Fix local Ollama timeout by removing global resilience handler from client

LocalAgent.ApiService/Program.cs

@@ -29,7 +29,33 @@
 
 if (aiConfig.IsLocalProvider())
 {
-    builder.AddOllamaApiClient("ollamaModel")
+    // The connection name used for Ollama
+    const string OllamaConnectionName = "ollamaModel";
+
+    // Configure increased timeout for local Ollama LLM requests BEFORE creating the client
+    // Local LLMs can be slower, especially on low-performance machines
+    // The HttpClient name follows the pattern: {connectionName}_httpClient
+    const int MinimumTimeoutSeconds = 30; // Minimum 30 seconds for LLM operations
+    var timeoutSeconds = Math.Max(MinimumTimeoutSeconds, aiConfig.TimeoutSeconds);
+
+    // IMPORTANT: We do NOT use AddStandardResilienceHandler for this client because:
+    // 1. The global AddStandardResilienceHandler cannot be effectively overridden per-client
+    // 2. It would stack with the global handler causing whichever timeout is shorter to win
+    // 3. For streaming LLM responses, we need a much longer timeout than the default 10 seconds
+    builder.Services.AddHttpClient($"{OllamaConnectionName}_httpClient", client =>
+    {
+        // Set a long timeout for LLM operations (streaming responses can take a while)
+        client.Timeout = TimeSpan.FromSeconds(timeoutSeconds);
+    })
+    // Remove any additional handlers that were added by ConfigureHttpClientDefaults
+    .ConfigureAdditionalHttpMessageHandlers((handlers, _) =>
+    {
+        // Clear the standard resilience handler that was added by defaults
+        // This allows us to have a custom timeout for Ollama without interference
+        handlers.Clear();
+    });
+
+    builder.AddOllamaApiClient(OllamaConnectionName)
         .AddChatClient()
         .UseFunctionInvocation()
         .UseOpenTelemetry(configure: t => t.EnableSensitiveData = true);

LocalAgent.AppHost/appsettings.json

@@ -11,6 +11,7 @@
   },
   "AIConfig": {
     "Provider": "Local",
-    "ModelId": "llama32"
+    "ModelId": "llama32",
+    "TimeoutSeconds": 90
   }
 }

LocalAgent.ServiceDefaults/AIConfig.cs

@@ -12,6 +12,14 @@ public class AIConfig
 
     public string ModelId { get; set; } = string.Empty;
 
+    /// <summary>
+    /// Timeout in seconds for HTTP requests to the AI service.
+    /// Default is 90 seconds to accommodate slower local LLMs while maintaining reasonable responsiveness.
+    /// Note: This timeout applies specifically to HTTP client requests. Some AI providers may have
+    /// their own internal timeout mechanisms that are independent of this setting.
+    /// </summary>
+    public int TimeoutSeconds { get; set; } = 90;
+
     /// <summary>
     /// Determines if the Azure provider is configured
     /// </summary>