tb-qti-config entries may link to QTI XML, but response processing, scoring, outcomes, and answer semantics are QTI-owned.
• tb-applet-config entries may link to an application manifest; application code is outside platform3.
• The renderer/host may exchange CONTENT_READY, CONTENT_LOAD, CONTENT_STATE_CHANGE, CONTENT_RESIZE, and TTS messages, but must not become the durable progress store.

Invalid examples

• Treating a renderer README as the only definition of stored Content HTML semantics.
• Putting gate thresholds, attempts, scores, or mastery state into tb-catalog or renderer contentState.
• Using data-catalog-idref to reference a global/private catalog outside the Content body.
• Copying QTI outcome processing into Content HTML instead of referencing QTI-owned XML.

Example

{
  "html": "<div class=\"tb-article-container\"><section class=\"tb-segment\"><div class=\"tb-interaction tb-qti-assessment-item\" data-catalog-idref=\"qti-1\"></div></section><div class=\"tb-catalog\" hidden><div id=\"qti-1\" class=\"tb-qti-config\"><a href=\"/qti/items/qti-1.xml\" type=\"application/xml\">QTI item</a></div></div></div>",
  "renderer_ref": "renderer:interactive-lesson",
  "iframe_url": "https://platform3-andymontgomery-9773s-projects.vercel.app/player",
  "versioned_iframe_url": "https://platform3-andymontgomery-9773s-projects.vercel.app/player/versions/2026-08-05.issue-2501-article-html/",
  "player_version": "2026-08-05.issue-2501-article-html",
  "capability_matrix_url": "https://platform3-andymontgomery-9773s-projects.vercel.app/player/renderer-capabilities.json",
  "registered_for_routing": true,
  "verified_rendering": true,
  "rendering_state": "verified_partial",
  "content_load_fields": [
    "html",
    "qtiXml",
    "testForm",
    "title",
    "contentId",
    "contentRef",
    "gateComponentId",
    "studentId",
    "interactiveComposition",
    "contentState",
    "config"
  ],
  "host_messages": [
    "CONTENT_LOAD",
    "TTS_GENERATE_RESPONSE"
  ],
  "renderer_messages": [
    "CONTENT_READY",
    "CONTENT_STATE_CHANGE",
    "CONTENT_RESIZE",
    "CONTENT_ERROR",
    "TTS_GENERATE_REQUEST"
  ],
  "preview_affordances": [
    "ad_hoc_html",
    "ad_hoc_qti_xml_a1_text_structural",
    "exact_test_form_fixture"
  ]
}

Queries

{"name":"renderer_launch_contract","sql":"Host selects Curriculum renderer_ref, separately checks verified_rendering and rendering_state against https://platform3-andymontgomery-9773s-projects.vercel.app/player/renderer-capabilities.json, and loads https://platform3-andymontgomery-9773s-projects.vercel.app/player (or pins https://platform3-andymontgomery-9773s-projects.vercel.app/player/versions/2026-08-05.issue-2501-article-html/ for automated QC). Article/interactive hosts post CONTENT_LOAD { html, contentId, contentState?, config? } or ad hoc preview { qtiXml, title?, contentId?, config? }. Exact governed articles may also carry type=html blocks; the player joins those blocks into the same centralized sanitized lesson corridor. Exact tests post one exact Content test contentRef. Exact test_bank launches also post gateComponentId and studentId; the player delegates deterministic selection to Curriculum next-form, verifies the returned bank identity, reads the selected exact Content form's ordered embeds and exact-version student views, and renders only selected_form_ref. Durable progress flows through Caliper Events/Results. Registry acceptance alone is not rendering proof, and unsupported form-member interactions fail visibly."}

Fields (8)

public read view · Content. Belongs here because it is a file, video, audio, image, caption, transcript, or source asset a content item needs to render.

media_asset — Media Asset

Renderable or support media attached to Content.

Purpose

Use media_asset to serve and audit files behind media_asset content items and supporting files used by questions, tests, articles, interactives, and external-link previews. Stimulus text remains question anatomy rather than a public media or Content row.

Relationships

• Belongs to one content_item and optionally one content_version.
• May map to qti.package_file when the asset came from an IMS/QTI package.

Constraints

• media_type is a MIME type. media_kind is the governed public API classification.
• byte_length and content_hash must match stored bytes or external object evidence.
• Primary video/mp4 on a freshly authored Incept video must use top-level source_asset with the exact allowlisted HTTPS origin, loop-artifacts bucket, canonical incept-skill-pack/video/{run}/media/{job}.mp4 path, bare lowercase SHA-256, positive byte length at most 128 MiB, and video/mp4 media type.
• Content disables redirects and encoded responses, rejects private/local DNS answers, timeout, overflow, non-MP4 bytes, and type/length/hash mismatch, then owns and serves the verified bytes.
• Governed source_asset attach rejects caller-supplied media_id/mediaId and mints the media UUID only after byte verification, so a request cannot alias an existing qti.package_file row.
• The original Incept bucket/path remains sanitized provenance metadata; qti.package_file.package_path is the unique Content-owned content/{contentId}/media/{serverMediaId} path, so successor versions may reuse one immutable source object without colliding.
• Caller metadata.inline_base64, metadata.inline_text, metadata.inlineText, and metadata.source_asset cannot create governed primary video. The signed source URL/query token is neither persisted nor returned.
• delivery_url must not contain secret tokens. Signed delivery is handled by authenticated endpoints or short-lived platform URLs.

Invalid examples

• A permanent signed URL with embedded secret stored in delivery_url.
• A governed source_asset request that selects an existing media_id/package_file UUID.
• A transcript stored as a free-form tag instead of a media_asset row.
• byte_length not matching the stored asset.

Example

{
  "media_id": "30225f0a-949d-49d5-a7fc-e2ea43863e96",
  "content_id": "58612cab-9c46-426c-8c20-0e9f19c807c5",
  "media_kind": "supporting",
  "media_type": "image/png",
  "byte_length": 18422,
  "delivery_url": "/tenants/demo/alpha/content/items/58612cab/media/30225f0a"
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/media

POST /tenants/{tenantId}/alpha/content/items/{contentId}/media with Idempotency-Key

Public API fields (11)

Alpha extension sidecar joined into public Content views · Content. Belongs here because it is a student-touchable link to material outside the platform; the link target is blank content, not an event or result.

external_link — External Link

Alpha-owned URL facts for content items whose content_kind is external_link.

Purpose

Use external_link when a student should open a governed outside resource while Content still owns catalog status, weighted KC tags, version identity, trust evidence, and effectiveness overlays.

Relationships

• One-to-one with content_item where content_kind='external_link'.
• Participates in content_kc_tag, content_effectiveness, content_trust_evidence, and catalog facets through the same content_id as every other Content kind.

Constraints

• external_url must be an absolute http or https URL after validation.
• URL values must not contain raw JWTs, OAuth tokens, student credentials, secret_ref values, or permanent signed query strings.
• External links are referenced by content_id/content_version_id like every other Content kind; no link-kind-specific curriculum or effectiveness tables are allowed.

Invalid examples

• external_url='javascript:alert(1)' or a data: URL.
• A signed URL with a long-lived token stored as the canonical external_url.
• A separate external_link_kc_tag table instead of the uniform content_kc_tag relationship.

Example

{
  "content_id": "fa90ce89-ea47-4e87-b44b-c0b9d786caa0",
  "external_url": "https://example.edu/simulations/number-line",
  "display_url": "example.edu/simulations/number-line",
  "open_policy": "new_tab",
  "last_validated_at": "2026-06-05T13:30:00Z"
}

Queries

GET /tenants/{tenantId}/alpha/content/items?content_kind=external_link

GET /tenants/{tenantId}/alpha/content/items/{contentId}

Fields (5)

Alpha extension sidecar for an article subtype · Content. Belongs here because a decodable is a student-readable article with a phonics constraint set; it is not a new content_kind and not Curriculum sequence.

decodable — Decodable Article

Phonics-constraint sidecar for content_item rows where content_kind='article' and subtype='decodable'.

