Root cause: inject_graph_pieces() added plain PHP arrays to wpseo_schema_graph_pieces.
Yoast Schema_Generator::filter_graph_pieces_to_generate() calls get_class() +
is_needed() on every piece expecting Abstract_Schema_Piece objects — a plain array
produces: PHP Fatal: get_class(): Argument #1 ($object) must be of type object,
array given. All peptide monograph pages returned HTTP 500; non-peptide pages were
unaffected (is_singular guard).
Fix: remove wpseo_schema_graph_pieces hook entirely. New inject_graph_nodes() method
hooks wpseo_schema_graph at priority 12 (after enrich_webpage_piece at 11), which
receives the fully-assembled graph as array-of-arrays — safe for plain array appends.
Also adds regression test (test-jsonld-peptide-piece.php) that:
- Reproduces the v0.6.1 get_class(array) TypeError (fails-before)
- Confirms inject_graph_nodes Drug+FAQ emission is correct (passes-after)
Agent-Session: jsonld-peptide-piece-fix v0.6.2
- Register repo_daily CPT at /daily/[slug]/ for editorial publications
- Register repo_daily_category taxonomy with 4 seeded terms
- Add Repo Daily metabox for author byline and clinical review flag
- Bulk-migrate _repo_daily_author on autoblogged posts (psa_source)
- Update ARCHITECTURE.md with content-type table and scanner scope
- Update CONVENTIONS.md with repo_daily meta field extension pattern
- Bump version to 0.5.0, add CHANGELOG entry
Closes v0.5.0 scope.
v0.2.0 removed TAX_FAMILY from PR_Core_Peptide_CPT but left two stale
references in PR_Core_Peptide_Repository:
- find_all() family filter branch (class-pr-core-peptide-repository.php:95-102)
- post_to_dto() families taxonomy lookup (class-pr-core-peptide-repository.php:138)
JSON-LD emission on single-peptide templates called find_by_id() ->
post_to_dto() which hit the undefined constant and killed every peptide
detail page render mid-wp_head().
Fix:
- family filter is silently ignored until a future minor bump
removes it from the REST schema.
- post_to_dto() hardcodes $families = [] so the DTO still carries
the field (REST response shape unchanged for existing clients).
Out-of-scope follow-ups:
- Post-mortem on why the v0.2.0 QA gate missed this (unit tests
exercised the CPT class in isolation, not the repository's
consumption of its constants).
- Add an integration-style test for JSON-LD emission on a single
peptide so any future CPT-class refactor catches repository drift.
Co-authored-by: peptiderepo <peptiderepo@users.noreply.github.com>
Consolidates the `peptide` CPT under PR Core ownership, dropping the `pr_peptide` slug-collision that was 404ing all 89 peptide detail pages.
- POST_TYPE: pr_peptide -> peptide
- TAX_CATEGORY: pr_peptide_category -> peptide_category
- TAX_FAMILY + pr_peptide_family taxonomy removed
- Defensive post_type_exists/taxonomy_exists guards on both registrations
- Args harmonized as superset of PSA + prior PR Core (adds thumbnail support)
- Activation hook + init:999 version-change flush handler
- Uninstall scoped to _pr_core_authored flag (PSA-authored posts survive)
- Plugin version 0.1.1 -> 0.2.0, CHANGELOG in same commit
- ARCHITECTURE.md + CONVENTIONS.md updated with CPT ownership rule
QA gate verdict: PASS (qa-reviews/peptide-repo-core/2026-04-22-874f93b.md)
Sibling release: peptiderepo/peptide-search-ai#15 (PSA v4.5.0)
Known follow-up: manage_peptide_content capability is not granted to any role as of v0.2.0. No current UI/theme reads or writes PR Core-defined meta keys, so no regression. Follow-up thread to decide role assignment.
1. NAMESPACE is a PHP reserved keyword — renamed to REST_NAMESPACE
in the REST controller. This caused a parse error on activation.
2. Autoloader can't resolve PR_Core class — the class name equals the
autoloader prefix exactly, so stripping "PR_Core_" yields an empty
suffix and the wrong filename. Added explicit require_once in
bootstrap, same pattern as PRAutoBlogger.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
