The AI Services page gives administrators a single control panel for runtime availability and scheduler visibility.

Use this page when you need to:

  • quickly enable or disable AI runtime processing,
  • confirm whether the override is persisted in admin settings,
  • inspect provider/configuration readiness,
  • review scheduler profile values used by the AI worker.

Open the page

  1. Sign in with an administrator account.
  2. Navigate to Admin → AI Services.
  3. Review the top summary card and status sections.

Runtime control

  • AI runtime enabled toggle stages the runtime state locally.
  • Select Save AI settings to persist the runtime override.
  • Use Refresh status to reload live status from the server.

Runtime changes are stored in admin settings and survive page refreshes.

AI runtime status

The status card reports:

  • whether an override is stored,
  • current runtime override value,
  • startup default value,
  • effective runtime value,
  • right rail assistant feature flag,
  • active conversation provider,
  • provider credential readiness (Azure OpenAI / Azure AI Foundry).

This helps quickly distinguish between feature disabled, provider not selected, and configuration missing states.

AI scheduling profile

The scheduling profile displays compact metrics with units in each title:

  • Retention (days)
  • Polling interval (seconds)
  • Max backoff (seconds)
  • Batch size (tasks/cycle)
  • Max retries (attempts)
  • Retry delay (milliseconds)

These values are read from environment configuration and shown for operational transparency.

Scheduling options block

The Scheduling options area exposes the same values in a read-only form layout, making it easier to compare and discuss candidate tuning values.

Current behavior:

  • values are informational,
  • effective changes are managed through environment configuration,
  • updated values apply after service restart.
  1. Confirm provider and runtime status are healthy.
  2. Stage runtime toggle if needed.
  3. Save AI settings.
  4. Refresh status and verify Effective runtime.
  5. Capture scheduler values for planned tuning in deployment/configuration tickets.