Purpose

Use decodable when an article must be rendered, searched, exported, and validated as text constrained to a specific phonics set. The article remains the one Content row; this sidecar stores only the constraint facts QTI does not define.

Relationships

• One-to-one with content_item where content_kind='article' and subtype='decodable'.
• May be referenced by Curriculum through the same content_ref contract as any other article.
• May reference images or audio through content_reference rows; it does not nest those rows.

Constraints

• content_item.content_kind must be article and content_item.subtype must be decodable.
• phonics_constraint is required JSONB and must be validated at write time against the governed phonics policy version.
• The constraint set is a Content render/discover fact; student reading attempts and scores live in Events/Results.

Invalid examples

• content_kind='decodable'.
• A phonics constraint stored as a catalog facet string instead of the decodable sidecar.
• Student reading accuracy or mastery stored in phonics_constraint.

Example

{
  "content_id": "4fd04a3e-2e51-4dd6-92d1-0ef196878fea",
  "phonics_constraint": {
    "focus_patterns": [
      "short_a",
      "short_i"
    ],
    "allowed_graphemes": [
      "a",
      "i",
      "m",
      "s",
      "t",
      "p"
    ],
    "exception_words": [
      "the"
    ]
  },
  "constraint_set_id": "alpha-phonics-k-01",
  "policy_version": "alpha.policy.decodable_constraints.v1"
}

Queries

GET /tenants/{tenantId}/alpha/content/items?content_kind=article&subtype=decodable

GET /tenants/{tenantId}/alpha/content/items/{contentId}

Fields (5)

public read view over content_item and assessment_test sidecars · Content. Belongs here because it is the plain public object for a blank renderable assessment form; blueprints are test_spec, banks are test_bank, and sittings/scores/mastery/gradebook rollups live outside Content.

test — Test

Plain public view for renderable content_item rows where content_kind='test'.

Purpose

Use test when an app-builder wants the plain Alpha object for a renderable form. Use test_spec for the blueprint and test_bank for the set of equivalent forms.

Relationships

• One-to-one with content_item where content_kind='test'.
• assessment_role='form' is a renderable test form and may be a bank member.
• test_spec and test_bank are separate first-class content_kind values, not hidden forms.

Constraints

• assessment_role is required and exactly form for content_kind='test' unless a legacy compatibility row is being read.
• test_spec and test_bank are content_kind values.
• Gate fields and cutoff numbers stay in Curriculum/Policy; Results owns attempts, scores, and mastery.

Invalid examples

• A blueprint or bank stored only as content_kind='test' with no first-class test_spec/test_bank kind.
• A bank membership list stored in a test JSON blob rather than content_reference bank_member rows.
• A student score stored on the test object.

Example

{
  "content_id": "8b41a559-98e3-4e33-9ad2-61a575f7fb3f",
  "content_kind": "test",
  "assessment_role": "form",
  "title": "Math diagnostic form A",
  "expected_xp": 30
}

Queries

GET /tenants/{tenantId}/alpha/content/items?content_kind=test&assessment_role=form

GET /tenants/{tenantId}/alpha/content/items?content_kind=test_spec

GET /tenants/{tenantId}/alpha/content/items?content_kind=test_bank

GET /tenants/{tenantId}/alpha/content/items/{contentId}/blueprint

GET /tenants/{tenantId}/alpha/content/items/{contentId}/bank-members

Fields (3)

Alpha extension sidecar joined into public test view · Content. Belongs here because this is the blank test instrument; sittings, submitted answers, scores, and gradebook rollups live outside Content.

assessment_test — Assessment Test

Typed Alpha fields for QTI-backed test, test_spec, and test_bank rows.

Purpose

Use assessment_test to classify QTI-backed assessment rows by educational purpose, external test family identity, and assessment_role before Results records arrive. Public content_kind is test, test_spec, or test_bank; assessment_role keeps form/spec/bank compatibility over shared QTI storage.

Relationships

• One-to-one with content_item where content_kind is test, test_spec, or test_bank.
• assessment_role='spec' links to alpha.test_spec_extension; assessment_role='bank' links to alpha.test_bank_extension.
• test_family_id is an intended future Results reconciliation target. Current Results ingest can persist producer family text without resolving it to Content.

Constraints

• assessment_role is exactly form, spec, or bank; content_kind=test_spec implies spec and content_kind=test_bank implies bank.
• test_type and is_mastery_gate are separate fields. test_type is optional for form, spec, and bank rows; when present it must use the governed enum.
• test_family_id is an optional opaque storage value for adapters, not a secret and not a score. The current API/database impose no lexical grammar, Content has no family resolver, TimeBack import writes null, and Results has no Content foreign key; a present string is not proof of reconciled family identity.
• Do not infer test_type from title, URL, producer, or QTI identifier at read time.
• Gate fields, passing_criteria, retake cooldown, and mastery cutoff numbers stay in Curriculum/Policy. Content owns only the blank forms, spec blueprint, and bank membership.

Invalid examples

• Setting is_mastery_gate by checking whether the URL contains mastery.
• Using placement as a display tag instead of assessment_test.test_type.
• Storing a student's test score on assessment_test.
• Creating content_kind='test_spec' with assessment_role other than spec, or content_kind='test_bank' with assessment_role other than bank.

Example

{
  "content_id": "8b41a559-98e3-4e33-9ad2-61a575f7fb3f",
  "assessment_role": "form",
  "test_type": "diagnostic",
  "is_mastery_gate": false,
  "test_family_id": "nwea:map-growth:math"
}

Queries

GET /tenants/{tenantId}/alpha/content/items?content_kind=test&test_type=diagnostic

GET /tenants/{tenantId}/alpha/content/items?content_kind=test_bank

GET /tenants/{tenantId}/alpha/content/items/{contentId}

Fields (5)

Alpha extension sidecar for a test role · Content. Belongs here because a test spec is the blank blueprint for equivalent test forms; gate decisions and student outcomes stay in Curriculum/Policy and Results.

test_spec — Test Spec

The validated table of specifications for a bank of equivalent test forms.

Purpose

Use test_spec when a mastery gate, generator, or reviewer needs to inspect the KC coverage, DOK coverage, difficulty constraints, passing rule, minimum form count, overlap limit, and enemy-item rule that define equivalent forms.

Relationships

• One-to-one with content_item where content_kind='test_spec' and assessment_test where assessment_role='spec'.
• Referenced by test_bank.spec_id.
• kc_coverage references active identities in the Curriculum-owned alpha.knowledge_component registry; Content stores ids and counts, not copied KC statements.

Constraints

• kc_coverage is a KC x DOK grid with target item counts and instructional weights. The accepted target requires every KC to resolve to an active Curriculum Knowledge Component at write time; current runtime does not yet perform that cross-module resolution.
• The accepted target requires every coverage cell and every counted item to carry one governed Webb DOK value (dok1 through dok4). Current runtime does not inspect coverage-cell contents or prove item-to-cell DOK conformance.
• passing_rule is canonical for every form generated from this spec, but the cutoff number and false-pass/false-fail loss weights stay in alpha.policy.
• min_forms is at least 1 for ordinary banks and at least 5 for grade gates; a Curriculum grade gate pointing at a bank below min_forms is a spec violation.
• Learning Scientists own the three blueprint semantics adopted here: caitd-016 item-to-KC cardinality, caitd-017 comparable-forms check, and caitd-018 passing-rule loss encoding.
• caitd-016 Item-to-KC cardinality: One primary KC per coverage-counted item; optional secondary Q-matrix tags may exist but are not double-counted in kc_coverage. Reopen trigger: If Learning Scientists revise caitd-016, this dictionary rolls back before changing coverage counting.
• caitd-017 Comparable-forms check: Two fixed forms are equivalent only when both satisfy the same blueprint constraints and the max_item_overlap rule; this is a binary same-blueprint validation, not a string label. Reopen trigger: If Learning Scientists revise caitd-017, test_bank membership validation rolls back before accepting new forms.
• caitd-018 Passing-rule loss encoding: The spec's passing_rule encodes the per-KC false-pass versus false-fail loss assumption; the cutoff number itself remains alpha.policy config. Reopen trigger: If Learning Scientists revise caitd-018, passing_rule validation rolls back before changing the schema contract.

