Enhance image compression handling and error logging

BackendSchema.cs

@@ -1,4 +1,5 @@
 using SixLabors.ImageSharp.Processing;
+using SwarmUI.Media;
 using SwarmUI.Utils;
 using SwarmUI.Media;
 using Image = SwarmUI.Utils.Image;
@@ -67,10 +68,15 @@ public static string CompressImageForVision(MediaContent media, string targetFor
         }
         try
         {
-            ImageFile image = ImageFile.FromDataString($"data:{media.MediaType};base64,{media.Data}");
-            // Skip compression for videos etc..
-            if (image.Type.MetaType != MediaMetaType.Image)
+            Image image = Image.FromDataString($"data:{media.MediaType};base64,{media.Data}") as SwarmUI.Utils.Image;
+            if (image == null)
+            {
+                Logs.Error("Failed to parse image data.");
+                return media.Data;
+            }
+            if (image.Type.MetaType != MediaMetaType.Image && image.Type.MetaType != MediaMetaType.Animation)
             {
+                Logs.Error("Media type is not supported for compression.");
                 return media.Data;
             }
             ISImage img = image.ToIS;
@@ -82,13 +88,18 @@ public static string CompressImageForVision(MediaContent media, string targetFor
                 int newHeight = (int)(img.Height * scaleFactor);
                 img.Mutate(i => i.Resize(newWidth, newHeight));
             }
-            // Set compression quality based on format TODO: This needs to be tested and adjusted
             int quality = targetFormat == "PNG" ? 60 : 40;
-            ImageFile tempImage = new Image(ImageFile.ISImgToPngBytes(img), image.Type);
-            ImageFile compressedImage = tempImage.ConvertTo(targetFormat, quality: quality);
-            // Return just the base64 data (without the data:image/webp;base64, prefix)
+            var imageFile = new Image(img) as SwarmUI.Media.ImageFile;
+            if (imageFile == null)
+            {
+                Logs.Error("Failed to convert image to ImageFile.");
+                return media.Data;
+            }
+            var compressedImage = imageFile.ConvertTo(targetFormat, quality: quality);
             return compressedImage.AsBase64;
+
         }
+
         catch (Exception ex)
         {
             Logs.Error($"Failed to compress image: {ex.Message}");
@@ -124,7 +135,11 @@ private static object OllamaRequestBody(MessageContent content, string model, Me
                 messages = messages.ToArray(),
                 stream = false,
                 keep_alive = content.KeepAlive,
-                options
+                options = new
+                {
+                    temperature = 1.0,
+                    top_p = 0.9,
+                }
             };
         }
         messages.Add(new { role = "user", content = content.Text });
@@ -134,7 +149,11 @@ private static object OllamaRequestBody(MessageContent content, string model, Me
             messages = messages.ToArray(),
             stream = false,
             keep_alive = content.KeepAlive,
-            options
+            options = new
+            {
+                temperature = 1.0,
+                top_p = 0.9,
+            }
         };
     }
 
@@ -190,7 +209,9 @@ private static object OpenAICompatibleRequestBody(MessageContent content, string
                 model,
                 messages = messages.ToArray(),
                 max_tokens = 1000,
-                temperature = 1.0,
+                temperature = 1.1,
+                min_p = 0.05,
+                top_p = 0.95,
                 stream = false
             };
         }
@@ -213,10 +234,11 @@ private static object OpenAICompatibleRequestBody(MessageContent content, string
         {
             model,
             messages = messages.ToArray(),
-            temperature = 1.0,
+            temperature = 1.1,
             max_tokens = 1000,
             top_p = 0.9,
-            stream = false
+            min_p = 0.05,
+            stream = false,
         };
     }
 
@@ -270,4 +292,4 @@ private static object AnthropicRequestBody(MessageContent content, string model,
             max_tokens = 1024
         };
     }
-}
+}

WebAPI/LLMAPICalls.cs

@@ -14,7 +14,15 @@ namespace Hartsy.Extensions.MagicPromptExtension.WebAPI;
 
 public class LLMAPICalls : MagicPromptAPI
 {
-    protected static readonly HttpClient HttpClient = NetworkBackendUtils.MakeHttpClient();
+    // Set HttpClient timeout to 120 seconds
+    protected static readonly HttpClient HttpClient = CreateHttpClientWithTimeout();
+
+    private static HttpClient CreateHttpClientWithTimeout()
+    {
+        var client = NetworkBackendUtils.MakeHttpClient();
+        client.Timeout = TimeSpan.FromSeconds(120);
+        return client;
+    }
 
     /// <summary>Fetches available models from the LLM API endpoint.</summary>
     /// <returns>A JSON object containing the models or an error message.</returns>
@@ -418,19 +426,19 @@ public static async Task<JObject> MagicPromptPhoneHome(JObject requestData, Sess
             {
                 // Typically thrown for validation issues (e.g., Grok vision requires direct JPG/PNG URLs)
                 Logs.Error($"Request build error for {backend}: {ex.Message}");
-                return CreateErrorResponse(ErrorHandler.FormatErrorMessage(ErrorType.UnsupportedParameterImage, ex.Message, backend));
+                return CreateErrorResponse(ErrorHandler.FormatErrorMessage(ErrorType.ValidationError, ex.Message, backend));
             }
-            string jsonContent = JsonSerializer.Serialize(requestBody);
-            request.Content = new StringContent(jsonContent, Encoding.UTF8, "application/json");
             try
             {
                 // Send request and handle response
                 // Detailed diagnostics to help trace hanging requests
                 Logs.Debug($"[MagicPrompt] Sending request | backend={backend} | type={(messageType == MessageType.Vision ? "vision" : "chat")} | endpoint={endpoint} | model={modelId}");
                 // Use KeepAlive (if provided) as a soft hint for timeout.
-                // Increase defaults and caps, especially for Ollama on slower machines.
-                int cap = backend == "ollama" ? 300 : 180;           // maximum allowed timeout (seconds)
-                int defaultTimeout = backend == "ollama" ? 120 : 60;  // default timeout when KeepAlive is not set
+                // Increase defaults and caps, especially for Ollama on slower machines and vision requests.
+                // Vision/image captioning requests get significantly longer timeouts due to image processing overhead.
+                bool isVisionRequest = messageType == MessageType.Vision;
+                int cap = backend == "ollama" ? 300 : (isVisionRequest ? 240 : 180);           // maximum allowed timeout (seconds)
+                int defaultTimeout = backend == "ollama" ? 120 : (isVisionRequest ? 120 : 60);  // default timeout when KeepAlive is not set
                 int timeoutSec = messageContent.KeepAlive.HasValue && messageContent.KeepAlive.Value > 0
                     ? Math.Min(messageContent.KeepAlive.Value, cap)
                     : defaultTimeout;