Skip to main content

Enhanced MoE Orchestrator - Integrates all enhancement modules.

This orchestrator extends the base MoE system with:

  • Semantic embeddings for improved similarity
  • Historical learning with dynamic weights
  • Memory system integration
  • Adaptive thresholds
  • Confidence calibration
  • Additional specialized judges
  • Batch corpus analysis

File: enhanced_orchestrator.py

Classes​

EnhancedOrchestratorConfig​

Configuration for enhanced orchestrator.

EnhancedMoEOrchestrator​

Enhanced MoE Orchestrator integrating all improvement modules.

Functions​

get_enhanced_orchestrator(config)​

Get or create singleton enhanced orchestrator.

reset_enhanced_orchestrator()​

Reset the singleton (useful for testing).

get_pre_classification_hints(document)​

Get hints from embeddings and memory before classification.

classify(document)​

Classify a document using enhanced classification with embeddings.

get_dynamic_analyst_weights()​

Get dynamic weights for analysts based on historical accuracy.

get_adaptive_thresholds()​

Get current adaptive thresholds.

calibrate_confidence(raw_confidence)​

Apply calibration to confidence score.

record_classification(result)​

Record classification for learning and threshold adjustment.

confirm_classification(document_path, actual_type)​

Confirm a classification for learning.

get_stats()​

Get statistics from all enhancement modules.

Usage​

python enhanced_orchestrator.py