Improved PowerPoint implementation for reading slide data (#517)

app/MindWork AI Studio/Tools/ContentStreamSseHandler.cs

@@ -6,7 +6,7 @@ namespace AIStudio.Tools;
 public static class ContentStreamSseHandler
 {
     private static readonly ConcurrentDictionary<string, List<ContentStreamPptxImageData>> CHUNKED_IMAGES = new();
-    private static readonly ConcurrentDictionary<string, int> CURRENT_SLIDE_NUMBERS = new();
+    private static readonly ConcurrentDictionary<string, SlideManager> SLIDE_MANAGERS = new();
 
     public static string? ProcessEvent(ContentStreamSseEvent? sseEvent, bool extractImages = true)
     {
@@ -44,31 +44,13 @@ public static class ContentStreamSseHandler
                         return sseEvent.Content;
 
                     case ContentStreamPresentationMetadata presentationMetadata:
-                        var slideNumber = presentationMetadata.Presentation?.SlideNumber ?? 0;
+                        var slideManager = SLIDE_MANAGERS.GetOrAdd(
-                        var image = presentationMetadata.Presentation?.Image ?? null;
+                            sseEvent.StreamId!,
-                        var presentationResult = new StringBuilder();
+                            _ => new()
-                        var streamId = sseEvent.StreamId;
+                        );
                         
-                        CURRENT_SLIDE_NUMBERS.TryGetValue(streamId!, out var currentSlideNumber);
+                        slideManager.AddSlide(presentationMetadata, sseEvent.Content, extractImages);
-                        if (slideNumber != currentSlideNumber)
+                        return null;
-                        {
-                            presentationResult.AppendLine();
-                            presentationResult.AppendLine($"# Slide {slideNumber}");
-                        }
-
-                        if(!string.IsNullOrWhiteSpace(sseEvent.Content))
-                            presentationResult.AppendLine(sseEvent.Content);
-                        
-                        if (extractImages && image is not null)
-                        {
-                            var imageId = $"{streamId}-{image.Id!}";
-                            var isEnd = ProcessImageSegment(imageId, image);
-                            if (isEnd && extractImages)
-                                presentationResult.AppendLine(BuildImage(imageId));
-                        }
-
-                        CURRENT_SLIDE_NUMBERS[streamId!] = slideNumber;
-                        return presentationResult.Length is 0 ? null : presentationResult.ToString();
                     
                     default:
                         return sseEvent.Content;
@@ -81,8 +63,8 @@ public static class ContentStreamSseHandler
                 return null;
         }
     }
-    
+
-    private static bool ProcessImageSegment(string imageId, ContentStreamPptxImageData contentStreamPptxImageData)
+    public static bool ProcessImageSegment(string imageId, ContentStreamPptxImageData contentStreamPptxImageData)
     {
         if (string.IsNullOrWhiteSpace(contentStreamPptxImageData.Id) || string.IsNullOrWhiteSpace(imageId))
             return false;
@@ -112,7 +94,7 @@ public static class ContentStreamSseHandler
         return isEnd;
     }
 
-    private static string BuildImage(string id)
+    public static string BuildImage(string id)
     {
         if (!CHUNKED_IMAGES.TryGetValue(id, out var imageSegments))
             return string.Empty;
@@ -128,4 +110,25 @@ public static class ContentStreamSseHandler
         CHUNKED_IMAGES.Remove(id, out _);
         return base64Image;
     }
+    
+    public static string? Clear(string streamId)
+    {
+        if (string.IsNullOrWhiteSpace(streamId))
+            return null;
+ 
+        var finalContentChunk = new StringBuilder();
+        if(SLIDE_MANAGERS.TryGetValue(streamId, out var slideManager))
+        {
+            var result = slideManager.GetAllSlidesInOrder();
+            if (!string.IsNullOrWhiteSpace(result))
+                finalContentChunk.Append(result);
+        }
+        
+        SLIDE_MANAGERS.TryRemove(streamId, out _);
+        var imageIdPrefix = $"{streamId}-";
+        foreach (var key in CHUNKED_IMAGES.Keys.Where(k => k.StartsWith(imageIdPrefix, StringComparison.InvariantCultureIgnoreCase)))
+            CHUNKED_IMAGES.TryRemove(key, out _);
+        
+        return finalContentChunk.Length > 0 ? finalContentChunk.ToString() : null;
+    }
 }

app/MindWork AI Studio/Tools/ISlideContent.cs

@@ -0,0 +1,3 @@
+namespace AIStudio.Tools;
+
+public interface ISlideContent;

app/MindWork AI Studio/Tools/Services/RustService.Retrieval.cs

@@ -15,39 +15,52 @@ public sealed partial class RustService
         if (!response.IsSuccessStatusCode)
             return string.Empty;
 
