Error handling

Design-partner phase. The public api.schneeai.com endpoint is activated at general availability. The error envelopes and status codes below are the production contract; until GA, exercise them against your partner-specific endpoint.

Error response shape

Every non-2xx response has the same envelope:

{
  "error": {
    "type": "rate_limit_error",
    "code": "rate_limit_exceeded",
    "message": "Rate limit exceeded. Retry after 2 seconds.",
    "param": null,
    "request_id": "req_01HQ..."
  }
}

HTTP status → action

SchneeAI-specific error codes

The OpenAI-compatible envelope carries SchneeAI-specific sub-codes in error.code. The ones you will actually hit:

Retry strategy

Retry only on 429, 500, 502, 503, 504. Use exponential backoff with jitter:

wait = min(base * 2^attempt, max_wait) + random_jitter
  base      = 1 second
  max_wait  = 60 seconds
  jitter    = 0–25% of wait
  attempts  = max 5

Honor the Retry-After header when present. 429 responses include it as an integer number of seconds.

When not to retry

• 400 / 401 / 403 / 404 / 422 — the request itself is wrong; retrying produces the same error.
• 402 with code: insufficient_credits — budget exhausted or credits depleted. Retrying burns budget/credits faster. Wait for the reset window or top up.

Idempotency

For POST /v1/chat/completions, send an Idempotency-Key header so retries do not double-charge or double-count:

curl https://api.schneeai.com/v1/chat/completions \
  -H "Authorization: Bearer $SCHNEEAI_API_KEY" \
  -H "Idempotency-Key: $(uuidgen)" \
  -H "Content-Type: application/json" \
  -d '{ "model": "gpt-4o-mini", "messages": [...] }'

• Format: any string ≤ 255 chars. UUIDv4 recommended.
• TTL: 24 hours. After that the same key is treated as a new request.
• Scope: scoped to your service + tenant, so two tenants using the same key do not collide.
• What is deduplicated: the charge and the usage record. The actual LLM call still happens if the original request never reached the provider.

Client examples

Python (requests + tenacity)

import uuid
import requests
from tenacity import retry, stop_after_attempt, wait_exponential_jitter, retry_if_exception_type

class RetryableError(Exception): pass

@retry(stop=stop_after_attempt(5),
       wait=wait_exponential_jitter(initial=1, max=60, jitter=2),
       retry=retry_if_exception_type(RetryableError))
def chat(messages, model="gpt-4o-mini"):
    res = requests.post(
        "https://api.schneeai.com/v1/chat/completions",
        headers={
            "Authorization": f"Bearer {API_KEY}",
            "Idempotency-Key": str(uuid.uuid4()),
            "Content-Type": "application/json",
        },
        json={"model": model, "messages": messages},
        timeout=60,
    )
    if res.status_code in {429, 500, 502, 503, 504}:
        raise RetryableError(res.text)
    res.raise_for_status()
    return res.json()

TypeScript (fetch)

async function chat(messages: Msg[], model = "gpt-4o-mini") {
  const sleep = (ms: number) => new Promise(r => setTimeout(r, ms));
  for (let attempt = 0; attempt < 5; attempt++) {
    const res = await fetch("https://api.schneeai.com/v1/chat/completions", {
      method: "POST",
      headers: {
        Authorization: `Bearer ${API_KEY}`,
        "Idempotency-Key": crypto.randomUUID(),
        "Content-Type": "application/json",
      },
      body: JSON.stringify({ model, messages }),
    });
    if (res.ok) return res.json();
    if (![429, 500, 502, 503, 504].includes(res.status)) {
      const body = await res.json();
      throw new Error(`schneeai ${res.status}: ${body.error?.code}`);
    }
    const retryAfter = Number(res.headers.get("Retry-After")) || 0;
    const wait = retryAfter || Math.min(2 ** attempt + Math.random(), 60);
    await sleep(wait * 1000);
  }
  throw new Error("schneeai: max retries exceeded");
}

Go (net/http + backoff)

var buf bytes.Buffer
json.NewEncoder(&buf).Encode(map[string]any{
    "model":    "gpt-4o-mini",
    "messages": messages,
})
req, _ := http.NewRequest("POST", "https://api.schneeai.com/v1/chat/completions", &buf)
req.Header.Set("Authorization", "Bearer "+apiKey)
req.Header.Set("Idempotency-Key", uuid.NewString())
req.Header.Set("Content-Type", "application/json")

bo := backoff.NewExponentialBackOff()
bo.MaxElapsedTime = 60 * time.Second
backoff.Retry(func() error {
    resp, err := http.DefaultClient.Do(req)
    if err != nil { return err }
    defer resp.Body.Close()
    switch {
    case resp.StatusCode >= 500, resp.StatusCode == 429:
        return errors.New("retry: " + resp.Status)
    case resp.StatusCode >= 400:
        return backoff.Permanent(parseError(resp))
    }
    return json.NewDecoder(resp.Body).Decode(&result)
}, bo)

Support

If an error persists after correct retrying, email [email protected] with the request_id from the error response. We can trace the exact call within minutes.

Note: SchneeAI is in design-partner stage. The error envelope above is stable and OpenAI-compatible; the budget / PII sub-codes are aligned with the platform features currently shipping. If you hit a code not listed here, send us the request_id and we will document it.