peptide-repo-core/ARCHITECTURE.md

Peptide Repo Core — Architecture

Peptide Repo Core is a WordPress plugin that provides the canonical peptide data layer for the peptiderepo.com ecosystem. It owns the peptide custom post type (consolidated in v0.2.0; previously registered by Peptide Search AI pre-schema-sprint — PSA v4.5.0 drops its registration in favor of this one), the peptide_category taxonomy, custom tables for dosing rows and legal status cells, typed repository classes, an AI candidate queue for semi-automated data extraction, a shared disclaimer component, and JSON-LD structured data output. Every consumer plugin (PR Theme, PRAutoBlogger, Peptide News, Peptide Search AI, and all Tier 1 tools) reads peptide data through this plugin's versioned PHP API, never directly.

Ownership model (v0.2.0+). PR Core is the sole registrant of the peptide CPT and peptide_category taxonomy. Registration is guarded by post_type_exists() / taxonomy_exists() so a second registration from any other plugin (historically PSA, which now defers) no-ops cleanly and deploy order between plugins does not matter. See CONVENTIONS.md for the CPT ownership rule.

---

Data Flow

                         ┌───────────────────────┐
                         │  PubMed / Literature   │
                         └──────────┬────────────┘
                                    │ (future: extraction pipeline)
                                    ▼
                         ┌───────────────────────┐
                         │ AI Candidate Queue     │  pr_ai_candidate_queue table
                         │ (pending → review)     │  Confidence-sorted
                         └──────────┬────────────┘
                                    │ Approve
                                    ▼
┌──────────────────┐    ┌───────────────────────┐    ┌──────────────────┐
│ Admin UI / REST  │───▶│   Dosing Repository   │───▶│ pr_dosing_rows   │
│ (manual entry)   │    │   Legal Repository    │    │ pr_legal_cells   │
└──────────────────┘    └───────────────────────┘    └──────────────────┘
                                    │
                                    ▼
                         ┌───────────────────────┐
                         │  Peptide Repository   │  peptide CPT + meta
                         └──────────┬────────────┘
                                    │
                    ┌───────────────┼───────────────┐
                    ▼               ▼               ▼
            ┌──────────┐    ┌──────────┐    ┌──────────┐
            │ REST API │    │ JSON-LD  │    │ Consumer │
            │ pr-core  │    │ Drug     │    │ Plugins  │
            │ /v1/     │    │ schema   │    │ via API  │
            └──────────┘    └──────────┘    └──────────┘

---

File Tree

peptide-repo-core/
├── peptide-repo-core.php              # Plugin bootstrap — constants, autoloader, activation hooks
├── uninstall.php                      # Selective teardown: plugin-owned posts + tables + options + caps
├── ARCHITECTURE.md                    # This file
├── CONVENTIONS.md                     # Naming patterns, extension guides
├── CHANGELOG.md                       # Semantic versioning changelog
├── composer.json                      # Dev dependencies (PHPCS, WPCS) and lint scripts
├── phpcs.xml.dist                     # PHPCS ruleset configuration
│
├── .github/
│   └── workflows/
│       └── ci.yml                     # PHP lint (8.1-8.3), PHPCS WordPress, 300-line check, unit tests
│
├── assets/
│   └── css/
│       └── admin.css                  # Minimal admin meta box styles
│
├── includes/
│   ├── class-pr-core.php              # Main orchestrator — boots subsystems, registers public filters
│   ├── class-pr-core-activator.php    # Activation + version-change flush handler
│   ├── class-pr-core-deactivator.php  # Deactivation: rewrite flush only (data preserved)
│   ├── class-pr-core-autoloader.php   # SPL autoloader for PR_Core_ prefixed classes
│   ├── class-pr-core-seed-data.php    # Dev fixture: 3 peptides, 10 dosing rows, 5 legal cells
│   │
│   ├── cpt/
│   │   ├── class-pr-core-peptide-cpt.php       # CPT + peptide_category registration (guarded), meta fields, sanitizers
│   │   └── class-pr-core-schema-sanitizers.php # Sanitizers for v0.6.0 schema-input meta (formula, weight, aliases, FAQ)
│   │   └── class-pr-core-repo-daily-cpt.php    # Repo Daily CPT registration for editorial content
│   │
│   ├── taxonomies/
│   │   └── class-pr-core-repo-daily-taxonomy.php  # repo_daily_category taxonomy + term seeding
│   │
│   ├── migrations/
│   │   ├── class-pr-core-migration-runner.php           # Sequential migration engine
│   │   ├── class-pr-core-migration-0001-dosing-rows.php # pr_dosing_rows table
│   │   ├── class-pr-core-migration-0002-legal-cells.php # pr_legal_cells table + unique constraint
│   │   ├── class-pr-core-migration-0003-candidate-queue.php # pr_ai_candidate_queue table
│   │   ├── class-pr-core-migration-0004-backfill-peptide-meta.php # Backfill _pr_* schema meta from PSA psa_* keys
│   │   └── class-pr-core-pubchem-client.php              # PubChem REST client (used by migration 0004)
│   │
│   ├── dto/
│   │   ├── class-pr-core-peptide-dto.php     # Typed peptide value object
│   │   ├── class-pr-core-dosing-row-dto.php  # Typed dosing row value object
│   │   ├── class-pr-core-legal-cell-dto.php  # Typed legal cell value object
│   │   └── class-pr-core-candidate-dto.php   # Typed candidate queue value object
│   │
│   ├── repositories/
│   │   ├── class-pr-core-peptide-repository.php          # CRUD for peptide CPT
│   │   ├── class-pr-core-dosing-repository.php           # CRUD for pr_dosing_rows
│   │   ├── class-pr-core-legal-repository.php            # CRUD for pr_legal_cells
│   │   └── class-pr-core-candidate-queue-repository.php  # CRUD + approve/reject for queue
│   │
│   ├── admin/
│   │   ├── class-pr-core-admin.php                    # Admin initialization, menu, styles
│   │   ├── class-pr-core-peptide-metaboxes.php        # Identifiers, dosing, legal meta boxes
│   │   ├── class-pr-core-repo-daily-metaboxes.php     # Author byline and clinical review flag meta boxes
│   │   ├── class-pr-core-admin-columns.php            # Evidence, editorial, dosing count columns
│   │   └── class-pr-core-candidate-queue-page.php     # AI candidate review admin screen
│   │
│   ├── frontend/
│   │   ├── class-pr-core-disclaimer.php       # Shortcode + static API for surface disclaimers
│   │   ├── class-pr-core-jsonld.php           # Orchestrator: Yoast-integrated Drug/FAQ/MedicalWebPage emission
│   │   ├── class-pr-core-jsonld-drug.php      # Drug (+ MolecularEntity) schema piece builder
│   │   ├── class-pr-core-jsonld-webpage.php   # MedicalWebPage retype + lastReviewed/reviewedBy enrichment
│   │   └── class-pr-core-jsonld-faq.php       # FAQPage schema piece (emits only when _pr_faq_items populated)
│   │
│   └── api/
│       └── class-pr-core-rest-controller.php  # REST endpoints for peptides, dosing, legal
│
└── tests/
    ├── bootstrap.php                  # Lightweight WP function mocks for no-PHPUnit unit runs
    └── unit/
        └── test-peptide-cpt.php       # CPT + taxonomy guard + args payload assertions