Invalid examples

• A blueprint stored only inside editing_draft JSON with no queryable kc_coverage.
• A passing_rule that restates 89.5 or another cutoff number instead of referencing policy.
• A KC label copied into kc_coverage without a Curriculum KC id.
• DOK used as difficulty, or difficulty stored without a separate mastery-probability constraint.

Example

{
  "spec_id": "2d461e05-2148-4ec9-a4c0-eafc013cb8e7",
  "assessment_role": "spec",
  "kc_coverage": [
    {
      "curriculum_node_id": "case:math:7.ee.b.4",
      "dok": "dok2",
      "target_item_count": 4,
      "instructional_weight": 0.4
    },
    {
      "curriculum_node_id": "case:math:7.ee.b.3",
      "dok": "dok3",
      "target_item_count": 2,
      "instructional_weight": 0.2
    }
  ],
  "passing_rule": {
    "rule_ref": "alpha.policy.mastery_gate.strict_per_kc_v1"
  },
  "min_forms": 5,
  "max_item_overlap": 0
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/blueprint

GET /tenants/{tenantId}/alpha/content/items?content_kind=test_spec

Fields (10)

Alpha extension sidecar plus content_reference membership · Content. Belongs here because a test bank is a reusable blank pool of equivalent test forms; Curriculum references it but does not own its members.

test_bank — Test Bank

A stored fixed-form set or immutable adaptive question pool governed by one test_spec.

Purpose

Use test_bank when Curriculum needs one exact target_ref for a mastery gate and Content must expose either equivalent forms or the complete immutable adaptive pool. Ordered exact form shards are storage containers, not a claim that Results should select a whole form.

Relationships

• One-to-one with content_item where content_kind='test_bank' and assessment_test where assessment_role='bank'.
• Belongs to exactly one test_spec through spec_id.
• Has fixed-form members or adaptive pool shards through content_reference rows: parent_content_id=bank_id, referenced_content_id=form content id, referenced_version_id exact form version, relationship_kind='bank_member', form_index ascending.
• Adaptive/CAT pools may stay inside one adaptive QTI test through qti-selection or qti-adaptive-selection. PP100 v2 uses ordered exact form shards and the union of each shard's exact-version embeds as the frozen pool; Results owns runtime question selection and score state. No parallel membership table is created.

Constraints

• Create and PATCH require spec_id to resolve in the routed tenant to an unretired, non-archived, non-status-retired content_kind='test_spec', assessment_role='spec' item with a readable test_spec sidecar. Missing, cross-tenant, wrong-kind/role, retired, archived, or sidecar-less targets return 400 content.validation_failed before mutation.
• A fresh test_bank create resolves that spec target under the tenant graph fence before its idempotency claim. Completed/in-progress/conflicting same-key state is arbitrated first; a fresh invalid target creates neither item nor ledger row, and a corrected body may reuse the still-unclaimed key.
• PATCH requires If-Match and preserves an omitted bank-definition field from the existing sidecar. Explicit null spec_id or membership_rule returns 400 content.validation_failed. An actual definition change requires status='draft', retired_at IS NULL, and trust_status='unknown' (including missing trust); otherwise 409 content.bank_membership_frozen preserves the base item, sidecar, timestamp/ETag, and graph. With any active bank_member row—fixed form or PP100 adaptive shard—spec_id cannot change and membership_rule cannot become qti_adaptive_selection; either returns 409 content.invalid_test_bank_member_contract with the same atomic preservation.
• Create and soft-retire bank_member writes are allowed only when the parent bank has status='draft', retired_at IS NULL, and trust_status='unknown' (including a missing trust row). Published, trusted, degraded, failed, archived, retired, and draft-with-retired_at states return 409 content.bank_membership_frozen with no graph, item identity, timestamp, ETag, or fresh-idempotency-ledger mutation. Completed retries replay their prior response.
• Every new fixed-form or PP100 adaptive-shard member must pin a referenced_version_id that belongs to its assessment_role='form' test, embed at least one question in that exact form-version graph, and have at least one active canonical content_kc_tag on every ordered question occurrence in that exact graph. Parent/member role, exact-version graph, and question-attribution validation are enforced atomically at the bank_member write boundary; full same-blueprint comparability remains pending.
• The accepted target requires at least test_spec.min_forms unseen-equivalent forms before a Curriculum grade gate can target the bank. Current runtime does not enforce this release-eligibility rule.
• For same_blueprint_fixed_forms, equivalent form selection is Curriculum-owned. For pp100_adaptive_shards, Results unions the exact shard embeds, freezes that pool on the attempt, and owns deterministic runtime score-band selection. Content reads provide the ordered member list, usage_scope, referenced_usage_scope, exact referenced form version, that immutable version's ordered/multiplicity-preserving per-question canonical KC tags, and an explicit attribution-readiness audit. A missing, empty, or ambiguously mixed exact-version graph is incomplete; latest-version substitution and cross-version union are forbidden. PostgreSQL acquires the shared writer fence on client A before beginning the stable repeatable-read/serializable snapshot on client B, so a writer already waiting cannot be hidden behind a stale reader snapshot. After a fixed number of identity/metadata reads, two set queries hydrate active member facts and exact form/question/KC facts, keeping database round trips constant as graph cardinality grows. Each query uses a cap-plus-one sentinel and rejects before row mapping when one bank exceeds the inclusive ceiling of 1,024 active bank-member facts or one integrity read exceeds 4,096 joined question/KC facts. Both return 409 content.invalid_test_bank_member_contract with fieldErrors.name=bank_members or form_graph and require the graph to be split before retry. Full spec-validation evidence is a pending target. Routine selection is limited to practice_eligible members whose referenced form is also practice_eligible.
• membership_rule='qti_adaptive_selection' requires stock QTI adaptive structure and forbids parallel bank_member rows. membership_rule='pp100_adaptive_shards' requires ordered exact bank_member shards whose pinned embeds union to the immutable pool. A PP100 v2 blueprint additionally requires a minimum pool of eight unique item families spanning easy, medium, and hard, accepted tier calibration for every item, standardized-test-rigor evidence for hard items, mastery and spaced retrieval only at score 100, unseen-before-repeat, immediate-repeat avoidance, and no 90+ downward fallback unless prerequisite.

Invalid examples

• A bank with no spec_id.
• A PATCH that sets required spec_id or membership_rule to null.
• A published, trusted, retired, or otherwise non-draft bank whose definition is rewritten in place.
• A bank with active fixed-form or PP100 adaptive-shard members whose spec_id is swapped or whose membership_rule becomes qti_adaptive_selection.
• A string label saying equivalent with no blueprint validation.
• A one-form grade gate when the spec's min_forms is greater than 1.
• A local question-bank or playlist collection stored as Content instead of Curriculum.

Example

{
  "bank_id": "18f4995b-1326-45f1-b04d-4be3861da324",
  "assessment_role": "bank",
  "spec_id": "2d461e05-2148-4ec9-a4c0-eafc013cb8e7",
  "membership_rule": "same_blueprint_fixed_forms"
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/bank-members

GET /tenants/{tenantId}/alpha/content/items?content_kind=test_bank

PATCH /tenants/{tenantId}/alpha/content/items/{contentId} with If-Match for sparse spec_id or membership_rule changes

Fields (5)

public API aggregate projection over an Alpha extension composition table · Content. Belongs here because cue timing, article segment order, and embedded-question placement are instructional-delivery structure for a student-touchable Content object.

interactive_composition — Interactive Article/Video Composition

Typed segment/cue to QTI-backed question bindings for interactive articles and videos.

