問題文
An end-to-end pipeline builds an embedding worker image, deploys it to a container app that reads from a queue and writes vectors to a database. Where in this pipeline should the image reference be resolved to a digest so that the running workload is auditable?
選択肢
- At the point where the worker reads from the queue, since resolving it there records the content that actually processed each message
- At the point where the deployment is created, so the deployment records the exact content it runs
- At the point where the image is pushed to the registry, because the push is what produces the digest and therefore fixes what will run
- At the point where the vectors are written, so that each stored vector carries the image version that produced it and the data itself becomes the audit trail