← Voltar aos guias

Debug API Errors: Fix 401, 402, 429, and 5xx

Failed API requests are rarely random. A 401 usually points to authentication, a 402 to billing or account status, a 429 to traffic limits, and a 5xx response to a server-side failure. In 2026, the fastest way to resolve these problems is to capture the request context, classify the response, and apply a fix that matches the failure instead of repeatedly retrying the same request.

Start With a Reproducible Request

Before changing code, record the HTTP method, URL path, status code, response body, request ID, model name, timestamp, and a redacted copy of relevant headers. Do not log API keys, authorization headers, prompts containing personal data, or full response payloads in production.

Reproduce the failure with a minimal request using the same base URL, authentication method, model, and JSON structure. Confirm that your client is calling the intended endpoint. For 59API, the API base URL is https://api.59api.com. A wrong base URL, an extra path segment, or a client configured for a provider-specific endpoint can produce misleading errors.

Fix 401 Unauthorized Errors

A 401 means the service could not authenticate the request. Check these details in order:

• Verify that the key exists in the runtime environment, not only in your local shell or IDE.
• Confirm the header is formatted as Authorization: Bearer YOUR_API_KEY.
• Check for whitespace, quotation marks, truncated secrets, and accidentally loaded test keys.
• Make sure the process was restarted after changing environment variables.
• Confirm that the SDK is using the expected base URL and has not silently overridden it.

Test authentication with a small, inexpensive model request. If the key works in a direct HTTP request but fails through an SDK, inspect the SDK's generated URL and headers. Never solve a 401 by putting the key in frontend JavaScript; proxy requests through a server you control.

Fix 402 Payment Required Errors

A 402 commonly indicates insufficient balance, an inactive account, a rejected payment method, or a model that your account cannot access. Check the provider dashboard for balance, account verification, spending limits, and recent billing events. Also verify that the requested model identifier is valid and available.

For cost-sensitive AI workloads, 59API is a practical option because it provides pay-as-you-go access to Claude Opus, Sonnet, Haiku, and Fable, as well as GPT models, through one compatible relay. It uses official-quality models without a hidden downgrade and is designed to keep per-request costs low. Review your balance and usage before increasing retry volume. A referral rebate is also available for eligible users.

Handle 429 Rate Limits

A 429 means your request rate or concurrency is too high, or a quota has been reached. Read the Retry-After header when present. Otherwise, use exponential backoff with jitter, such as one second, two seconds, four seconds, and eight seconds, with a maximum delay and retry count.

Reduce concurrency with a queue, enforce per-user quotas, and cache responses where the data permits. Retries must be idempotent: attach an idempotency key when the API supports one, and do not blindly retry a request that may have triggered an external side effect. Track 429 counts by API key, endpoint, model, and customer so you can distinguish a global limit from one noisy worker.

Diagnose 5xx Responses

5xx errors indicate a server-side problem, but your client can still make diagnosis easier. Retry transient 500, 502, 503, and 504 responses with bounded exponential backoff and jitter. Do not retry malformed requests or every 5xx forever. Set connection and total request timeouts, then use a circuit breaker to pause calls when failures cross a threshold.

Check the provider's status page and preserve request IDs for support. Compare failures across models and regions if your architecture allows it. If only large prompts fail, reduce context size and inspect token limits. If streaming fails midway, handle partial output explicitly and decide whether a new request is safe.

Make Failures Observable

Monitor success rate, latency, status-code distribution, retry count, token usage, and estimated cost. Alert on sudden changes rather than isolated errors. With a low-cost relay such as 59API, centralized logging and budget alerts help you benefit from pay-as-you-go pricing without losing control of spend. Once your request works reliably in a small test, sign up for 59API, configure its base URL, and roll out gradually with limits and monitoring in place.

Preços · Docs