Release Notes 4.2 (January 2026)
Aqua AI Agent Improvements
Aqua now features several upgrades to improve speed, accuracy, and usability:
A new RAG-based knowledge base, replacing Amazon Neptune, which provides lower operational costs, greater answer accuracy, and reduced response time.
Auto-refresh to allow users to instantly see Data Assets created by Aqua.
Starter Environments have been added to Aqua's context for more relevant and informed responses.
A new Aqua page in the Admin Panel where Code Ocean deployment admins can easily toggle access.
Public API - Create Data Assets from running Cloud Workstation
The Code Ocean Public API's Create Data Asset endpoint has been extended to allow creation of Data Assets from running Cloud Workstation sessions via an optional cloud_workstation section in the request body. You can find more information in the User Guide here.
Copy Environment from an Existing Capsule
Users can now quickly copy the environment of another Capsule, replacing the need to manually recreate the environment in the UI and rebuild the Docker image from scratch. This makes reproducible environments easier to maintain and improves the UX for teams working with established workflows.
Custom S3 Endpoints
Admins can now configure custom S3 endpoints that behave like standard AWS S3 data sources. Once a custom S3 endpoint is configured, users can create Internal Data Assets or External Data Assets (remote links) from the buckets and objects stored on the service and write External Results to it.
This provides full support for storage systems (e.g. Weka, MinIO) that expose an S3-compatible interface, offering a flexible, vendor-agnostic data integration path.
Copy from Private Git
A more flexible Copy from Git option allows users to import code from both public and private Git repositories, automatically using their stored credentials when authentication is required—mirroring the seamless experience of cloning repos today.
Expanded Asset Favoriting
Users can now add assets to Favorites directly from each asset's specific page, including Data/Model details, Capsule/Pipeline IDE headers, and Capsule/Pipeline details panels. This UX update makes it far easier to bookmark important assets right from where you’re working with them.
Assumable Roles added to Accounts Page
Users can now see all of their AWS Assumable Roles directly on their User Account Page under Roles and Secrets, making it easier to understand available permissions.
MLflow Upgrade to v3.6
Code Ocean has upgraded to MLflow v3.6, ensuring our users can stay aligned with the latest MLflow capabilities, including production-ready generative AI features.
Infrastructure Updates
Additional Flex worker instance type
R6id instance types are now supported for Flex workers, providing enhanced cost-performance. Customers are recommended to update to an R6id instance type as part of the deployment via the Worker EC2 Instance Type CloudFormation parameter, after first verifying that the selected instance type is available in their region.
New Dedicated Machine options
New dedicated machine families are now supported for users to select as Capsule compute resources:
g6e
r8i
r8i-flex
x2iedn
System Capacity Tier
A new CloudFormation parameter was added to define resource sizes for the services instance and RDS. With this change, the Services Machine EC2 Instance Type and Analytics RDS Instance Type parameters were removed.
Aqua CloudFormation parameter removed
The Aqua CloudFormation parameter was removed since Bedrock resources are no longer deployed. Aqua infrastructure is now enabled by default for all customers in all regions and UI access can still be managed from the Admin Panel.
Internal postgres service PostgreSQL version update
The internal postgres service's PostgreSQL version has been updated to v18.1 with seamless automated data migration included in the deployment upgrade.
NVIDIA driver upgrade
The NVIDIA driver was upgraded from version 525.105.17 to 580.105.08.
Code Ocean systemd service reliability improvement
The Code Ocean systemd service now includes an improved retry mechanism on failures.
Analytics materialized views refresh schedule
Refresh of analytics materialized views now runs once per day instead of every hour to improve performance. The execution time of the refresh is set based on the preferred time of the deployment's AWS Region.
Last updated