Interface AIFactoryConfig

Max provider calls in flight at once across this factory; extra requests queue. A stream holds its slot until it ends. Default: unlimited.

Preferred provider when no modelId is specified

Fallback provider if default fails

Fallback providers, tried in order after fallbackProvider.

Observe every attempt: called before each provider call, after each answer, and on each failure. Awaited; a throw propagates to the caller.

Optional logger; if not set, no logging

Order of providers to try when no default/model match (first available wins)

Custom provider instances; if set, only these are used (no built-in list)

Max retries per request on a retryable failure (default 2). Shorthand for retry.retries.

Backoff settings; see RetryOptions.

Default idle timeout (ms) for streams; per-request streamIdleTimeout wins.

Default whole-request timeout (ms) for non-streaming calls; per-request timeout wins.