Handling Rate Limits (429)

Rhythm APIs may return a 429 Too Many Requests response when a client sends requests at a high rate. This is standard behavior for REST APIs and is not specific to your account or credentials.

When you receive a 429, your client should pause and retry the request after a delay rather than immediately retrying — which would likely trigger another 429.

The recommended approach is exponential backoff with jitter: each successive retry waits longer than the last, with a small random offset to avoid synchronized retries from multiple clients.

A basic pattern:

1. Receive a 429 response
2. Wait base * 2^attempt + jitter milliseconds
3. Retry the request
4. If the retry also returns 429, increase the attempt count and repeat
5. Stop after a maximum number of attempts and surface an error

Rather than implementing backoff from scratch, use a well-tested library for your language:

• p-retry — Promise-based retry with full backoff control (ESM only; requires a bundler or dynamic import() in CommonJS projects)
• axios-retry — drop-in retry plugin for Axios (ensure your axios dependency is kept up to date independently)

• tenacity — full-featured retry library with exponential, fibonacci, and custom wait strategies

• Resilience4j Retry — production-grade retry with exponential backoff and jitter; the recommended choice for Java 17+ projects

• Polly — comprehensive resilience library with retry, circuit breaker, and backoff policies

• retry-go — simple retry with configurable backoff strategies including exponential and fibonacci

• GuzzleRetryMiddleware — retry middleware for Guzzle with Retry-After header support
• PHP Backoff — standalone backoff library with exponential, linear, and polynomial strategies

• A 429 response means your client is sending too many requests — it is not an auth error
• Always implement retry logic with exponential backoff in any integration that calls Rhythm APIs under load
• Honor the Retry-After header when present
• Use an established library rather than rolling your own implementation