For the complete documentation index, see llms.txt. This page is also available as Markdown.

Release Notes 4.5 (May 2026)

Single-Machine Pipeline Execution

Pipelines can now be executed on a single machine in both Reproducible Runs (RR) and Cloud Workstations (CW), improving the developer experience by enabling faster iteration and more efficient pipeline testing

  • Run without AWS batch infrastructure — Ideal for development, testing, and quick iteration.

  • Faster startup times — Eliminate delays from infrastructure spin-up required for batch execution.

Compute resources can be configured in Pipeline Settings.

Nextflow v25 Support

Pipelines now support Nextflow v25.10.4 (latest), which introduces improvements to language validation and syntax handling, enhanced AWS Batch reliability, and multiple bug fixes for more stable pipeline execution. This ensures compatibility with modern workflows (e.g., nf-core pipelines) and keeps users aligned with the latest Nextflow capabilities.

Nextflow v25.10.4 will also be the default Nextflow version for all new Pipelines.

Hugging Face Credential Support

Users can now add Hugging Face user credentials and use them when importing models.

  • Hugging Face credentials support — Add your credentials to access gated resources.

  • Automatic credential detection — The system checks for credentials when importing from Hugging Face and uses them if available.

This enables seamless access to gated Hugging Face models that require authentication and agreement to provider terms.

Improved Collections Search Defaults

Search behavior has been updated to provide more intuitive default scopes across Collections:

  • My Organization defaults to Internal Collections

  • Public Collections defaults to Public Collections

  • Nested Collections default to Current Collection

This helps users find assets more easily by reducing empty search results caused by narrow default scopes.

Infrastructure Updates

Resource Watchdog

The services machine now reports CloudWatch metrics for all containers and CouchDB, to act as a memory safeguard that enhances overall system stability.

The existing top.py script, which monitored CPU and memory usage of OS processes on the services machine every minute and reported the top 7 processes to CloudWatch, has been replaced with a new comprehensive system monitor and memory safeguard called watchdog.py that runs on all Code Ocean EC2 instances and provides full visibility of the system by monitoring the following:

  • The top 25 processes

  • All running containers

  • Various OS parameters

RDS Postgres Upgrade

The RDS Postgres engine version was upgraded from version 17 to 18.

This change may slightly increase the duration of deployment updates.

OpenSearch Automatic Software Update

The OpenSearch configuration was changed to allow automatic software updates, ensuring it stays up to date with the latest features and security patches.

Internal Docker Registry Upgrade

The internal Docker registry was upgraded from version 2.7.0 to version 3.1.0, bringing enhanced performance, and improved security and maintainability.

Aqua Model Upgrade

The model used by the Aqua AI Agent has been upgraded from Claude Sonnet 4.5 to Claude Sonnet 4.6, bringing improvements in performance, accuracy, and overall model capabilities.

Source AMI

al2023-ami-2023.11.20260413.0-kernel-6.1-x86_64

Last updated