Error Handling

Handling Pattern

1. Catch and classify error type.
2. Retry only transient errors.
3. Capture request_id/job_id in logs.
4. Escalate permanent errors to user-facing state.

Sample Error Envelope

{
  "error": {
    "code": "insufficient_credits",
    "message": "Not enough credits to process this video",
    "request_id": "req_123456"
  }
}

Recommended SDK Usage Pattern

1. Initialize singleton client per service process.
2. Wrap SDK calls with your own timeout/retry decorators.
3. Map SDK/domain errors into internal error taxonomy.
4. Capture job_id and business context in your tracing system.

SDK Hardening Checklist

• Add per-call timeout defaults and circuit breaking.
• Track SDK version in logs for post-incident debugging.
• Test SDK upgrade in staging with archived production payloads.
• Implement fallback strategy by model and output format.

Error Recovery Flow

1. Classify error as retriable vs non-retriable.
2. Apply retry policy with capped attempts.
3. Escalate with full diagnostic context after budget exhausted.