Timeouts and Retries

Connections built by the SDK time out and retry transient failures by default:

client = PaystackSdk::Client.new(
  secret_key: "sk_test_xxx",
  timeout: 30,        # seconds to wait for a response
  open_timeout: 5,    # seconds to wait for the connection to open
  max_retries: 2,     # retries after the first attempt (0 disables retrying)
  retry_interval: 0.5 # base backoff in seconds, doubled on each retry with jitter
)

• Read-only requests (GET) are retried on network failures and 429/502/503/504 responses.
• Anything that writes (POST, PUT, DELETE) is retried only on 429, where Paystack rejected the request for exceeding the rate limit. A timeout or 5xx on a write may mean Paystack processed it, so it is not retried. Pass retry_non_idempotent: true only if you deduplicate yourself (e.g. verify by reference first).
• On a 429 the SDK waits for Paystack's x-ratelimit-reset header (seconds) before retrying. If the wait exceeds 10 seconds it stops retrying and raises PaystackSdk::RateLimitError (#retry_after holds the value).
• Errors that survive all retries are raised as PaystackSdk::RateLimitError, PaystackSdk::ServerError, PaystackSdk::TimeoutError or PaystackSdk::ConnectionError, all of which inherit from PaystackSdk::Error.
• These options only apply to connections the SDK creates. Passing them together with your own Faraday::Connection raises ArgumentError; configure your connection yourself. Invalid values (e.g. a negative max_retries) also raise ArgumentError.