---

Content Types

CPT	Slug	Taxonomy	Purpose	Registrant
peptide	/peptides/[slug]/	peptide_category	Canonical peptide monographs with dosing, legal, evidence metadata	PR Core v0.2.0 (consolidated from PSA)
repo_daily	/daily/[slug]/	repo_daily_category	Editorial publications (articles, guides, comparisons, news) generated by PRAutoBlogger, published under Boo Sheeran byline	PR Core v0.5.0

Verification scanner scope: The verification scanner (class-pr-core-verification-scanner.php) queries post_type=peptide only. repo_daily posts are excluded automatically by CPT separation and do not appear in the overdue queue.

---

Custom Tables

{prefix}pr_dosing_rows

High-cardinality dosing data, 1:many with peptide. Indexed on peptide_id, route, population, pubmed_id. Supersede pattern: corrections create a new row and set superseded_by_id on the old one.

{prefix}pr_legal_cells

Per-country legal status, 1:many with peptide. Unique constraint on (peptide_id, country_code, superseded_by_id) — only one active cell per peptide x country.

{prefix}pr_ai_candidate_queue

AI-extracted dosing candidates. Status flow: pending → approved/rejected → merged (if approved). Approved rows copy into pr_dosing_rows with source = 'ai-candidate-approved'.

---

Public API

Repositories (typed DTOs, never raw objects)

• PR_Core_Peptide_Repository — find_by_id, find_by_slug, search, find_all, count
• PR_Core_Dosing_Repository — find_by_id, find_by_peptide, insert, supersede, count_by_peptide
• PR_Core_Legal_Repository — find_by_id, find_by_peptide, find_by_peptide_and_country, find_by_country, insert, supersede
• PR_Core_Candidate_Queue_Repository — find_by_id, find_by_status, insert, approve, reject, count_by_status

Filters

• pr_core_get_indexable_corpus — returns { id, url, title, body, type } entries for search indexing
• pr_core_disclaimer_for_surface — returns disclaimer text for a given surface identifier
• pr_core_evidence_strength_label — maps enum value to localized label
• pr_core_jsonld_peptide — filter JSON-LD schema before output
• pr_core_register_peptide_fields — add computed fields to peptide DTO (planned)

Actions

• pr_core_before_peptide_publish, pr_core_after_peptide_publish
• pr_core_before_dosing_row_publish, pr_core_after_dosing_row_publish
• pr_core_before_legal_cell_publish, pr_core_after_legal_cell_publish
• pr_core_candidate_approved, pr_core_candidate_rejected

REST API (namespace: pr-core/v1)

Method	Endpoint	Auth	Description
GET	/peptides	Public	List peptides with filters
GET	/peptides/{id}	Public	Single peptide
GET	/peptides/{id}/dosing	Public	Dosing rows for peptide
POST	/peptides/{id}/dosing	manage_peptide_content	Add dosing row
GET	/peptides/{id}/legal	Public	Legal cells for peptide
POST	/peptides/{id}/legal	manage_peptide_content	Add legal cell

---

Key Decisions

#1: Separate plugin for canonical schema

Consumer plugins read through versioned PHP API + hooks, never direct table access. Decouples domain schema from feature lifecycle. See ADR-0002.

