DarkMatter AIDC 2026
Unified Release Notes
Current Version: 0.4.2
========================================

0.4.2 - DOCUMENTATION CONSOLIDATION
-----------------------------------
- Reproduced index.html, project.html, and harvester.html using the established DarkMatter AIDC visual theme and existing image filenames.
- Added direct links from all three pages to README.md, RELEASE_NOTES.txt, VERSION, and PROJECT_CANON.txt.
- Consolidated prior release notes into this single release-notes file.
- Updated all current documentation identity and visible release labels to 0.4.2.
- Documented worker.php as the controlled Harvester orchestration layer.
- Documented extractor.php as the Ollama structured-JSON interpretation layer.
- Documented provenance fields, extraction engine identity, endpoint URL, endpoint category, excerpts, confidence, claims, and proposals.
- Documented human review as the mandatory publication boundary.
- Documented review-order-safe canonical company creation and duplicate-safe primary publication.
- Documented proper typed publication for companies, endpoints, company_endpoints, technologies, services, brands, and graph relationships.
- Defined the split between human-readable HTML interfaces and machine-readable JSON/data outputs.
- Preserved the original project documentation as PROJECT_CANON.txt.

0.4.1 - PUBLICATION ORDER AND COMPANY ENRICHMENT
-------------------------------------------------
- Company enrichment proposals are generated even when the company is new.
- Publishing description, website, sector, or company type first creates or resolves the canonical company record.
- Primary company publication is duplicate-safe.
- AI company classification publishes as AI & Machine Intelligence / Artificial Intelligence Company.
- Endpoint detail panels show directly linked companies from company_endpoints.
- Main company cards show the official website and avoid PRIVATE / OTHER or UNCLASSIFIED COMPANY placeholders.
- Endpoint relationships remain outside generic relationship lists.
- Bundled harvested and test records were removed for a clean proof run.
- PHP syntax, SQLite integrity, schema additions, endpoint publication, review-order behavior, and ZIP integrity were validated in the supplied test report.

0.4.0 - DATA MODEL AND ENDPOINT PUBLICATION CORRECTION
------------------------------------------------------
- Rebuilt Harvester publication around the data model used by the main AIDC application.
- Added AI & Machine Intelligence discovery category.
- Added companies.sector.
- Added endpoints.source_url.
- Added company_endpoints for real company-to-endpoint linkage.
- Corrected LINKED ENDPOINTS counts on company cards.
- Removed endpoint records from generic relationship display.
- Replaced weak company-label fallbacks with ticker, sector, company type, or an explicit unclassified state.
- Removed regex heuristic claim extraction.
- Added strict canonical-name validation for Ollama output.
- Preserved deterministic HTML endpoint discovery through DOMDocument.
- Added correct endpoint publication for the required category_id schema.
- Added installation and Apache permission instructions.

0.3.1 - RELATIONSHIP-FIRST PUBLICATION
--------------------------------------
- Added automatic entity_relationships graph-table creation.
- Backfilled previously published Harvester claims into the relationship graph.
- Populated unmapped brands.company_id from approved owns_brand claims.
- Reused existing entities during publication.
- Added company technology, service, and relationship display.
- Added extraction-engine diagnostics.
- Added a research-phrase root-entity guard.
- Retained Ollama endpoint and model environment overrides.

0.2.2 - EXTRACTION AND OWNERSHIP SAFETY
---------------------------------------
- Added an entity-first extraction prompt with company, brand, technology, service, endpoint, and infrastructure classifications.
- Improved clause and sentence boundary handling in the former heuristic extractor.
- Filtered sentence fragments, dates, attribution phrases, and self-references.
- Treated acquisitions as companies rather than brands.
- Passed the research goal into the Ollama extraction prompt.
- Made new-company research the default queue type.
- Changed approved owns_brand publication to map a brand to an existing parent company.
- Preserved human review as the publication boundary.

ORIGINAL PROJECT CANON
----------------------
AIDC began as an interactive technology ownership map. Its enduring investigative path is:

What am I using?
→ What technology powers it?
→ Which brand provides it?
→ Which corporation owns the brand?
→ Which data-center systems support it?

The complete original design document is preserved in PROJECT_CANON.txt.

INSTALLATION TARGET
-------------------
/var/www/html/AIDC_2026/

Recommended command when included with the application release:

sudo bash /var/www/html/AIDC_2026/install_permissions.sh