- 00
- Features
- Auto Fallback
A provider fails. The next one solves.
When a provider is out of attempts, the task moves down the order your routing mode built. On for every new API key.
Resilience
- On a new key
- On
- Fallbacks by default
- 2
- Most per task
- 20
- Billed per task
- Once
01Order
The order is your routing mode.
Excluded providers are dropped first. A pinned provider has no next one, so fallback is off.
02Config
On by default, tunable per request.
settings.fallback, in POST /createTask
{
"settings": {
"routing": {
"mode": "autoCheapest"
},
"fallback": {
"enabled": true,
"maxFallbacks": 2
}
}
}| Field | Range | Default |
|---|---|---|
| enabled | true, false | true on a new key |
| maxFallbacks | 0 to 20 | 2 |
| onErrors | Any trigger code | The 11 below |
Auto Retry runs first on each provider. One charge follows the task and settles on the provider that solved it. If the task ends in failure, the last provider and error decide whether the charge is refunded.
03Triggers
Everything that retries, plus 3 that only another provider can fix.
Fallback triggers
- CAPTCHA_
UNSOLVABLE - TASK_
TIMEOUT - SERVICE_
UNAVAILABLE - PROVIDER_
UNREACHABLE - PROVIDER_
RATE_ LIMITED - PROXY_
CONNECTION_ FAILED - PROXY_
BANNED - ERROR_
PROVIDER_ ZERO_ BALANCE - PROVIDER_
GLOBAL_ LIMIT_ REACHEDFallback only - PROVIDER_
IP_ BLOCKEDFallback only - PROVIDER_
PAUSEDFallback only
04FAQ
Questions about Auto Fallback.
01
How is the fallback order decided?
It is your routing mode applied to every eligible provider: cheapest by price, fastest by p90 solve time, most reliable by success rate, or priority by the order you set.
02
Can I fall back from a pinned provider?
No. Forcing a single provider with settings.routing.provider disables fallback, because there is no next candidate to move to.