klingai-prod-checklist

# Kling AI Production Checklist ## Overview Checklist covering authentication, error handling, cost controls, monitoring, security, and content policy before deploying Kling AI video generation to production. ## Authentication - [ ] AK/SK stored in secrets manager (not `.env` in repo) - [ ] JWT auto-refresh with 5-min buffer before 30-min expiry - [ ] Separate API keys per environment (dev/staging/prod) - [ ] Key rotation schedule (quarterly minimum) - [ ] `Authorization: Bearer <token>` format verified ## Error Handling - [ ] HTTP 400/401/402/403/429/5xx all handled - [ ] Exponential backoff with jitter for 429/5xx retries - [ ] Max retry limit set (3-5, not infinite) - [ ] `task_status: "failed"` logs `task_status_msg` - [ ] 30s timeout on all HTTP calls - [ ] `duration` sent as string `"5"` not integer `5` ## Cost Controls - [ ] Daily credit budget enforced in code - [ ] Alert at 80% daily budget consumption - [ ] `standard` mode used for non-final renders - [ ] Max poll attempts cap (no infinite loops) - [ ] Credit estimate before batch submission ```python # Pre-batch credit check credits_needed = len(prompts) * 10 # 10 credits per 5s standard if credits_needed > DAILY_BUDGET: raise RuntimeError(f"Batch needs {credits_needed}, budget is {DAILY_BUDGET}") ``` ## Task Management - [ ] All task_ids logged with timestamp - [ ] Stuck task detection (>10 min in processing) - [ ] `callback_url` used instead of polling in production - [ ] Failed tasks queued for r...