rp serverless batch

Submit large sets of inference requests to an endpoint as one managed batch (beta).

rp serverless batch <verb> [flags]

Notes

A batch is endpoint-scoped and rides the control plane (unlike run, which targets the data plane). Lifecycle: DRAFT → FINALIZED → FAILED|CANCELLED — work starts only after finalize, there is no RUNNING/COMPLETED status, and completion is inferred when completed + failed counts equal the total. Batch workers are isolated from /run traffic, so a batch never delays interactive jobs. Batches are addressed by id (display names are not unique). Sub-verbs: list, create, add, remove, finalize, cancel, get, requests, update.

API: /v2/{endpoint_id}/batch

Commands

• rp serverless batch list — List an endpoint's batches, newest first (beta).
• rp serverless batch create — Create a new DRAFT batch (beta).
• rp serverless batch add — Append requests to a DRAFT batch (beta).
• rp serverless batch finalize — Lock a DRAFT batch and make it eligible for execution (beta).
• rp serverless batch cancel — Cancel a batch (beta).
• rp serverless batch remove — Remove one request from a DRAFT batch (beta).
• rp serverless batch get — Show a batch's progress summary (beta).
• rp serverless batch requests — List a batch's child requests with per-request status and results (beta).
• rp serverless batch update — Rename a batch (beta).