Purpose

Use interactive_composition when an article or video needs embedded questions at known article segments or video timestamps. The renderer reads this canonical structure instead of parsing tb-* markup for composition.

Relationships

• Belongs to one content_item and may carry optional content_version UUIDs on its physical rows. Current storage enforces independent version FKs only; validating that every version belongs to the parent item and that all active rows use one aggregate version is pending.
• interactive_video media_asset_ref references a media_asset attached to the same Content item.
• segments and cues contain question_ref objects that resolve to content_kind=question.
• Each question_ref may include an assessment_component_id that resolves through the question's assessment_component rows.
• Curriculum sequences and gates the parent Content item through its own Content references and registry-KC scope; content_kc_tag records Content-to-KC relevance, and QTI stores the referenced questions only.
• interactive_cluster is a third interactive composition kind stored on the version student_view (student_view.kind=interactive_cluster), not as interactive_composition rows: it carries cluster-owned shared_stimuli plus exact-version Content component references, and is authored via POST /items/{contentId}/versions rather than this endpoint.

Constraints

• composition_type is interactive_article or interactive_video.
• interactive_cluster (a coherent set of interdependent items — shared stimulus plus part-a/part-b components — judged at arc grain) is stored on the version student_view with student_view.kind=interactive_cluster, never as interactive_composition rows; composition_type on this table remains interactive_article or interactive_video.
• interactive_cluster components are exact-version Content reference tuples (content_id plus content_version_id), never inlined copies, so each member keeps its own exact-version eligibility answer and provenance identity.
• interactive_cluster component part_label is a non-empty string; component stimulus_ref, when supplied, resolves to a declared shared_stimuli id (shared_stimuli ids are unique).
• interactive_cluster expected_xp is the sum of component expected_xp and is never double-counted with the referenced members; the arc-grain quality judgment of a cluster is the producer's (Incept), not Content's.
• interactive_article responses contain ordered segments with positive sequence_index values and may carry segment_ref; cues are omitted.
• interactive_video responses require media_asset_ref, duration_seconds > 0, and ordered cues whose at_seconds values are >0, less than duration_seconds, and strictly increasing.
• Governed article/video publication requires every question_ref to carry canonical content_id plus exact content_version_id returned by QTI import; each exact question version must exist and be release eligible.
• Governed video publication requires media_asset_ref to name same-item, same-exact-version, primary video/mp4 bytes created by the verified Incept handoff, with matching duration and a valid caption track or nonempty transcript.
• Each segment or cue carries tests_prior_content=true to attest that the referenced question tests already-shown material.
• Optional parent and question version UUIDs currently have independent foreign keys; same-item version ownership validation is pending and must not be inferred from the FK alone.
• The HTTP body currently accepts a version_id on each segment/cue. Active-row uniqueness omits version_id, GET loads all active rows for the content_id, and the aggregate response reports the first row's version_id. Rejecting mixed parent versions or defining an explicit mixed-version contract is pending; consumers must not infer aggregate version coherence today.
• No answer keys, correctness, scores, attempts, XP thresholds, or Curriculum gate fields belong here.

Invalid examples

• Embedding video cue timings inside a QTI assessmentItem or assessmentTest.
• Putting article segment order on a Curriculum node.
• Parsing tb-qti-assessment-item positions as the canonical source after a typed composition row exists.
• Referencing a non-question Content row as question_content_id.

Example

{
  "content_id": "2f4b4c12-7c41-4b40-a33a-9637581d0e51",
  "composition_type": "interactive_video",
  "media_asset_ref": {
    "media_id": "30225f0a-51fd-4a7e-b3b2-ecfe24bca7b6"
  },
  "duration_seconds": 120,
  "cues": [
    {
      "sequence_index": 1,
      "at_seconds": 10,
      "question_ref": {
        "content_id": "58612cab-9c46-426c-8c20-0e9f19c807c5",
        "content_version_id": "efcf3561-3a66-4825-9588-e792ef20c312",
        "assessment_component_id": "51ef0fdf-7ebf-4cc7-9242-45432b04ad4f"
      },
      "tests_prior_content": true
    }
  ]
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/interactive-composition

PUT /tenants/{tenantId}/alpha/content/items/{contentId}/interactive-composition

Public API fields (9)

read-only public API projection over a frozen Alpha archive table · Content migration archive. It records how earlier Content rows were linked to CASE or Curriculum nodes before weighted registry-KC tags became canonical.

content_curriculum_link — Content Curriculum Link Archive

Frozen historical Content-to-Curriculum link evidence.

Purpose

Use content_curriculum_link only to read historical relationships during reconciliation. New lesson-to-Content routing is one Curriculum-owned use per item: currently one targeted child practice/review component per additional item, and one course_component_member target after verified cutover. New KC alignment reads and writes use content_kc_tag, which never encodes lesson membership.

Relationships

• Historically belonged to one content_item and optionally one independently constrained content_version. Same-item/tenant version validation was not universal.
• Historical curriculum_node_id values may reference CASE or Curriculum components and are not governed registry-KC identity.

Constraints

• The archive is read-only through the public API; POST returns content.curriculum_link_write_deprecated with HTTP 410.
• Existing archive rows returned through GET/items?curriculum_node_id have no automatic TTL. They remain available until a separately approved migration issue inventories dependent consumers and proves their cutover before removal.
• Archived curriculum_node_id values are historical evidence and must not be interpreted as governed kc_id values.
• A producer routes imported Content from Curriculum with one ordered use per item. The accepted target is one course_component_member per Content use. While the deployed API remains CASE-backed, a many-item lesson uses one targeted child practice/review component per item under the lesson; repeatedly PATCHing the lesson target replaces the prior direct target.
• Curriculum remains the source for membership and order. A BFF may discover ordered current targets through the deployed Curriculum lesson tree or parent-filtered component list; a thin Content reader may use GET /items?curriculum_node_id={lessonId} to project the active target identities without copying placement state.
• The Content compatibility projection starts from an exact same-tenant active component id and traverses active isChildOf descendants. Component activity means retired_at is null; the projection does not reinterpret scope_kc_refs, KC tags, or publication labels as lesson membership.
• content_kc_tag is weighted item-to-governed-KC attribution and must not be used as lesson membership or course sequence.
• New applications must not use link_kind or kc_refs facets as substitutes for content_kc_tag weight, primary role, provenance, and confidence.

Invalid examples

• Storing 'fractions' as a content tag when a Curriculum node exists.
• Copying a full standard text into Content and treating it as the graph.
• Linking to a retired Curriculum node without explicit migration evidence.

Example

{
  "content_curriculum_link_id": "e148f69b-f0c6-44c4-b9dc-c6ec50f9ccee",
  "content_id": "58612cab-9c46-426c-8c20-0e9f19c807c5",
  "curriculum_node_id": "case:math:7.ee.b.4",
  "link_kind": "assesses",
  "active_from": "2026-06-03T18:20:00Z",
  "active_until": null
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/curriculum-links (frozen archive)

POST /tenants/{tenantId}/alpha/content/items/{contentId}/curriculum-links returns HTTP 410

GET /tenants/{tenantId}/alpha/content/items?curriculum_node_id={lessonId}&content_kind=question returns the union of historical archive matches and active Content targets in that Curriculum component subtree

For one primary lesson object, GET then PATCH the Curriculum lesson target once; for multiple items, create one targeted child practice/review component per item under the lesson

GET the deployed Curriculum lesson tree or parent-filtered component list when membership order is required; after verified cutover read course_component_member rows

Public API fields (7)

Alpha extension registry table · Content. Belongs here because it defines governed discovery values used to find Content; it is not the curriculum graph and not a copy of QTI metadata.

content_catalog_facet — Content Catalog Facet

Governed catalog-facet definition for discovery metadata.

