Interface: ProviderBatchOptions

Defined in: packages/core/src/adapters/ai-sdk/provider-batch.ts:152

Configuration for batch({ providerBatch }).

Properties

fallback?

optional fallback?: "error" | "live"

Defined in: packages/core/src/adapters/ai-sdk/provider-batch.ts:173

What a call does when its provider has no batch endpoint, or refuses the submission as unsupported: 'live' (default) makes a normal call at the standard price and emits a fallback event; 'error' fails it.

---

maxRequests?

optional maxRequests?: number

Defined in: packages/core/src/adapters/ai-sdk/provider-batch.ts:167

Most requests in one provider batch. A full queue is submitted at once. Default 10 000 — below every provider's count limit, and small enough that typical requests stay under their payload-size limits.

---

maxWaitMs?

optional maxWaitMs?: number

Defined in: packages/core/src/adapters/ai-sdk/provider-batch.ts:161

Longest a queued call waits for work that is busy elsewhere — running a tool, making a live call — before the queue is submitted without it. When every item is waiting on a batch, the queue goes out at once. Default 2 000 ms.

---

onEvent?

optional onEvent?: (event) => void

Defined in: packages/core/src/adapters/ai-sdk/provider-batch.ts:175

Batch lifecycle events. A throwing listener does not affect the run.

Parameters

event

ProviderBatchEvent

Returns

void

---

pollIntervalMs?

optional pollIntervalMs?: number

Defined in: packages/core/src/adapters/ai-sdk/provider-batch.ts:154

How often to ask the provider whether a submitted batch has finished. Default 15 000 ms.