Python: Fix: Skip model_deployment_name validation for application endpoints (Azure new foundary agents)

[Ravinduprasath]

Summary

• Skip model_deployment_name validation when using application endpoints in AzureAIClient. Azure new foundary agents

Problem

When using AzureAIClient with an application endpoint (pre-configured Foundry agent), the framework incorrectly requires model_deployment_name even though:

1. The model is already pre-configured on the server for the application
2. The model field gets stripped from run_options anyway for application endpoints (lines 404-417)

This forces users to provide a dummy/placeholder value like model_deployment_name="placeholder" which is confusing and unnecessary.

Error before fix:
ValueError: model_deployment_name must be a non-empty string

Solution

Added a check in _check_model_presence() to skip validation when _is_application_endpoint is True:

@override
def _check_model_presence(self, run_options: dict[str, Any]) -> None:
    # Skip model check for application endpoints - model is pre-configured on server
    if self._is_application_endpoint:
        return
    if not run_options.get("model"):
        if not self.model_id:
            raise ValueError("model_deployment_name must be a non-empty string")
        run_options["model"] = self.model_id
        
        
Test plan

Tested with existing Foundry agent via application endpoint
Verified agent responds correctly without model_deployment_name

[markwallace-microsoft]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/azure-ai/agent_framework_azure_ai
_client.py	195	13	93%	345, 347, 395, 424–429, 472, 509, 511, 587
TOTAL	16252	1984	87%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
3742	221 💤	0 ❌	0 🔥	1m 7s ⏱️