feat: wire prune_releases.py into publish_muse_release.sh (issue #128, Phase 2)
S3 cleanup now pipes candidate filenames through prune_releases.py instead of sort -V | awk. Server cleanup restructured into two SSM round-trips: list files remotely, compute the stale set locally (same source of truth as the S3 step -- Python/packaging availability isn't guaranteed on the bare EC2 host), then send an explicit rm of the exact stale filenames. The remote shell never sorts or re-implements precedence logic.
PRUNE_06/06b/06c/07/07b/07c verified via source-level assertions against the script (matching test_deployment.py's existing convention) rather than live AWS calls -- live verification against the real staging bucket/ instance is Phase 3, a deliberate action requiring explicit go-ahead, not something a test suite should trigger. All 17 tests across both files green; bash -n syntax check passes.
Semantic Changes
26 symbols
0 comments
muse hub commit comment sha256:893883a1ec8d2317de0bdc78df5c1522ef0bacbc3bdacf4a005be8f391484d0b --body "your comment"
No comments yet. Be the first to start the discussion.