Azure AI Foundry

Connect an agent on Azure AI Foundry to AI GO! as a native model, using the typed azure_foundry connection. The integration authenticates with an Azure service principal, creates a conversation on the first turn, and round-trips the conversation_id automatically, so multi-turn evaluations preserve state.

Prerequisites

• An agent deployed on Azure AI Foundry, its project endpoint URL, and the agent name.
• An Azure service principal (tenant ID, client ID, client secret) with access to the project.

Configuration

Define the model with connection_type: azure_foundry and reference the service-principal credentials as secrets:

key: "azure-foundry-agent"
display_name: "Azure AI Foundry Agent"
description: "Agent on Azure AI Foundry, via the typed `azure_foundry` connection config."
task: "chat_completion"
rate_limit: 15
config:
  connection_type: "azure_foundry"
  project_endpoint: $AZURE_AI_PROJECT_ENDPOINT
  agent_name: $AZURE_FOUNDRY_AGENT_NAME
  tenant_id: "<< secrets.AZURE_TENANT_ID >>"
  client_id: "<< secrets.AZURE_CLIENT_ID >>"
  client_secret: "<< secrets.AZURE_CLIENT_SECRET >>"
secrets:
  AZURE_TENANT_ID: $AZURE_TENANT_ID
  AZURE_CLIENT_ID: $AZURE_CLIENT_ID
  AZURE_CLIENT_SECRET: $AZURE_CLIENT_SECRET

Config fields

Field	Required	Secret	Description
connection_type	yes	no	Must be azure_foundry.
project_endpoint	yes	no	Endpoint URL of the Azure AI Foundry project.
agent_name	yes	no	Name of the agent to invoke.
tenant_id	yes	yes	Service-principal tenant ID.
client_id	yes	yes	Service-principal client ID.
client_secret	yes	yes	Service-principal client secret.

Credentials

The secrets block uploads values (from your .env or environment) to server-side secret storage, and << secrets.NAME >> references them from the config so no secret is stored in plaintext. Provide the values in your .env:

AZURE_AI_PROJECT_ENDPOINT=https://my-project.services.ai.azure.com/api/projects/my-project
AZURE_FOUNDRY_AGENT_NAME=my-agent
AZURE_TENANT_ID=...
AZURE_CLIENT_ID=...
AZURE_CLIENT_SECRET=...

Add and test

lf add model -f azure_foundry.yaml
lf test model azure-foundry-agent

Use in an evaluation

Reference the model key from a task specification in your run config:

evaluation:
  task_specifications:
    - task_key: my-task
      model_key: "azure-foundry-agent"

Conversation state

On the first turn the integration creates a conversation; the conversation_id is attached to the assistant message and reused on each subsequent turn, so the agent remembers earlier turns.