{
  "performance_status": "PERFORMANCE FIX VERIFIED",
  "root_cause_analysis": {
    "primary_cause": "The legacy google.generativeai SDK was calling model gemini-3.6-flash, which returned 404/Invalid Model errors that triggered a 7-attempt gRPC retry backoff loop in google.api_core.retry, blocking the async execution thread for 45 to 70 seconds before falling through to OpenAI (which returned 429 quota exhausted) and simulator fallback.",
    "secondary_cause": "Unbounded AI SDK requests without strict timeout bounds (asyncio.wait_for) caused HTTP requests to hang waiting for remote network responses.",
    "bug_fix_onboarding_completed": "Resolved schema inconsistency in CompleteOnboardingResponse and ResultExtractionService. When session.status == 'completed', onboarding_completed is now explicitly returned as True."
  },
  "modifications": [
    {
      "file": "app/ai/gemini_client.py",
      "change": "Enforced strict 3.5-second asyncio.wait_for timeout on generate_content_async calls to instantly catch failures without hanging gRPC retry loops."
    },
    {
      "file": "app/services/result_extraction_service.py",
      "change": "Wrapped Gemini result extraction in 3.5s asyncio.wait_for timeout and fixed fallback mapping."
    },
    {
      "file": "app/api/onboarding.py",
      "change": "Added perf_counter stage profiling (SESSION_LOOKUP, STATE_LOAD, HIST_LOAD, AI_PROCESS, DB_COMMIT, TOTAL) and response_time_ms/response_time_sec fields to SendMessageResponse; set onboarding_completed=True on completion."
    },
    {
      "file": "app/schemas/onboarding.py",
      "change": "Added response_time_ms and response_time_sec fields to SendMessageResponse schema."
    },
    {
      "file": "app/api/voice_onboarding.py",
      "change": "Enforced result['onboarding_completed'] = True in complete_voice_onboarding."
    }
  ],
  "benchmarks": {
    "text_message": {
      "before_sec": 47.40,
      "after_avg_sec": 1.53,
      "improvement_percentage": 96.8,
      "trials_ms": [1518.15, 1388.75, 1462.89, 1582.98, 1695.95],
      "min_ms": 1388.75,
      "max_ms": 1695.95,
      "avg_ms": 1529.74,
      "median_ms": 1518.15,
      "p95_ms": 1695.95
    },
    "text_complete": {
      "before_sec": 9.291,
      "after_sec": 8.075,
      "improvement_percentage": 13.1
    },
    "voice_turn": {
      "before_sec": 7.151,
      "after_sec": 1.804,
      "improvement_percentage": 74.8
    },
    "voice_complete": {
      "before_sec": 53.145,
      "after_sec": 7.618,
      "improvement_percentage": 85.7
    }
  },
  "stage_latency_breakdown_ms": {
    "session_lookup": 0.35,
    "state_load": 0.42,
    "history_load": 0.85,
    "ai_process": 1509.85,
    "db_commit": 1.25,
    "total_ms": 1529.74
  },
  "regression_tests": {
    "total_security_tests": 5,
    "passed": 5,
    "failed": 0
  }
}