-        await using var stream = await response.Content.ReadAsStreamAsync();
-        using var reader = new StreamReader(stream);
-
         var resultBuilder = new StringBuilder();
-        var chunkCount = 0;
 
-        while (!reader.EndOfStream && chunkCount < maxChunks)
+        try
         {
-            var line = await reader.ReadLineAsync();
+            await using var stream = await response.Content.ReadAsStreamAsync();
-            if (string.IsNullOrWhiteSpace(line))
+            using var reader = new StreamReader(stream);
-                continue;
+            var chunkCount = 0;
-            
-            if (!line.StartsWith("data:", StringComparison.InvariantCulture))
-                continue;
-            
-            var jsonContent = line[5..];
 
-            try
+            while (!reader.EndOfStream && chunkCount < maxChunks)
             {
-                var sseEvent = JsonSerializer.Deserialize<ContentStreamSseEvent>(jsonContent);
+                var line = await reader.ReadLineAsync();
-                if (sseEvent is not null)
+                if (string.IsNullOrWhiteSpace(line))
+                    continue;
+
+                if (!line.StartsWith("data:", StringComparison.InvariantCulture))
+                    continue;
+
+                var jsonContent = line[5..];
+
+                try
                 {
-                    var content = ContentStreamSseHandler.ProcessEvent(sseEvent, extractImages);
+                    var sseEvent = JsonSerializer.Deserialize<ContentStreamSseEvent>(jsonContent);
-                    if(content is not null)
+                    if (sseEvent is not null)
-                        resultBuilder.AppendLine(content);
+                    {
-                    
+                        var content = ContentStreamSseHandler.ProcessEvent(sseEvent, extractImages);
-                    chunkCount++;
+                        if (content is not null)
+                            resultBuilder.AppendLine(content);
+
+                        chunkCount++;
+                    }
+                }
+                catch (JsonException)
+                {
+                    this.logger?.LogError("Failed to deserialize SSE event: {JsonContent}", jsonContent);
                 }
             }
-            catch (JsonException)
+        }
-            {
+        catch(Exception e)
-                this.logger?.LogError("Failed to deserialize SSE event: {JsonContent}", jsonContent); 
+        {
-            }
+            this.logger?.LogError(e, "Error reading file data from stream: {Path}", path);
+        }
+        finally
+        {
+            var finalContentChunk = ContentStreamSseHandler.Clear(streamId);
+            if (!string.IsNullOrWhiteSpace(finalContentChunk))
+                resultBuilder.AppendLine(finalContentChunk);
         }
         
         return resultBuilder.ToString();

app/MindWork AI Studio/Tools/Slide.cs

@@ -0,0 +1,10 @@
+namespace AIStudio.Tools;
+
+public sealed class Slide
+{
+    public bool Delivered { get; set; }
+
+    public int Position { get; init; }
+    
+    public List<ISlideContent> Content { get; } = new();
+}

app/MindWork AI Studio/Tools/SlideImageContent.cs

@@ -0,0 +1,8 @@
+using System.Text;
+
+namespace AIStudio.Tools;
+
+public sealed class SlideImageContent(string base64Image) : ISlideContent
+{
+    public StringBuilder Base64Image => new(base64Image);
+}

