> ## Documentation Index
> Fetch the complete documentation index at: https://docs.findly.icu/llms.txt
> Use this file to discover all available pages before exploring further.

# Rate limits

> Per-address and per-account limits, size limits and how to retry.

## Limits

| Limit                      | Scope                                                 | Error                                      | `Retry-After`               |
| -------------------------- | ----------------------------------------------------- | ------------------------------------------ | --------------------------- |
| 60 calls per minute        | IP address, all `/api/v1` endpoints, valid key or not | `429 too_many_requests`                    | Until a slot frees up       |
| 2 searches running at once | Account (dashboard and API together)                  | `429 rate_limited`, `reason: "concurrent"` | 3 seconds                   |
| 20 searches per minute     | Account (dashboard and API together)                  | `429 rate_limited`, `reason: "per-minute"` | Until a slot frees up       |
| Daily quota                | Account                                               | `429 quota_exceeded`                       | Until the 02:00 Paris reset |

Minute limits use a sliding 60-second window. Refused calls are never billed and do not count toward the window that refused them.

In order, a search is checked for: IP limit, key and plan, module, format, body, input validity, concurrency, per-minute limit, then quota. So an invalid body does not use a per-minute slot, but a valid search refused for quota does.

## Retrying

Every `429` sends `Retry-After` in whole seconds (at least 1). Wait that long, then retry once; do not retry in a tight loop.

```python theme={null}
import time
import requests

def search(module, body, key, attempts=3):
    for _ in range(attempts):
        response = requests.post(
            f"https://findly.icu/api/v1/search/{module}",
            headers={"Authorization": f"Bearer {key}"},
            json=body,
            timeout=60,
        )
        if response.status_code != 429:
            return response
        code = response.json()["error"]["code"]
        if code == "quota_exceeded":
            return response  # Waiting until 02:00 Paris is rarely what you want.
        time.sleep(int(response.headers.get("Retry-After", "5")))
    return response
```

To run many searches, keep at most **two in flight** and stay under **20 per minute** per account. Remember that searches made in the dashboard at the same time count too.

## Size and time limits

| Limit                           | Value      | What happens                                                                                                                                                        |
| ------------------------------- | ---------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Request body                    | 16 KB      | `413 body_too_large`                                                                                                                                                |
| Items in one list response      | 5,000      | The list is cut; `truncated: true`, `total` keeps the real count                                                                                                    |
| Raw file delivered              | 8 MB       | The text is cut; `file.truncated: true`                                                                                                                             |
| Search response read by Find.ly | 16 MB      | Lists: `502 upstream_error` (refunded). Files: `502 upstream_error` (refunded) when the size is known in advance, otherwise the first 8 MB with `total_bytes: null` |
| Search duration                 | 30 seconds | `502 upstream_error` (refunded)                                                                                                                                     |

Set your HTTP client timeout above 30 seconds (60 is a good value), or a slow search may be billed on our side while your client has already given up.