Refactor Initiative Analyzer to Use Initiative-Centric Processing

[PlagueHO]

Technical Debt: Architectural Improvement Opportunity

Current State

The initiative analyzer currently uses an item-centric approach where each backlog item is individually analyzed against all initiatives. This creates several inefficiencies:

• API Inefficiency: N API calls for N backlog items
• Context Dilution: LLM must consider all initiatives for each single item
• Suboptimal Goal Alignment: Process doesn't directly match the stated goal of "enriching initiatives with relevant backlog items"
• Limited Batch Analysis: No opportunity for comparative analysis within initiative context

Proposed Solution

Implement an initiative-centric approach with the following architecture:

Core Algorithm Change

# Current: Item → Initiatives mapping
for backlog_item in backlog_items:
    best_initiative = analyze_against_all_initiatives(backlog_item)

# Proposed: Initiative → Items mapping  
for initiative in initiatives:
    for chunk in chunk_backlog_items(backlog_items, size=20):
        relevant_items = analyze_chunk_for_initiative(initiative, chunk)

Benefits

1. Strategic Alignment: Directly matches goal of enriching initiatives
2. API Efficiency: Reduces API calls by ~80% through batching
3. Enhanced Context: LLM focuses on single initiative context
4. Richer Analysis: Comparative analysis within initiative scope
5. Better Token Utilization: Optimal use of context windows

Implementation Tasks

• Create chunk_backlog_items() function for batching
• Implement analyze_chunk_for_initiative() with structured output
• Design new system prompt for initiative-centric analysis
• Update data models to handle batch results
• Implement result aggregation and deduplication
• Add comprehensive testing for new approach
• Performance benchmarking against current implementation

JSON Schema for Structured Output

{
  "type": "object",
  "properties": {
    "relevant_items": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "backlog_item_title": {"type": "string"},
          "relevance_score": {"type": "integer", "minimum": 0, "maximum": 100},
          "impact_analysis": {"type": "string"},
          "strategic_value": {"type": "string"},
          "implementation_synergies": {"type": "string"},
          "confidence_reasoning": {"type": "string"}
        }
      }
    }
  }
}

Risk Mitigation

• Implement both approaches initially for A/B comparison
• Ensure backward compatibility with existing CSV formats
• Add configuration flag to switch between approaches
• Comprehensive testing with real data sets

Impact Assessment

• Performance: 80% reduction in API calls
• Quality: More focused, context-aware analysis
• Maintainability: Clearer separation of concerns
• Scalability: Better handling of large datasets

Implementation Priority

High - This change significantly improves both efficiency and analytical quality, directly supporting the primary use case.