Rate Limits, Credits & Reliability

Rate limits

Two limits apply per minute: a plan-wide limit across all endpoints, and a per-endpoint limit (60 requests/minute unless stated otherwise). The one that runs out first applies. There is no separate concurrency limit.

Every API response includes these headers:

Timeouts and retries

Most calls finish in a few seconds, but a call can wait up to 20 seconds for the AI provider and then try a backup model. Set your client timeout to 60 seconds. Retry only on 429, 502, 503, 504 and network errors, and add random jitter so many clients do not retry at the same moment. Do not retry 400, 401, 402, 403 or 422: the same request will fail again.

import random
import time

import requests

RETRYABLE = {429, 502, 503, 504}

def call_rsflowhub(url, headers, payload, max_retries=4, base_delay=1.0, max_delay=16.0):
    for attempt in range(max_retries + 1):
        try:
            response = requests.post(url, headers=headers, json=payload, timeout=60)
        except requests.RequestException:
            response = None

        if response is not None and response.status_code not in RETRYABLE:
            return response
        if attempt == max_retries:
            break

        retry_after = response.headers.get("Retry-After") if response is not None else None
        if retry_after and retry_after.isdigit():
            delay = float(retry_after)
        else:
            delay = random.uniform(0, min(max_delay, base_delay * (2 ** attempt)))
        time.sleep(delay)

    raise RuntimeError("RSFlowHub request failed after retries")

Provider fallback and its limits

• Each endpoint has a primary AI model and, for most endpoints, a backup model from a different provider.
• If the primary model errors or times out, the same request is sent to the backup. If a model keeps failing, a circuit breaker pauses it for a short time and calls go straight to the backup.
• The price does not change when the backup is used: credits depend only on the endpoint and the size of your text.
• Image generation has no backup model. If its provider is down, the call fails with 502 AI_PROVIDER_ERROR and is not charged.
• If both models fail, you get 502 AI_PROVIDER_ERROR and nothing is charged. A backup model can phrase answers a little differently from the primary.
• Live provider health is on the status page.

How a call is charged

Freemium Utility API Quotas (Per Plan)

Selected utility endpoints (e.g. Sentiment Analysis, PII Redaction, Spam Detection, Intent Detection) are available completely free without deducting credits from your balance, up to a monthly global quota determined by your active plan:

The credit formula

The charge for a successful, non-cached call is:

credits = base_credits
        + ceil(input_characters / 1000)
        + ceil(output_characters / 500)   (text-generating endpoints only)
minimum 2 credits

• Input characters are the text you send. Our system prompts are never billed.
• Output characters count only for endpoints that write text (replies, summaries, chat, translation, extraction and similar). Classifiers such as sentiment or spam detection are billed on input only.
• Free freemium calls and cache hits cost 0 credits. Every charge is listed per call in Dashboard → Usage.

Endpoint Base Credits Reference