Comparison
Canary deploymentvsPost-deploy verification
Canary deployment
the new prompt goes to two percent of traffic first, and you watch quality and cost before letting the rest through.
Rolling a prompt or model change to a small traffic slice first, watching quality and cost metrics, then widening. Treat prompts as deployable artefacts with the same ceremony as code, because they break things the same way.
Full entry →Post-deploy verification
the pipeline hits three real endpoints after the rollout and fails the deploy if any of them is unhappy.
Automated checks run against the deployed system immediately after a release, distinct from the tests that ran against the artefact. They catch the class of failure that only exists in the real environment: missing configuration, absent permissions, an unreachable dependency. They are the precondition for automated rollback, because something has to decide the deploy failed.
Full entry →