Migrate from Google Cloud Batch

GCP Batch (and Vertex AI batch prediction) handles general-purpose batch compute. For the inference subset — embeddings, transcription, OCR, image captioning — Common Compute is a drop-in replacement at materially lower cost.

Concept mapping

• TaskGroup → workload
• JobSpec → submit() body
• Allocation policy → none (we own the fleet)
• GCS input → R2 presigned upload
• Service account → API key (no IAM bindings to manage)
• Vertex AI batch prediction → POST /v1/jobs with workload_id

Migration code

from commoncompute import Client

cc = Client()
job = cc.jobs.submit(
    workload_id="coreml_embed",
    model_id="bge-large",
    payload={"texts": [...]},
    priority="batch",
    max_spend_usd=50.00,
)