This release improves flattening reliability and performance, adds human-readable job names, and hardens TORCH polling.
What's Changed
Improvements
- Flattening: route resources by Provenance instead of meta.profile — Use Provenance references for more reliable resource-to-profile mapping during flattening #231
- Batched streaming for flattening — Replace load-all approach with single-pass batched streaming to prevent OOM errors on large datasets #228
- Human-readable job names — Job IDs now use
YYYYMMDD_HHMM_UUIDformat for easier identification #230
Bug Fixes
- TORCH status polling resilience — Make TORCH status polling resilient to transient HTTP errors instead of failing immediately #222
CI & Housekeeping
- Update TORCH to v1.0.0-test4
- Update Go to v1.26.1
- Update debian:bookworm-slim Docker digest
- Pin minio/minio Docker tag
- Update Vue to v3.5.30
- Update GitHub Actions (codecov, codeql, gh-release, sbom, github-script)