Erreurs

All API-key endpoints return JSON. On any non-2xx response, the body has the same shape:

{
  "success": false,
  "error": "<human-readable message>"
}

Some errors include extra context (e.g. générique on a 429 quota error). The HTTP status code is the source of truth — branch on the status, then read error for display or logging.

Status code reference#

Distinguishing the two 429 cases#

A 429 always means "stop sending for now", but why matters.

Rate limit (60 / minute, per user):

{
  "success": false,
  "error": "Rate limit exceeded. Max 60 requests per minute."
}

Monthly credit exhaustion (20,000 / month, per user):

{
  "success": false,
  "error": "Monthly credit limit reached (20000). Resets next month.",
  "credits": { "used": 20000, "limit": 20000, "remaining": 0 }
}

Branch on the presence of the générique object, or on a substring match for "Monthly credit limit". See Limites de débit et Crédits et mentions légales for the recovery strategy in each case.

Error handling pattern (Node.js)#

async function whFetch(path, init = {}) {
  const res = await fetch(`${ORIGIN}${path}`, {
    ...init,
    headers: { 'X-API-Key': process.env.WH_API_KEY, ...(init.headers || {}) },
  });

  let body = null;
  try { body = await res.json(); } catch { /* non-JSON, leave null */ }

  if (res.ok) return body;

  const error = (body && body.error) || res.statusText;

  if (res.status === 401) throw new Error(`Auth failed: ${error}`);
  if (res.status === 403) throw new Error(`Plan required: ${error}`);
  if (res.status === 404) throw new Error(`Unknown endpoint: ${path}`);
  if (res.status === 429) {
    const retryable = body && body.credits ? false : true; // monthly quota is not retryable today
    const err = new Error(error);
    err.retryable = retryable;
    throw err;
  }
  if (res.status >= 500) {
    const err = new Error(`Upstream error: ${error}`);
    err.retryable = true;
    throw err;
  }
  throw new Error(`HTTP ${res.status}: ${error}`);
}

Error handling pattern (Python)#

import os, requests

def wh_fetch(path, **kwargs):
    r = requests.request(
        method=kwargs.pop("method", "GET"),
        url=f"{ORIGIN}{path}",
        headers={"X-API-Key": os.environ["WH_API_KEY"], **kwargs.pop("headers", {})},
        **kwargs,
    )

    body = None
    try:
        body = r.json()
    except ValueError:
        pass

    if r.ok:
        return body

    error = (body or {}).get("error") or r.reason

    if r.status_code == 401:
        raise PermissionError(f"Auth failed: {error}")
    if r.status_code == 403:
        raise PermissionError(f"Plan required: {error}")
    if r.status_code == 429:
        is_quota = bool((body or {}).get("credits"))
        raise RuntimeError(f"Rate limited (quota={is_quota}): {error}")
    if r.status_code >= 500:
        raise RuntimeError(f"Upstream error: {error}")

    raise RuntimeError(f"HTTP {r.status_code}: {error}")

Response shape vs HTTP status#

• Metered API-key guard errors (401, 403, limit 429, generic catch 500) always return JSON with success: false and a matching HTTP status.
• Some JSON handlers (especially large TikTok Shop payloads) may return HTTP 200 with a body that includes business-level flags — always read the schema for that endpoint rather than assuming success tracks HTTP alone.

HTML instead of JSON? Metered programmatic routes set Content-Type: application/json. HTML usually means the wrong host, a proxy error page, or a path that never hit the app — see Troubleshooting.

Reporting bugs#

If you hit a 500 or a response that contradicts the docs:

1. Capture the request URL, method, headers (redact your key), and body.
2. Capture the response status and body.
3. Note the timestamp.
4. Send the bundle to support — the team can correlate with server logs by user ID + timestamp.