[QTI] Accept QTI items through the internal API for ricecooker import

[rtibbles]

Overview

Land assessment items uploaded through the internal/ricecooker API uniformly as QTI: store editor-shaped and arbitrary QTI directly, convert uploaded legacy/structured items to QTI via the #6 machinery, validate against the schema, and map referenced media.

Complexity: High
Target branch: unstable

Context

• create_exercises (contentcuration/contentcuration/views/internal.py) builds items from structured fields and validates only assessment_id uniqueness; type and content go unchecked until publish.
• The internal endpoint and utils/nodes.py duplicate logic that the AssessmentItemViewSet/serializer (Introducing Content Curation #1) already owns; delegating to it cleans up that tech debt.
• QTI items carry content as raw_data XML and reference media mapped via map_files_to_assessment_item (utils/nodes.py).
• Arbitrary QTI may use interactions Studio can't edit; these are non-editable passthrough, published verbatim.
• The Perseus custom interaction (data-type='perseus') is a publish-time-only wrapper (Browserify #8); if our own published QTI is re-imported, it must be unwrapped back to a raw perseus_question.

The Change

• Refactor create_exercises to delegate to the AssessmentItemViewSet/serializer machinery (Introducing Content Curation #1), removing duplicated logic in the internal endpoint and nodes utils.
• Accept QTI items (raw_data XML + media) directly; validate each against the schema (Ant's backend plus setting up front end files #2), rejecting non-compliant items with actionable errors.
• Convert uploaded legacy/structured items to QTI on ingest via the resolves merge conflict that i allowed in... #6 conversion; leave raw Perseus as raw Perseus (handled by Browserify #8).
• Unwrap any Perseus custom interaction (data-type='perseus'): read its referenced data-perseus-path payload from the upload and restore it as a raw perseus_question, rather than storing the custom interaction.
• Map media/file dependencies so they are present at publish.

Acceptance Criteria

• create_exercises delegates to the AssessmentItem serializer/machinery; duplicated internal-endpoint and nodes logic is removed
• The API accepts and persists QTI items (raw_data XML + media)
• Uploaded QTI is schema-validated (Ant's backend plus setting up front end files #2); non-compliant items are rejected with a clear error
• Uploaded legacy/structured items are converted to QTI on ingest; raw Perseus is preserved as-is
• A Perseus custom interaction is unwrapped by reading its data-perseus-path payload and restoring the raw perseus_question
• Referenced media is mapped and present at publish
• Tests cover QTI upload (editor-shaped + arbitrary), legacy-converted-on-ingest, an invalid upload, and media mapping

References

• contentcuration/contentcuration/views/internal.py
• contentcuration/contentcuration/utils/nodes.py

AI usage

Architecture decided with the maintainer across an iterative session: blanket QTI type with the item XML in raw_data; XSD-authoritative validation across all sources; a legacy→QTI global migration with an API-layer dual-read; ricecooker upload delegating to the AssessmentItem serializer; and a Perseus custom-interaction contract confirmed against the QTI 3.0 specification. Claude mapped the existing publish/validation/ricecooker code, proposed the breakdown, and drafted each issue; the maintainer steered every decision and reviewed throughout.