Purpose

Use content_catalog_facet to define report-only or discovery-only values such as difficulty band, format grouping, accessibility label, or authoring collection when those values are not relationship keys or computed inputs.

Relationships

• Parent of content_catalog_facet_assignment through facet_key.
• Does not reference Curriculum nodes. Knowledge coverage uses weighted content_kc_tag rows to registry KCs.

Constraints

• facet_key must be exactly one of the twelve named catalog discovery facets: subject, grade_band, language, accessibility_profile, rights_license, provenance, quality_rating, kc_refs, standards_framework_refs, content_kind, expected_xp_band, or modality.
• allowed_values is required for value_type='enum' and must be an array of governed values with plain-language labels.
• Facet definitions are not relationship targets for app joins. Promote the facet to a real column before consumers compute or join on it.
• Facet assignments are stored by authoring/import/catalog workflows and read directly by LOR/library/catalog UIs; they are never recomputed at read time from Content base rows or related Curriculum rows.

Invalid examples

• facet_key='curriculum_node_id' because current knowledge alignment uses weighted content_kc_tag relationship rows.
• Copying qti.artifact.title into a facet definition.
• A free-form bag of tags with no allowed values or grain.

Example

{
  "facet_key": "accessibility_profile",
  "label": "Accessibility profile",
  "value_type": "enum",
  "grain": "media_asset",
  "status": "active"
}

Queries

GET /tenants/{tenantId}/alpha/content/catalog-facets

GET /tenants/{tenantId}/alpha/content/items?facet=accessibility_profile:captioned

Fields (7)

Alpha extension assignment table · Content. Belongs here because it assigns governed discovery metadata to a Content item, version, or media asset; it is not a curriculum link, event, result, or copied base row.

content_catalog_facet_assignment — Content Catalog Facet Assignment

Assignment of one governed catalog facet value to Content.

Purpose

Use content_catalog_facet_assignment to drive catalog discovery filters without copying Content base fields, Curriculum statements, or QTI metadata into a parallel catalog table.

Relationships

• Belongs to one content_catalog_facet.
• Belongs to one content_item and optionally one content_version or media_asset depending on facet grain.
• Does not replace content_kc_tag, subject_id, target_grade_id, test_family_id, expected_xp, or other real operational columns.

Constraints

• facet_value must validate against content_catalog_facet.value_type and allowed_values.
• Only one active assignment may exist for the same facet_key/content_id/version_id/media_id/facet_value at the same grain.
• Assignments must not contain secrets, direct student PII, or copied QTI/Curriculum text.
• LOR/library/catalog UIs query assigned facets directly. They must not rebuild subject, language, standards, expected_xp_band, or modality by reading base rows or related Curriculum rows at request time.

Invalid examples

• Assigning a KC id only as a facet instead of a weighted content_kc_tag row.
• Using a facet assignment for subject_id or target_grade_id after those are real fields.
• facet_value contains a bearer token or copied standard statement.

Example

{
  "facet_assignment_id": "47a637af-7f0f-45a1-a042-dcad51ed2761",
  "facet_key": "accessibility_profile",
  "content_id": "58612cab-9c46-426c-8c20-0e9f19c807c5",
  "media_id": "30225f0a-949d-49d5-a7fc-e2ea43863e96",
  "facet_value": "captioned",
  "retired_at": null
}

Queries

GET /tenants/{tenantId}/alpha/content/items?facet=accessibility_profile:captioned

POST /tenants/{tenantId}/alpha/content/items/{contentId}/catalog-facets with Idempotency-Key

Fields (8)

reference contract, not a table · Content. Belongs here as the uniform reference contract other modules use to point at any Content kind; it is not a stored Content table and not a Curriculum member.

content_ref — Content Reference

The one reference shape for every Content kind.

Purpose

Use content_ref when Curriculum, LOR/library exports, or another module needs to refer to a question, test, test_spec, test_bank, article, video, audio, image, interactive, media_asset, external_link, or script without kind-specific foreign keys.

Relationships

• content_id references exactly one content_item.
• content_version_id is intended to reference zero or one version for content_id. Current concrete relationship storage enforces independent UUID foreign keys only; same-item validation is pending.
• content_kind is the discriminator used for validation and renderer routing hints; renderers must still dereference content_id.

Constraints

• Allowed for all Content kinds: question, test, test_spec, test_bank, article, video, audio, image, interactive, media_asset, external_link, and script.
• No kind-specific reference tables or fields are allowed unless a future architecture ITD supersedes the uniform contract.
• A reference to a retired item remains valid for historical Events/Results, but new Curriculum assignments should use published content.

Invalid examples

• course_component_member.question_id/test_id/article_id separate columns.
• A playlist table stored in Content instead of Curriculum referencing content_ref rows.
• A reference that stores the content title or Curriculum statement instead of ids.

Example

{
  "content_id": "58612cab-9c46-426c-8c20-0e9f19c807c5",
  "content_version_id": "efcf3561-3a66-4825-9588-e792ef20c312",
  "content_kind": "question"
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}

GET /tenants/{tenantId}/alpha/content/items/{contentId}/versions/{versionId}

Fields (3)

public API projection over an Alpha extension relationship table · Content. Belongs here because it records that one Content row embeds, depends on, captions, thumbnails, or otherwise references another Content row; it is not Curriculum sequencing and not nested Content storage.

content_reference — Content Reference Relationship

A stored relationship from one Content row to another using the uniform content_ref contract.

Purpose

Use content_reference when a content item needs another Content item for rendering, accessibility, provenance, lossless export, script lineage, or bank membership. Examples: a video points to a transcript, an article embeds an image, a question points to a supporting image while its stimulus text stays inside question anatomy, a produced video points to its script Content item, and a test_bank points to member test forms whose usage_scope controls whether routine routing may select them.

Relationships

• Is tenant-scoped by the route and JWT; the tenant key is not repeated in the public payload.
• parent_content_id references exactly one content_item.
• referenced_content_id references exactly one content_item.
• For bank_member rows, both exact version ids must be supplied and belong to their corresponding content ids. Omitted or explicit-null inputs fail closed and never resolve latest_version_id.
• Not a Curriculum sequence, playlist, question bank, lesson section, or course-component member.

Constraints

• Both parent and referenced sides use the same content_ref contract and the same public content_kind enum.
• Clients do not submit parent_content_kind or referenced_content_kind. The server derives both from Content; it stores only the parent snapshot needed by the physical relationship table and derives referenced_content_kind on reads.
• A content item may reference another item, but the referenced item remains its own row and version history.
• relationship_kind is required and governs renderer/export behavior; consumers must not infer it from filenames or MIME type.
• display_order is only order within this parent relationship set; it is null or an exact PostgreSQL int4 in 0..2147483647, and Curriculum owns instructional sequence order.
• form_index is required only for relationship_kind='bank_member', must already be an exact PostgreSQL int4 in 1..2147483647 without coercion, and orders equivalent fixed forms inside a test_bank; it is null for non-bank relationships.
• usage_scope is required and defaults to practice_eligible. For relationship_kind='bank_member', routing may select only rows whose member usage_scope and referenced content_item usage_scope are both practice_eligible.
• relationship_kind='source_script_for' must point from produced audio/video to a script Content item; relationship_kind='bank_member' must point from a content_kind='test_bank', assessment_role='bank' parent using membership_rule='same_blueprint_fixed_forms' or 'pp100_adaptive_shards' to a content_kind='test', assessment_role='form' target. qti_adaptive_selection cannot carry parallel bank_member rows. The bank and form version UUIDs must be owned by their corresponding artifacts, and KC attribution is resolved only from embeds whose parent_version_id equals the stored referenced_version_id.
• A bank_member may be added or soft-retired only when its parent is status='draft', retired_at IS NULL, and trust_status='unknown'; missing trust normalizes to unknown. Every other current lifecycle/trust state returns 409 content.bank_membership_frozen. Wrong parent/member kind or assessment role, wrong-owner versions, native-QTI parallel rows, non-integer/invalid indexes, malformed active historical rows, and empty fixed-form exact versions return 409 content.invalid_test_bank_member_contract; a missing, empty, or ambiguous exact question graph or missing KC attribution retains 409 content.bank_kc_attribution_incomplete. Empty native-QTI banks remain valid because their selection graph is native. A fresh create takes the exclusive graph fence and repeats replay arbitration before semantic validation and claim; the relationship write, item touch, claim, and completion are atomic. Integrity reads acquire the shared writer fence on client A before starting the stable reader snapshot on client B. After fixed identity/metadata reads, the active-member and exact form/question/KC hydrations are two set queries, so database round trips do not grow with the graph. The supported inclusive ceilings are 1,024 active bank-member facts for one bank and 4,096 joined question/KC facts for one integrity read. PostgreSQL selects at most cap plus one and rejects the sentinel before mapping; in-memory contract tests apply the same limits before cloning. An overflow returns 409 content.invalid_test_bank_member_contract with fieldErrors.name=bank_members or form_graph and requires the graph to be split before retry. Rejected fresh requests leave no graph or idempotency row and completed retries still replay.

