AIARCOASC Docs
Sign inStart free

Batch

Fan-out to 1M+ containers with retries + DLQ, iterators + collectors, concurrency caps. $0.23 per 1M CPU calls.

What it is

Batch is the fan-out product. Map a callable over a (potentially huge) iterable and collect results. Built-in retries, dead-letter queue, per-tenant concurrency caps. Used for ETL, eval runs, document processing, dataset generation.

When to use it

  • Process 10M documents through a model in a few hours.
  • Run a batched eval over your test set.
  • Backfill embeddings for a 100M-row dataset.

Quickstart

asc batch submit embed-docs --image my/embed:latest --inputs-file s3://my/inputs.jsonl --concurrency 200 --retries 3
client.batch.submit(
    name='embed-docs',
    image='my/embed:latest',
    inputs_uri='asc://my/inputs.jsonl',
    concurrency=200,
    retries=3,
    dlq='embed-dlq',
)
await client.batch.submit({ name: 'embed-docs', image: 'my/embed:latest', inputsUri: 'asc://my/inputs.jsonl', concurrency: 200, retries: 3, dlq: 'embed-dlq' });

Limits & quotas

LimitDefaultBurstNotes
Per 1M CPU calls$0.23—
Max concurrency10,000100,000Lift via sales
Max retries10—Per item
DLQ retention14 days—

Pricing

See pricing. Pay-as-you-go, billed monthly via Stripe.

API surface

  • POST /v1/batch/jobs
  • GET /v1/batch/jobs/{id}/items

Required scope(s): pods:write. See Scopes.

Security

All access is authenticated and scoped. See Auth & scopes and Network controls.