rdmbair15m5-changelog-20260822-1634-rddb-manual-labeling-and-rule-learning-engine

• Scope: rdmbair15m5 (native UI, database & learning engine) & rdmsm4x:~/dev/apps/rdDB.
• Summary: Architected and implemented a thoughtful Self-Learning Engine & Manual Labeling Studio for rdDB.app. Users can now manually edit file taxonomy labels (with instant USER_OVERRIDE badge lock & Finder tag sync), teach persistent classification rules directly from any selected file, manage rules in a dedicated "Learning & Rules" studio, and execute fleetwide rule evaluation across 31,000+ files in SQLite.

Architecture & Features Delivered

1. Interactive Manual Labeling & Tag Editor:
   • ManualTagEditorSheet: Inline modal for editing L1 Domain, L2 Vendor, L3 Solution, L4 Customer, and L5 Archetype.
   • On save: updates SQLite record with confidence_score: 1.0, locks USER_OVERRIDE badge, and writes native macOS Finder tags.
2. "Teach Rule from This File..." Flow:
   • TeachRuleSheet: Smart rule creator pre-populated with the selected file's directory path, filename, and target taxonomy.
   • Saves persistent ClassificationRule records in SQLite classification_rules table.
3. Dedicated "Learning & Rules" Studio:
   • New sidebar view displaying active learning rules, match patterns, target taxonomy chips, and live match counters.
   • One-click "Apply All Rules to Database Now" re-evaluates all user rules against the database and updates matching records.
4. Classifier Engine Integration:
   • ClassifierEngine.swift evaluates user-defined learning rules with top priority before falling back to heuristics or LLM inference.
5. Multi-Source Connectors Staged by Teamwork:
   • Apple Notes (AppleNotesConnector.swift), Apple Mail (AppleMailConnector.swift), Safari & Chrome (BrowserConnector.swift), and Document Extractor (DocumentExtractor.swift) staged with test fixtures in Tests/RDDBCoreTests/Fixtures/.