Invalid examples

• Embedding a transcript JSON blob directly inside the video content_item.
• A playlist or question bank stored as content_reference rows instead of Curriculum collection membership.
• A referenced content title copied onto the relationship row.
• A relationship_kind parsed from a package path such as /captions/.

Example

{
  "content_reference_id": "7bcb0edb-7df1-4f1c-89b8-b7f5f0ca2451",
  "parent_content_id": "2f4b4c12-7c41-4b40-a33a-9637581d0e51",
  "parent_content_kind": "video",
  "referenced_content_id": "91fa40b1-f8a1-4f4a-8a60-c19a93dce4f1",
  "referenced_content_kind": "media_asset",
  "relationship_kind": "transcript_for",
  "display_order": 1,
  "form_index": null,
  "usage_scope": "practice_eligible",
  "referenced_usage_scope": "practice_eligible"
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/references

GET /tenants/{tenantId}/alpha/content/items/{contentId}/bank-members

Public API fields (14)

public read view · Content. Belongs here because it describes the blank place a student can answer; the student's actual answer lives in Events or Results.

answer_slot — Answer Slot

QTI response declaration exposed in plain language.

Purpose

Use answer_slot to know what responses a question or test expects without scanning QTI XML or treating student submissions as Content.

Relationships

• Belongs to one content_version and content_item.
• May have one answer_key row for authorized teacher/author reads.

Constraints

• Only variable_kind='response' rows are public answer slots.
• identifier, cardinality, and base_type preserve QTI semantics.
• No student response values are stored here.

Invalid examples

• A student selected choice stored as answer_slot.value.
• A variable_kind='outcome' row exposed as a student answer slot.
• Cardinality parsed from an interaction title.

Example

{
  "answer_slot_id": "0c6f5271-9d9e-4df1-8d5d-2ad59dacde66",
  "content_id": "58612cab-9c46-426c-8c20-0e9f19c807c5",
  "version_id": "efcf3561-3a66-4825-9588-e792ef20c312",
  "identifier": "RESPONSE",
  "cardinality": "single",
  "base_type": "identifier"
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/answer-slots for the latest effective grading version

GET /tenants/{tenantId}/alpha/content/items/{contentId}/answer-slots?version_id={versionId} for immutable history

Fields (7)

authorized public read view · Content. Belongs here because it is part of the blank instrument; access is teacher/author-only and it is never a student's submitted answer.

answer_key — Answer Key

Teacher-only correct response and mapping for an answer slot.

Purpose

Use answer_key to inspect what the blank question expects. Choice identifiers and text-entry string answers both stay in correct_response.values. Results owns whether a specific student was correct.

Relationships

• One-to-one with answer_slot when a correct response exists.
• Belongs to content_version through the answer_slot.

Constraints

• Requires read:answer_key scope.
• Never included in student_view.
• correct_response JSON shape must match answer_slot.cardinality and base_type.
• For textEntryInteraction answer slots with base_type string, each accepted string answer appears as one exact correct_response.values entry.

Invalid examples

• Answer keys returned to a student-safe content read.
• correct_response used as a student's answer.
• Values coerced from identifiers to display text.

Example

{
  "answer_slot_id": "0c6f5271-9d9e-4df1-8d5d-2ad59dacde66",
  "correct_response": {
    "values": [
      "choiceA"
    ]
  },
  "required_scope": "read:answer_key"
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/answer-keys with scope read:answer_key for the latest effective grading version

GET /tenants/{tenantId}/alpha/content/items/{contentId}/answer-keys?version_id={versionId} with scope read:answer_key for immutable history

Fields (4)

public read view · Content. Belongs here because it describes the blank test or question structure before any student attempt exists.

assessment_component — Assessment Component

Sections, parts, choices, rubrics, and pre-authored feedback from the QTI component tree.

Purpose

Use assessment_component to inspect how a blank assessment is assembled: sections, parts/interactions, question references, distractors, rubrics, and pre-authored feedback. Results owns the student's responses, scores, AI grading, and feedback after an attempt.

Relationships

• Belongs to one content_version and content_item.
• May point to a parent assessment_component for section/part/choice hierarchy.
• Works alongside answer_slot, answer_key, and scoring_plan; it does not replace them.

Constraints

• component_role must be one of the published assessment_component_role values.
• Only blank-instrument structure belongs here.
• Distractors are authored choices, not a record of what a student selected.
• Rubrics and feedback are authored instrument text, not AI grading audit logs.

Invalid examples

• A student's selected distractor stored as an assessment_component.
• AI-generated feedback from a completed attempt stored as preauthored_feedback.
• A section title parsed from display text instead of read from the QTI component row.

Example

{
  "assessment_component_id": "51ef0fdf-7ebf-4cc7-9242-45432b04ad4f",
  "content_id": "58612cab-9c46-426c-8c20-0e9f19c807c5",
  "version_id": "efcf3561-3a66-4825-9588-e792ef20c312",
  "parent_assessment_component_id": "3f2bf843-b8d7-49d0-b4d1-17dbb3152c43",
  "component_role": "distractor",
  "ordinal": 1,
  "element_name": "qti-simple-choice",
  "qti_identifier": "choiceA",
  "response_identifier": "RESPONSE",
  "display_text": "x = 4"
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/assessment-components for the latest effective grading version

GET /tenants/{tenantId}/alpha/content/items/{contentId}/assessment-components?version_id={versionId} for immutable history

GET /tenants/{tenantId}/alpha/content/items/{contentId}/assessment-components?component_role=rubric

Fields (14)

public read view · Content. Belongs here because it documents how the blank instrument will be scored; actual scored outcomes live in Results.

scoring_plan — Scoring Plan

QTI processing provenance for scoring a content version.

Purpose

Use scoring_plan to inspect processing rules and operands without confusing the plan with a student's processed result.

Relationships

• Belongs to one content_version and content_item.
• May reference answer_slot identifiers through operands.

Constraints

• Only blank-instrument scoring provenance belongs here.
• No outcome_state, processing_trace for a student, or gradebook rollup is stored here.
• rule_scope must use the governed processing scope.

Invalid examples

• A student's SCORE value stored on scoring_plan.
• Processing trace from a live attempt exposed through Content.
• rule_name rewritten into school-language text that cannot round-trip to QTI.

Example

{
  "scoring_plan_id": "74b1c279-e85f-407e-971e-16408ad792c0",
  "version_id": "efcf3561-3a66-4825-9588-e792ef20c312",
  "rule_scope": "response",
  "rule_name": "qti-map-response",
  "sequence_number": 10
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/scoring-plan for the latest effective grading version

GET /tenants/{tenantId}/alpha/content/items/{contentId}/scoring-plan?version_id={versionId} for immutable history

Fields (9)

Results-owned public read dependency · Content documents this read contract because apps display it beside Content, but Results owns the storage, recomputation, and public alpha.content_effectiveness view.

content_effectiveness — Content Effectiveness

Results-owned derived effectiveness read for one content item.

