Govern models, apps, agents, and files

Governing an asset links it to a bundle, subjecting it to your organization’s review and approval workflows. Depending on the asset type, governance may be required before deployment, publishing, or use in validated workflows.

Governance workflow

The workflow is the same for all asset types:

  1. Link the asset: Navigate to the asset and click Add to Bundle to link it to a new or existing bundle. Confirm the link in the bundle’s Attachments section.

  2. Collect evidence: Provide the documentation and artifacts required by your policies. Evidence types vary by asset.

  3. Obtain approvals: Request reviewer approvals to reach compliance.

Models

From the Model card or Model endpoint version page, click Add to Bundle to link the model.

For models, you can use two specialized evidence types:

  • Model metrics: Capture performance data such as accuracy, RMSE, and custom KPIs.

  • Metrics checks: Define pass/fail thresholds that automatically validate performance and auto-create findings.

Tip
 You can use gates to block Model endpoint deployment when a bundle is non-compliant.

Apps and agents

From the app or agent page, click Add to Bundle to link the asset.

For apps and agents, evidence typically covers security and data privacy:

  • Scripted checks Run automated tests to verify dependencies, check for vulnerabilities, or validate data access permissions.

  • Manual evidence: Provide screenshots, UAT sign-offs, or data source documentation.

Tip
 You can use gates to block app publishing or viewing when a bundle is non-compliant.

Files

From the Code section of your Project, locate the file. Click the three dot-menu, then select Add to Bundle. Confirm the link in the bundle’s Attachments section.

For files, evidence typically focuses on validation and traceability. This is particularly relevant in Life Sciences, where scripts, data pipelines, and clinical reports must meet GxP standards before use in validated studies or production workflows.

  • Scripted checks: Run automated tests to verify data integrity, perform checksums, or validate formatting standards such as CDISC.

  • Manual evidence: Upload code review logs, validation plans, or QA reports.

Next steps