Join the LadVen OS testing programRequest a demo
Skip to main content

AI connections

The AI connections area lists channels used by AI scenarios. Access to view and edit is controlled by administrator rights and company settings.

Connection fields

Connection fields — conceptual guidance, not UI evidence

The list row shows the key, transport and model, vision support, and enabled state; response time appears only in the Playground result panel. The form requires both a machine key and a human-readable name. Enter the secret in the protected field; it is not shown again.

Select a connection in the scenario or keep the module default. After saving, confirm that the intended scenario uses it.

Access scope

Access scope — conceptual guidance, not UI evidence

The usage mode limits a connection to a module or company policy. Configure company, department, or user access in administrator settings; check scope, role, and permission to use it together.

Secrets and deactivation

Secrets and deactivation — conceptual guidance, not UI evidence

The provider key is not shown again. In edit mode, a blank field means “keep the existing secret,” not delete it. Never put keys, provider URLs, or real responses in documentation, demonstrations, or chat. Deactivate a temporary connection instead of deleting it so scenario history remains understandable.

Form and safe refresh

Form and safe refresh — conceptual guidance, not UI evidence

Create and edit open in a separate modal. Saving requires both a key and a human-readable name; a secret is required only when the selected provider or transport needs one. On edit, an empty secret field preserves the existing key. Check any provider address before saving. Refresh reloads the connection and reports when another administrator changed or removed it. Reset clears only the current form draft; it does not delete the connection.

Errors and verification

Errors and verification — conceptual guidance, not UI evidence

Account for loading, missing rights, changes by another administrator, and provider failure. Before using a scenario, run an anonymized test that verifies the model, delay, vision, and effective access. Keep real secrets and money out of demonstrations.

Related: AI Playground, AI scenarios, AI and Assistant settings.