Chapter D · 4 of 4~9 min

API errors & retries

Production LLM calls hit **429 rate limits**, timeouts, and bad JSON — handle them on purpose.

Reviewed July 2026 · interactive lab

Paths & resources for this lesson

1Try it yourself

Playground

API errors & retries

429 rate limits are normal — back off, retry, and cap attempts in production.

Attempt 1

429 Rate limit — retry after 2s

2Read & reflect

Recap

#What you just did

You walked through 429 → wait → retry → success — the shape every SDK wrapper needs.

Teach

#How it works

| Code | Meaning | Typical action | |------|---------|----------------| | 429 | Rate limit | Exponential backoff + cap retries | | 401 | Bad key | Alert ops — don't retry blindly | | 500 | Server error | Retry with jitter, then fallback | | 400 | Bad request | Fix payload — don't retry same body |

Read

#Try it

Add a retry loop around your chat call — max 3 attempts, double wait each 429. Log each failure with request id.

3Spark check

Continue learning · glossary & guides