v1.8.3
2026-04-10
Fixed
- Fixed GPU allocation race on multi-tenant clusters
- Pipeline logs now stream reliably on long-running jobs
Changelog
Release notes and product updates.
v1.8.0
2026-03-18
Added
- Shadow deployment mode for real-time models
- Native support for A100 and H100 GPU families
Changed
- Improved pipeline DAG resolution performance by 2.5x
v1.7.0
2026-02-12
Added
- PSI-based drift detection built into model registry
- Cost reports exported to S3 in CSV/Parquet
v1.6.0
2026-01-08
Added
- BYOC support for AWS (GCP and Azure coming)
- Prometheus metrics scraping
v1.5.0
2025-11-22
Added
- Feature store (beta) with offline/online parity
Fixed
- Fixed memory leak in long-running training jobs
v1.4.0
2025-10-05
Added
- Model promotion workflow with approval gates
- Audit log for compliance (Enterprise)
v1.3.0
2025-08-14
Added
- Pipeline caching across runs (up to 90% cost reduction)
v1.2.0
2025-07-01
Added
- Multi-region support (eu-west-1 + eu-north-1)
- GDPR-compliant data residency controls
v1.1.0
2025-05-20
Added
- Jupyter notebook integration
- Native HuggingFace Hub import
v1.0.0
2025-04-12
Added
- Initial production release
- Pipeline orchestration, model registry, experiment tracking