Purpose

Use content_effectiveness to see whether content helped students validate or fail later outcomes without making apps maintain counters.

Relationships

• The Content endpoint aggregates Results-owned rows by content_id and returns exactly the seven fields in publicApiProjection.
• The raw Results view may carry content_version_id, kc_id, subject_id, evidence_count, policy_version, and source_result_refs; those fields are inputs to aggregation/provenance and are not top-level Content endpoint members.

Constraints

• Read-only to Content and to consumers; Results is the only writer/recompute owner.
• Raw DB contract: alpha.content_effectiveness has top-level tenant_id, object_id, and data JSONB. Fields marked Results raw view only are projected from data JSON and are not top-level members of GET /items/{contentId}/effectiveness.
• Content API contract: GET /items/{contentId}/effectiveness returns exactly content_id, validated_by_count, failed_by_count, effectiveness_rate, freshness_status, last_recomputed_at, and provenance.
• validated_by_count and failed_by_count are non-negative integers.
• effectiveness_rate is null until policy says enough data exists; otherwise 0.0000 through 1.0000.
• freshness_status is the Content response status derived from Results view presence/freshness and fallback policy; it is not a client-maintained field.

Invalid examples

• An app increments validated_by_count after a quiz.
• A dashboard recomputes effectiveness_rate from raw Results.
• A Content migration creates alpha.content_effectiveness or alpha.content_effectiveness_view.
• A raw query filters alpha.content_effectiveness.content_id instead of data ->> 'content_id'.
• Mastery state stored here.

Example

{
  "content_id": "58612cab-9c46-426c-8c20-0e9f19c807c5",
  "validated_by_count": 42,
  "failed_by_count": 8,
  "effectiveness_rate": 0.84,
  "freshness_status": "fresh",
  "last_recomputed_at": "2026-06-03T18:22:00Z",
  "provenance": {
    "sourceModule": "results",
    "policy": "alpha.policy.content_effectiveness.v1",
    "resultCount": 50
  }
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/effectiveness

SELECT data ->> 'content_id', (data ->> 'validated_by_count')::int, (data ->> 'effectiveness_rate')::numeric FROM alpha.content_effectiveness WHERE tenant_id = :tenant_id AND data ->> 'content_id' = :content_id

Fields (13)

Alpha extension table · Content. Belongs here because it is validation and provenance evidence for a blank content item; it is not a conformance job runner or a learner event.

content_trust_evidence — Content Trust Evidence

Read-only trust status and validation provenance for Content.

Purpose

Use content_trust_evidence to read the compatible item-level validation summary and emergency revocation state. trust_status=failed is the single post-visibility revocation lever; positive trust and the internal append-only QC receipts written by QTI import or governed publication are recorded provenance, never release preconditions (owner ruling #2534: approval never gates student visibility).

Relationships

• One-to-one with content_item.
• Internal content_version_qc_receipt has one row per exact immutable version that passed a Content-owned QC path.

Constraints

• Read-only to ordinary consumers.
• Public evidence must redact package bytes, secrets, tokens, and direct student PII.
• trust_status is not publication status; a published item can become degraded or failed.
• Positive trust and append-only version receipts are recorded provenance; release never waits for them. trust_status=failed fails every version closed on the next read (owner ruling #2534).
• Only Content-owned QTI import or governed spec, bank, or article publication writes a version receipt. No caller verdict, producer attestation, or generic trust mutation participates.

Invalid examples

• A public route that starts a conformance run.
• trust_status used to hide a retired content item.
• Problem detail includes raw package bytes.

Example

{
  "content_id": "58612cab-9c46-426c-8c20-0e9f19c807c5",
  "trust_status": "trusted",
  "evidence_url": "https://platform3-andymontgomery-9773s-projects.vercel.app/qti/1edtech/surface_qc/",
  "checked_at": "2026-06-03T18:20:00Z"
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/trust

Fields (5)

module-local policy table/view plus bundled source-adapter policy · Content. Belongs here only for Content-specific policy names and values; cross-student outcome formulas stay in Results or platform policy.

content_policy — Content Policy

Current Content policy projection, including module-local rows and the bundled versioned source-adapter policy.

Purpose

Use content_policy to inspect policy metadata the Content surface exposes. Current TimeBack import behavior is governed by the bundled content-source-adapter-policy.json resource loaded by the implementation; it does not dynamically read alpha.content_policy. Consumers must not copy magic numbers or source-kind mappings into dashboards or skill packs.

Relationships

• Referenced by content_effectiveness.provenance.
• The bundled content-source-adapter-policy.json resource, not a request-time database lookup, is referenced by the TimeBack production import adapter for processed-facts and resources content_kind folding plus expected_xp calibration.
• Does not own Results formulas, K-8 advancement, tenure buckets, or NWEA norms policy.

Constraints

• policy_key starts with alpha.policy.content.
• value_json must be typed and versioned where a policy row exists; clients must not duplicate policy numbers.
• alpha.policy.content.expected_xp_band declares bands 0_to_10, 10_to_20, 20_to_45, and 45_plus with thresholds, legacy aliases, and numeric range filter aliases.
• The bundled alpha.policy.content.source_adapter resource declares TimeBack production processed-facts recognition, raw resources identity fields, specific content-kind text mappings, an interactive fallback for otherwise opaque application-backed resources, duration-to-XP calibration, and per-content-kind expected_xp fallbacks. This is the Content half of Curriculum's course_components -> component_resources -> resources reconcile; the adapter response returns source_record_id with normalized content_id/content_kind/expected_xp for the Curriculum import to consume.
• Secrets are never policy values.

Invalid examples

• A dashboard hardcodes minimum effectiveness sample size.
• NWEA Growth X target stored in Content policy.
• A student app credential stored as policy JSON.

Example

{
  "policy_key": "alpha.policy.content.source_adapter.v2026-07-15",
  "value_json": {
    "resources": {
      "source_identity_fields": [
        "sourced_id",
        "vendor_resource_id"
      ],
      "fallback_content_kind": "interactive"
    },
    "processed_facts": {
      "content_kind": "interactive"
    },
    "expected_xp": {
      "seconds_to_xp": {
        "divisor": 8,
        "round": "nearest_integer",
        "min": 1,
        "max": 90
      },
      "defaults_by_content_kind": {
        "interactive": 15,
        "article": 8,
        "test": 25
      }
    }
  },
  "policy_version": "2026-07-15.1"
}

Queries

GET /tenants/{tenantId}/alpha/content/policies

Fields (4)

derived public read · Content owns the single exact-version lifecycle and item-revocation answer consumed by Curriculum, Incept, gates, and launch surfaces.

content_release_eligibility — Content Release Eligibility

Current lifecycle and item-revocation answer for one exact Content version.

Purpose

Return one exact immutable Content version identity, its kind and usage scope, and a fail-closed answer where release_eligible always equals Content serveability.

Relationships

• One answer per requested exact content_id/content_version_id.
• content_kind and usage_scope are projected from the owning Content item.
• trust_status is the current item-level trust state: failed for item revocation, trusted or degraded when positive evidence exists, otherwise unknown; unknown never blocks release.
• producer_surface remains a validated compatibility query input but is not returned and does not select an eligibility lane.
• release_eligible is exactly equal to serveable.
• Curriculum retains the exact version reference and asks Content again at launch; it does not copy or recompute the verdict.

