Deep Analysis Agents for MoE Escalation Handling

Second-tier specialists invoked when first-tier classification is uncertain. These agents perform more expensive but more accurate analysis.

Architecture: Escalation → DeepAnalysisOrchestrator → [DeepAnalysts] → FinalDecision

Deep Analysts: 1. SemanticSimilarityAnalyst - Compare against known exemplars using embeddings 2. ContextualAnalyst - Analyze directory context, sibling files, naming patterns 3. ContentReasoningAnalyst - Deep content analysis with structured reasoning 4. CrossReferenceAnalyst - Check references, links, and relationships

File: deep_analysts.py

Classes

DeepAnalystType

Types of deep analysis agents.

DeepAnalysisVote

Vote from a deep analysis agent.

DeepAnalysisResult

Result from the deep analysis pipeline.

SemanticSimilarityAnalyst

Analyzes document similarity to known exemplars.

ContextualAnalyst

Analyzes document context including directory structure,

ContentReasoningAnalyst

Performs deep content analysis with structured reasoning.

CrossReferenceAnalyst

Analyzes cross-references, links, and relationships to other documents.

DeepAnalysisOrchestrator

Orchestrates deep analysis for escalated documents.

Functions

analyze(doc, original_votes)

Compare document against known exemplar patterns.

analyze(doc, original_votes)

Analyze document context for classification hints.

analyze(doc, original_votes)

Deep content analysis with structured reasoning.

analyze(doc, original_votes)

Analyze document references and relationships.

analyze_escalation(doc, original_result)

Perform deep analysis on an escalated document.

should_escalate_to_human(result)

Determine if document needs human review after deep analysis.

get_analysis_summary(result)

Get a human-readable summary of deep analysis.