Managing Context Platform UI Visibility: Hiding Evals and Agents While Enabling Context Feedback – DataHub Customer Support Portal

Area: UI Issues

Sub-Area: AI Feature Visibility and Navigation Controls

Issue

DataHub Cloud administrators may need to selectively show or hide AI-related navigation items — specifically Agents and Evals — in the sidebar UI, while simultaneously enabling other Context platform features such as Context Feedback. These sections can appear in the UI when a tenant is moved onto a Context beta build or when platform-level or release-level defaults enable the underlying feature flags. Understanding which flags and platform privileges control each item is essential for administrators who want to tailor the UI experience for their users.

You Might Be Asking

How do I hide the Agents and/or Evals sections from the DataHub sidebar navigation?
What feature flags control visibility of Agents, Evals, and Context Feedback in the UI?
Will hiding Evals also hide Context Feedback or Context Documents?
Who can see the Context Feedback page, and how do I control that access?
Can I hide Agents and Evals without a deployment change, using only permissions?

Solution

Visibility of the Agents, Evals, and Context Feedback navigation items is controlled by a combination of environment-variable feature flags and platform privileges. You can use either or both mechanisms depending on your needs.

Understanding the Visibility Logic

The sidebar navigation (governed by the useNavSidebarCloudItems component) applies the following rules:

Agents is shown only when both of the following are true:
- The aiAgentsEnabled flag is true (environment variable: AI_AGENTS_ENABLED)
- The user has the manageAgents platform privilege
Evals is shown only when both of the following are true:
- The showContextHub flag is true (environment variable: SHOW_CONTEXT_HUB)
- The user has the manageEvals platform privilege
The Evals tab on the Agent detail page is separately controlled by aiAgentEvalsEnabled (environment variable: AI_AGENT_EVALS_ENABLED)
Context Feedback is controlled independently and is visible to users who hold the manageContextFeedback platform privilege. It is documented under Validation > Feedback in the navigation.

Option 1: Hide Items via Feature Flags (Recommended — Affects All Users)

Set the following environment variables in your tenant deployment configuration. This approach hides the pages for all users regardless of their privileges and is the most thorough solution.

# Hide the Agents navigation item
AI_AGENTS_ENABLED=false

# Hide the Evals navigation item and Context settings page
SHOW_CONTEXT_HUB=false

# Also remove the Evals tab on the Agent detail page
AI_AGENT_EVALS_ENABLED=false

Apply these values in your tenant's deployment values file, then roll out the change through your standard deployment process. A pod restart is required for the changes to take effect.

Option 2: Hide Items via Platform Privileges (No Deployment Required)

If a deployment change is not immediately feasible, you can hide the navigation items for specific users by removing the relevant platform privileges from their roles and policies. Navigate to Settings > Permissions > Policies and remove the following privileges from any roles that should not see these items:

Remove manageAgents to hide the Agents navigation item
Remove manageEvals to hide the Evals navigation item

Note: This approach is user-scoped. Any user or role that retains these privileges will still see the navigation items. Feature flags are the preferred solution for instance-wide suppression.

Enabling Context Feedback Independently

Hiding Evals and Agents does not affect the Context Feedback feature. Context Feedback has its own enablement setting. To enable it:

Navigate to Settings > AI in the DataHub UI to confirm that feedback collection is enabled for your instance.
Verify that the relevant platform policy grants the manageContextFeedback privilege to the appropriate roles (e.g., Admin, Editor).
To extend visibility to additional users, add the manageContextFeedback privilege to a platform policy under Settings > Permissions > Policies and assign it to the desired groups or users.

Once enabled, Context Feedback appears under Validation > Feedback (or as a Feedback item within the Context section, depending on which other Context features are active).

Feature Flag Reference Summary

| Feature               | Environment Variable      | Platform Privilege        |
|-----------------------|---------------------------|---------------------------|
| Agents (nav item)     | AI_AGENTS_ENABLED         | manageAgents              |
| Evals (nav item)      | SHOW_CONTEXT_HUB          | manageEvals               |
| Evals (agent detail)  | AI_AGENT_EVALS_ENABLED    | —                         |
| Context Feedback      | (separate enablement)     | manageContextFeedback     |
| Ask DataHub           | SHOW_ASK_DATAHUB          | —                         |

Additional Notes

The Cloud defaults for AI_AGENTS_ENABLED and SHOW_CONTEXT_HUB are false. If these items appear in your UI despite no explicit override in your tenant values file, a platform-level or release-level default may be enabling them — check the rendered environment of your running GMS and frontend pods to confirm the active values.
Disabling SHOW_CONTEXT_HUB (Evals) also removes the Context settings page under Platform Settings. It does not affect Context Documents.
Ask DataHub and custom agents report feedback gaps automatically under Settings > AI. Third-party MCP clients only report gaps if their instructions are explicitly configured to do so.
These controls apply per-tenant. Changes to one environment (e.g., production) do not affect other environments (e.g., staging) unless the same configuration is applied there.
This behavior is specific to DataHub Cloud deployments. Self-hosted DataHub configurations may differ.