Agent Configuration

For AI agents to function in a specific environment, every provider used in your project must be configured with valid credentials.

This configuration is managed in the DevOps portal on a per-portal, per-environment basis; meaning you will need to set it up separately for each environment (e.g. staging and live).

On a Shared Runtime currently only WEM Support can setup the AI Agent Configuration.

You need to provide the information in a support ticket and WEM Support will setup the configuration.

Check the required settings per Provider below.

If you feel hesitant about sharing all details in a single Support Ticket, you can send the API key via email to [e-mail] and include either ticket id or project id in the message so it can be linked to the correct application.

You should also mention specifically Staging and/or Live if different settings need to be used in either case, or if one setting can be used for both runtimes.

WEM Support will check which Runtime your project is currently using from the settings in Modeler and apply the settings there.

Opening the configuration

In the DevOps portal, navigate to your portal and open the AI agent configuration panel. Any providers already configured are listed here. Click + Add configuration to add a new provider.

Adding a provider

Clicking + Add configuration opens a Choose a provider dialog with the currently available model providers. Each provider has its own set of required credentials, and each provider will have its own authentication options. There can be multiple providers configured at the same time, allowing you to use different agents from different providers in the same project.

Notes

• The same provider configuration can be shared across multiple agents within the same portal.

• Configuration must be set up per environment, a configuration added to staging is not automatically available on live.

• This configuration is only accessible to users with DevOps portal access. If you need a provider configured, contact your portal administrator.

• The model selection in the modeler chooses the credentials used, ensure developers are aware of the cost implications.

• Below per provider the necessary information to setup in DevOps (or ask WEM Support to setup in DevOps in case your project runs in a Shared Runtime).

Anthropic

• API Key only

Azure OpenAI

• Resource Name (ending in .openai.azure.com)

and

• Authentication Mode API Key:

  • API Key

• Authentication Mode Client Secret:

  • Tenant ID

  • Client ID

  • Client Secret

Cerebras

• API Key only

Google

• Gemini Developer API

  • API Key

• Vertex AI

  • Project

  • Location

  • Service Account Key

OpenAI

• Organization ID

• Project ID

• API Key