🐛 Fix: analyze_health_topic timeout on comprehensive analysis (MCP 10-second limit)

[ma3u]

🐛 Bug Report

Problem Description

The analyze_health_topic tool times out when using "comprehensive" depth setting through MCP clients. The timeout occurs due to MCP clients' 10-second default timeout, while comprehensive analysis takes >10 seconds to process 43,373 documents.

Current Behavior

• Error: MCP error -32001: Request timed out
• Occurs when: Using depth="comprehensive" parameter
• Processing time: >10 seconds for comprehensive analysis
• MCP client timeout: 10 seconds (hard limit in most clients)

Root Cause

1. Large dataset: Processing 43,373 documents in vector database
2. Complex operations: Multiple vector searches, result aggregation, AI synthesis
3. MCP timeout: 10-second client timeout is shorter than processing time

🎯 Proposed Solution: Streaming Progressive Responses

Implementation Plan

Phase 1: Streaming Response Architecture (High Priority)

Implement SSE-based progressive response streaming to keep connections alive:

async def analyze_health_topic_streaming(topic: str, depth: str):
    # Send initial acknowledgment (< 1 second)
    yield {"type": "progress", "status": "started", "progress": 0}
    
    # Stream search results (2-4 seconds)
    search_results = await vector_search(topic)
    yield {"type": "progress", "progress": 25, "partial_results": search_results[:5]}
    
    # Stream categorization (4-6 seconds)
    categories = await categorize_results(search_results)
    yield {"type": "progress", "progress": 50, "categories": list(categories.keys())}
    
    # Stream AI synthesis (6-9 seconds)
    for category, analysis in process_categories_async(categories):
        yield {"type": "progress", "progress": 75, "partial_analysis": {category: analysis}}
    
    # Final result (< 10 seconds total)
    yield {"type": "complete", "progress": 100, "result": comprehensive_analysis}

Phase 2: Optimization (Medium Priority)

1. Parallel Processing: Run searches concurrently
2. Smart Caching: Cache common topics with 1-hour TTL
3. Result Truncation: Limit to 30 documents for comprehensive

Phase 3: Background Processing (Low Priority)

Implement job queue for truly comprehensive analysis with polling.

Quick Fix (Immediate)

if depth == "comprehensive" and len(results) > 25:
    results = results[:25]  # Temporary fix to stay under 10 seconds

Success Metrics

• Comprehensive analysis completes in <10 seconds
• No timeout errors in MCP Inspector
• Progressive updates visible to user
• Claude.ai integration works smoothly

References

• TIMEOUT_ANALYSIS.md
• MCP Production Test Report