Constraints

• The exact version must belong to the requested Content item in the caller tenant; a same-tenant mismatch is content.release_wrong_version and a cross-tenant or unknown version is content.not_found.
• Every success repeats the requested content_id and content_version_id plus that item's governed content_kind and usage_scope; consumers must fail closed on any identity or kind mismatch.
• serveable=true exactly when status is published, retired_at is null, and item trust is not failed; positive trust or a per-version QC receipt is never a precondition (owner ruling #2534).
• release_eligible must always equal serveable.
• Failed item trust is the all-version revocation lever and fails closed on the next read; per-version QC receipts are append-only provenance evidence and never gate the answer.
• No producer release evidence, required quality-bar tuple, owner acceptance tuple, or version allowlist participates.
• A non-serveable exact version fails closed as content.version_not_serveable.
• Routine Curriculum launch and next-form routing additionally require usage_scope=practice_eligible.

Invalid examples

• A published version is hidden while waiting for a QC receipt, producer sign-off, or any other positive approval signal.
• A published trusted version waits for a producer evidence append.
• An owner allowlist makes a draft, retired, or failed-trust version eligible.
• release_eligible differs from serveable.
• A consumer copies the verdict instead of reading the exact version at launch.

Example

{
  "content_id": "58612cab-9c46-426c-8c20-0e9f19c807c5",
  "content_version_id": "efcf3561-3a66-4825-9588-e792ef20c312",
  "content_kind": "question",
  "usage_scope": "practice_eligible",
  "trust_status": "trusted",
  "serveable": true,
  "release_eligible": true,
  "problem_code": null
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/versions/{versionId}/release-eligibility?producer_surface=incept%2Falpha

Public API fields (8)

virtual field dictionary · Content. Belongs here as the Content-specific field dictionary for the inherited platform RFC 7807 error envelope.

content_problem — Content Problem

Stable problem codes and field errors returned by Content APIs.

Purpose

Use content_problem to implement clients that react to validation, stale-version, authorization, and relationship errors without parsing human text.

Relationships

• May be replayed by platform.idempotency_key for failed_permanent requests.
• requestId and traceId tie to platform request tracing, not Content rows.

Constraints

• Every Problem response includes type, title, status, code, detail, requestId, traceId, and a fieldErrors array. cause and docUrl are the only optional envelope fields.
• detail text is plain language and must not leak package bytes, secrets, tokens, or direct student PII.
• Unsupported filters are rejected, never silently ignored.

Invalid examples

• A 400 response with only 'bad request'.
• A problem detail that includes raw QTI package bytes.
• A stale If-Match save accepted instead of content.stale_version.

Example

{
  "type": "https://platform.timeback.com/problems/content/stale-version",
  "title": "Content version is stale",
  "status": 409,
  "code": "content.stale_version",
  "detail": "Read the content again and retry with the current ETag.",
  "requestId": "req_01HX...",
  "traceId": "tr_01HX...",
  "fieldErrors": []
}

Queries

Returned from all Content API endpoints when a documented constraint fails.

Fields (10)

Content-owned authorization configuration · Content owns the exception to its own ordinary release lane; People & Orgs remains the identity/reality authority.

content_test_serve_grant — Content Test-Serve Grant

Soft-revocable demo authorization for one test/synthetic identity.

Purpose

Permit governed draft-bank exercise without publishing, trusting, or making the material ordinarily serveable.

Relationships

• tenant_id resolves platform.tenant and test_identity_id resolves the same tenant's People & Orgs alpha.person at authorization time.
• One row exists per tenant, test_identity_id, and producer_surface.
• Successful authorizations reference this row through test_serve_grant_id.

Constraints

• Runtime authorizes only tenant_key=demo and a live enabled person with reality in test or synthetic and an active source status.
• expires_at and revoked_at fail closed; no caller flag or header can create or bypass a grant.
• The grant changes no Content item status, trust status, serveable value, or release_eligible value.

Invalid examples

• A real student person_id or non-demo tenant.
• A wildcard identity, tenant, or producer surface.
• An upsert that clears revoked_at during deploy.

Example

{
  "test_serve_grant_id": "18240000-0000-4000-8000-000000000001",
  "test_identity_id": "person_reality_test_student",
  "producer_surface": "incept/alpha",
  "authorization_ref": "https://github.com/andymontgomery-byte/platform3/issues/1824",
  "expires_at": null,
  "revoked_at": null
}

Queries

SELECT test_serve_grant_id, test_identity_id, producer_surface, authorization_ref, expires_at, revoked_at FROM alpha.content_test_serve_grant WHERE tenant_id=:tenant_uuid AND test_identity_id=:person_id AND producer_surface=:producer_surface

Fields (8)

append-only Content authorization audit · Content records the exact exceptional asset serve; Results still owns attempts, answers, scores, seen-form evidence, and remediation state.

content_test_serve_authorization — Content Test-Serve Authorization

Immutable exact bank/form test-serve receipt.

Purpose

Prove who exercised which draft material under which grant without changing production governance.

Relationships

• test_serve_grant_id references the exact Content grant.
• bank/form ids and versions reference qti.artifact and qti.artifact_version.
• gate_component_id references the Curriculum gate by immutable identifier; Content does not copy the gate.

Constraints

• Eligibility and final authorization routes require an authenticated service, demo, or operator role; ordinary student, teacher, writer, reviewer, and reader tokens cannot invoke the lane even when they know the test identity.
• Bank must be content_kind=test_bank and form content_kind=test with an active exact bank_member relationship.
• Both exact items must be practice eligible, unretired, non-failed trust, and not explicitly rejected, blocked, or revoked.
• UPDATE and DELETE fail in the database.

Invalid examples

• A receipt written before bank membership validation.
• Latest-version substitution.
• Mutation or deletion of a historical receipt.
• A receipt containing bearer credentials, answers, scores, or real-student PII.

Example

{
  "test_serve_authorization_id": "18240000-0000-4000-8000-000000000051",
  "test_serve_grant_id": "18240000-0000-4000-8000-000000000001",
  "test_identity_id": "person_reality_test_student",
  "producer_surface": "incept/alpha",
  "gate_component_id": "18240000-0000-4000-8000-000000000030",
  "bank_content_version_id": "18240000-0000-4000-8000-000000000011",
  "form_content_version_id": "18240000-0000-4000-8000-000000000021",
  "authorized_at": "2026-07-24T12:00:00Z"
}

Queries

SELECT * FROM alpha.content_test_serve_authorization WHERE tenant_id=:tenant_uuid AND test_identity_id=:person_id ORDER BY authorized_at DESC

Fields (14)

current authorization read · Content is the only authority for exceptional draft material serving.

content_test_serve_eligibility — Content Test-Serve Eligibility

Current fail-closed answer for one exact Content version and one test identity.

Purpose

Allow Curriculum to validate every exact bank/form candidate before final atomic authorization.

Relationships

• Reads one exact content_item/content_version pair, active grant, live person reality, and current trust.
• Final bank/form authorization repeats these checks and additionally proves exact bank membership in one transaction.

Constraints

• The route requires an authenticated service, demo, or operator role before repository resolution.
• Eligibility never makes serveable or release_eligible true and never mutates lifecycle/trust.
• A grant mismatch (including a producer_surface not matching the active grant), revoked/expired grant, real identity, failed trust, retired/archived lifecycle, or non-practice use returns a typed 403/409 Problem.
• A successful eligibility read is not a launch receipt; final authorization is mandatory.

Invalid examples

• A preview=true parameter.
• A cached eligible answer reused after revocation.
• An eligible answer for a real identity or non-demo tenant.

Example

{
  "content_id": "18240000-0000-4000-8000-000000000010",
  "content_version_id": "18240000-0000-4000-8000-000000000011",
  "content_kind": "test_bank",
  "usage_scope": "practice_eligible",
  "producer_surface": "incept/alpha",
  "test_identity_id": "person_reality_test_student",
  "test_serve_grant_id": "18240000-0000-4000-8000-000000000001",
  "test_serve_eligible": true,
  "problem_code": null
}

Queries

GET /tenants/{tenantId}/alpha/content/items/{contentId}/versions/{versionId}/test-serve-eligibility?producer_surface=incept%2Falpha&test_identity_id=person_reality_test_student

Public API fields (9)