fix: switch to gpt-4o for azure_ai_search tool support

- gpt-5-mini doesn't support azure_ai_search tool in Persistent Agents
- Update model deployment from gpt-5-mini to gpt-4o
- Update default model config in appsettings and FoundryOptions
- Fix foundry_endpoint output to use project URL format

Author: mjhoffmeister

infra/terraform/foundry.deployments.tf

@@ -1,9 +1,9 @@
 # Azure AI Foundry Model Deployments for Health Plan Chat
 
-# GPT-5-mini deployment for chat completions
-resource "azapi_resource" "gpt5_mini_deployment" {
+# GPT-4o deployment for chat completions (supports azure_ai_search tool)
+resource "azapi_resource" "gpt4o_deployment" {
   type      = "Microsoft.CognitiveServices/accounts/deployments@2024-10-01"
-  name      = "gpt-5-mini"
+  name      = "gpt-4o"
   parent_id = azapi_resource.ai_services.id
 
   body = {
@@ -14,8 +14,8 @@ resource "azapi_resource" "gpt5_mini_deployment" {
     properties = {
       model = {
         format  = "OpenAI"
-        name    = "gpt-5-mini"
-        version = "2025-08-07"
+        name    = "gpt-4o"
+        version = "2024-11-20"
       }
       raiPolicyName = "Microsoft.DefaultV2"
     }
@@ -49,7 +49,7 @@ resource "azapi_resource" "embedding_deployment" {
   }
 
   # Ensure deployments are created sequentially to avoid conflicts
-  depends_on = [azapi_resource.gpt5_mini_deployment]
+  depends_on = [azapi_resource.gpt4o_deployment]
 
   timeouts {
     create = "30m"

infra/terraform/outputs.tf

@@ -32,8 +32,8 @@ output "redis_hostname" {
 }
 
 output "foundry_endpoint" {
-  description = "Azure AI Foundry endpoint"
-  value       = azapi_resource.ai_services.output.properties.endpoint
+  description = "Azure AI Foundry project endpoint for Persistent Agents API"
+  value       = "https://${azapi_resource.ai_services.name}.services.ai.azure.com/api/projects/${azapi_resource.foundry_project.name}"
 }
 
 # GitHub Actions WIF outputs

src/backend/HealthPlanChat.Infrastructure.AgentFramework/FoundryOptions.cs

@@ -18,7 +18,7 @@ public sealed class FoundryOptions
     /// <summary>
     /// Model deployment name for chat completions. Default: "gpt-5-mini".
     /// </summary>
-    public string ChatModelDeployment { get; set; } = "gpt-5-mini";
+    public string ChatModelDeployment { get; set; } = "gpt-4o";
 
     /// <summary>
     /// Model deployment name for embeddings. Default: "text-embedding-3-small".

src/backend/HealthPlanChat.WebApi/appsettings.Development.json

@@ -41,7 +41,7 @@
 
   "Foundry": {
     "Endpoint": "https://<your-foundry-endpoint>.cognitiveservices.azure.com",
-    "ChatModelDeployment": "gpt-5-mini",
+    "ChatModelDeployment": "gpt-4o",
     "EmbeddingModelDeployment": "text-embedding-3-small",
     "MaxTokens": 1024,
     "Temperature": 0.7

src/backend/HealthPlanChat.WebApi/appsettings.json

@@ -31,7 +31,7 @@
 
   "Foundry": {
     "Endpoint": "",
-    "ChatModelDeployment": "gpt-5-mini",
+    "ChatModelDeployment": "gpt-4o",
     "EmbeddingModelDeployment": "text-embedding-3-small",
     "MaxTokens": 1024,
     "Temperature": 0.7