app/MindWork AI Studio/Tools/SlideManager.cs

@@ -0,0 +1,106 @@
+using System.Text;
+
+namespace AIStudio.Tools;
+
+public sealed class SlideManager
+{
+    private readonly Dictionary<int, Slide> slides = new();
+
+    public void AddSlide(ContentStreamPresentationMetadata metadata, string? content, bool extractImages = false)
+    {
+        var slideNumber = metadata.Presentation?.SlideNumber ?? 0;
+        if(slideNumber is 0)
+            return;
+        
+        var image = metadata.Presentation?.Image ?? null;
+        var addImage = false;
+        if (extractImages && image is not null)
+        {
+            var isEnd = ContentStreamSseHandler.ProcessImageSegment(image.Id!, image);
+            if (isEnd)
+                addImage = true;
+        }
+        
+        if (!this.slides.TryGetValue(slideNumber, out var slide))
+        {
+            //
+            // Case: No existing slide content for this slide number.
+            //
+            
+            var contentBuilder = new StringBuilder();
+            contentBuilder.AppendLine();
+            contentBuilder.AppendLine($"# Slide {slideNumber}");
+            
+            // Add any text content to the slide?
+            if(!string.IsNullOrWhiteSpace(content))
+                contentBuilder.AppendLine(content);
+
+            //
+            // Add the text content to the slide:
+            //
+            var slideText = new SlideTextContent(contentBuilder.ToString());
+            var createdSlide = new Slide
+            {
+                Delivered = false,
+                Position = slideNumber
+            };
+            
+            createdSlide.Content.Add(slideText);
+            
+            //
+            // Add image content to the slide?
+            //
+            if (addImage)
+            {
+                var img = ContentStreamSseHandler.BuildImage(image!.Id!);
+                var slideImage = new SlideImageContent(img);
+                createdSlide.Content.Add(slideImage);
+            }
+            
+            this.slides[slideNumber] = createdSlide;
+        }
+        else
+        {
+            //
+            // Case: Existing slide content for this slide number.
+            //
+            
+            // Add any text content?
+            if (!string.IsNullOrWhiteSpace(content))
+            {
+                var textContent = slide.Content.OfType<SlideTextContent>().First();
+                textContent.Text.AppendLine(content);
+            }
+            
+            // Add any image content?
+            if (addImage)
+            {
+                var img = ContentStreamSseHandler.BuildImage(image!.Id!);
+                var slideImage = new SlideImageContent(img);
+                slide.Content.Add(slideImage);
+            }
+        }
+    }
+
+    public string? GetAllSlidesInOrder()
+    {
+        var content = new StringBuilder();
+        foreach (var slide in this.slides.Values.Where(s => !s.Delivered).OrderBy(s => s.Position))
+        {
+            slide.Delivered = true;
+            foreach (var text in slide.Content.OfType<SlideTextContent>())
+            {
+                content.AppendLine(text.Text.ToString());
+                content.AppendLine();
+            }
+
+            foreach (var image in slide.Content.OfType<SlideImageContent>())
+            {
+                content.AppendLine(image.Base64Image.ToString());
+                content.AppendLine();
+            }
+        }
+        
+        return content.Length > 0 ? content.ToString() : null;
+    }
+}

app/MindWork AI Studio/Tools/SlideTextContent.cs

@@ -0,0 +1,8 @@
+using System.Text;
+
+namespace AIStudio.Tools;
+
+public sealed class SlideTextContent(string textContent) : ISlideContent
+{
+    public StringBuilder Text => new(textContent);
+}

runtime/Cargo.lock

@@ -3408,9 +3408,9 @@ checksum = "439ee305def115ba05938db6eb1644ff94165c5ab5e9420d1c1bcedbba909391"
 
 [[package]]
 name = "pptx-to-md"
-version = "0.3.0"
+version = "0.4.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e26f6df203425a22367de642b415c18f1456de2bc870fbd7d2be83d5f57ae058"
+checksum = "25f7bef20173da9d560ffb6b67cba2d2b834375d0d262e5aeb86f44e069ae446"
 dependencies = [
  "base64 0.22.1",
  "image 0.24.9",

runtime/Cargo.toml

@@ -38,7 +38,7 @@ calamine = "0.28.0"
 pdfium-render = "0.8.33"
 sys-locale = "0.3.2"
 cfg-if = "1.0.1"
-pptx-to-md = "0.3.0"
+pptx-to-md = "0.4.0"
 
 # Fixes security vulnerability downstream, where the upstream is not fixed yet:
 url = "2.5"