#2: Custom tables for dosing and legal data

Post-meta doesn't scale for high-cardinality data queried by multiple dimensions (peptide + route + population). Custom tables with proper indexes.

#3: Supersede pattern instead of hard deletes

Corrections create new rows; old rows get superseded_by_id. Preserves full audit trail for regulatory accountability.

#4: AI candidate queue with human-in-the-loop

Automated extraction populates the queue; humans approve/reject. Approved rows copy into dosing table with provenance tracking (source, ai_candidate_id).

#5: Disclaimer component owned by core

Single editorial review point. All consumer plugins render the same versioned disclaimer text. Surface-specific copy (dosing, legal, reconstitution, AI answer) stored in one wp_option.

#6: JSON-LD from day one (v0.6.0: Yoast-integrated)

Drug schema on single pages increases LLM citation rate. Filter hook allows consumer plugins to extend the schema object.

v0.6.0 Yoast integration (ratified jsonld-contract-v1 2026-06-11):

• Integrated path (Yoast active): PR Core hooks into wpseo_schema_graph_pieces (priority 11) to inject Drug and FAQPage pieces. Yoast owns WebPage/BreadcrumbList — we never emit those. wpseo_schema_webpage_type filter retypes the page node to MedicalWebPage. wpseo_schema_graph (priority 11) enriches the existing WebPage piece with lastReviewed, reviewedBy, and audience.
• Standalone fallback (no Yoast): wp_head action (priority 99) emits a complete @graph (Drug + FAQPage) only. Suppressed automatically when function_exists('YoastSEO') is true.
• Drug @id: {permalink}#drug — stable for cross-plugin about linkage from PRAutoBlogger articles.
• Molecular data source: _pr_molecular_formula, _pr_molecular_weight, _pr_aliases (written by migration 0004, sourced from PSA psa_* keys or PubChem REST API).
• FAQ emission: _pr_faq_items post-meta (JSON array of {question, answer} objects). Node emitted only when items exist.
• Dosing omitted from schema: YMYL constraint — no dosing data in JSON-LD.

JSON-LD class tree (v0.6.0):

Class	File	Responsibility
PR_Core_Jsonld	frontend/class-pr-core-jsonld.php	Orchestrator: registers hooks, Yoast vs. standalone routing
PR_Core_Jsonld_Drug	frontend/class-pr-core-jsonld-drug.php	Builds Drug (+ MolecularEntity) schema node
PR_Core_Jsonld_Webpage	frontend/class-pr-core-jsonld-webpage.php	MedicalWebPage retype + lastReviewed/reviewedBy enrichment
PR_Core_Jsonld_Faq	frontend/class-pr-core-jsonld-faq.php	FAQPage node (emit-only-when-populated)

#7: PR Core owns the peptide CPT and peptide_category taxonomy (v0.2.0)

Prior to v0.2.0, PR Core registered pr_peptide while Peptide Search AI registered peptide — both claimed the public rewrite slug peptides, and WP's rewrite resolver picked PR Core's empty CPT, 404'ing all 89 production peptide detail pages. v0.2.0 consolidates both registrations onto a single peptide CPT, owned by PR Core. PSA v4.5.0 drops its CPT/taxonomy registration; its meta boxes (psa_peptide_data, psa_extended_data), directory shortcode, KB article renderer, and search widget continue operating on the shared peptide CPT regardless of who registers it. Registration on both sides is guarded with post_type_exists() / taxonomy_exists() so deploy order is forgiving.

§2.9 Uninstall specification

PR Core uninstall.php removes plugin-owned data only:

1. Drops custom tables — pr_dosing_rows, pr_legal_cells, pr_ai_candidate_queue.
2. Deletes peptide posts only if they carry the _pr_core_authored meta flag. The 89 canonical peptide posts predate PR Core and were authored via PSA; they are never blanket-deleted on PR Core uninstall.
3. Does not delete peptide_category terms. Shared-ownership taxonomy; term metadata the site relies on outlasts this plugin.
4. Deletes pr_core_* options.
5. Removes manage_peptide_content capability from all roles.
6. Deletes v0.6.0 schema-input meta keys (_pr_molecular_formula, _pr_molecular_weight, _pr_aliases, _pr_faq_items) from all peptide posts. These are PR Core-authored values (sourced from PubChem / PSA) and are safe to remove on uninstall. They do not overlap with PSA's psa_* namespace — PSA's own meta is unaffected.