Fix Ollama API timeout for local LLM inference

LocalAgent.ApiService/Program.cs

@@ -26,6 +26,28 @@
     .UseFunctionInvocation()
     .UseOpenTelemetry(configure: t => t.EnableSensitiveData = true);
 
+// Configure extended timeouts for local LLM operations
+// The HttpClient naming convention is "{connectionName}_httpClient"
+builder.Services.AddHttpClient("llama32_httpClient")
+    .ConfigurePrimaryHttpMessageHandler(() => new SocketsHttpHandler
+    {
+        // Configure connection pooling to keep connections alive longer for LLM requests
+        PooledConnectionLifetime = TimeSpan.FromMinutes(10),
+        PooledConnectionIdleTimeout = TimeSpan.FromMinutes(5)
+    })
+    .ConfigureHttpClient(client =>
+    {
+        // Set to infinite to let the resilience pipeline handle timeouts
+        client.Timeout = Timeout.InfiniteTimeSpan;
+    })
+    .AddStandardResilienceHandler(options =>
+    {
+        // Increase total request timeout to 5 minutes for local LLM inference
+        options.TotalRequestTimeout.Timeout = TimeSpan.FromMinutes(5);
+        // Increase attempt timeout to 2 minutes, leaving room for retry logic
+        options.AttemptTimeout.Timeout = TimeSpan.FromMinutes(2);
+    });
+
 builder.Services.AddSingleton<IClientTransport>(sp =>
 {
     var mcpServerEndpoint = builder.Configuration.GetConnectionString("McpServer")!;