MAJOR FEATURES ADDED:
======================
🤖 ML Analysis System:
- Comprehensive scene image analysis with per-scene predictions
- Enhanced database schema with scene_ml_analysis table
- Advanced detection for clothing colors, body types, age categories, positions, settings
- Support for multiple prediction types (clothing, body, sexual acts, etc.)
- Confidence scoring and ML source tracking
🧠 Enhanced Search Capabilities:
- Natural language parser for complex queries (e.g., "Teenage Riley Reid creampie older man pink thong black heels red couch")
- Category-based search with confidence-weighted results
- ML-enhanced tag matching with automatic fallback to traditional search
- Support for "Money Shot: Creampie" vs "Cum in Open Mouth" detection
🗄️ Advanced Database Schema:
- Male detection: circumcised field (0/1)
- Pubic hair types: natural, shaved, trimmed, landing strip, bushy, hairy
- Scene ML analysis table for storing per-scene predictions
- Comprehensive seed tags for all detection categories
🏗️ Dual Scraper Architecture:
- Flexible import service supporting both TPDB and Adult Empire scrapers
- Bulk scraper implementation for Adult Empire using multiple search strategies
- Progress tracking with Server-Sent Events (SSE) for real-time updates
- Graceful fallback from Adult Empire to TPDB when needed
📝 Enhanced Import System:
- Individual bulk imports (performers, studios, scenes, movies)
- Combined "import all" operation
- Real-time progress tracking with job management
- Error handling and retry mechanisms
- Support for multiple import sources and strategies
🔧 Technical Improvements:
- Modular component architecture for maintainability
- Enhanced error handling and logging
- Performance-optimized database queries with proper indexing
- Configurable import limits and rate limiting
- Comprehensive testing framework
This commit establishes Goondex as a comprehensive adult content discovery platform with ML-powered analysis and advanced search capabilities, ready for integration with computer vision models for automated tagging and scene analysis.
Stu Leak
3b8adad57d