feat: Add PDF header/footer filter and DOCX table complexity analyzer

- Add PdfHeaderFooterFilter for repetitive line detection and removal
- Add DocxTableComplexityAnalyzer (Phase 1 of complex table roadmap)
- Enhance PPTX slide title extraction with shape type analysis
- Improve XLSX Markdown table output formatting
- Add RefineOptions.FilterPdfHeaderFooter/PdfHeaderFooterThreshold
- Bump version to 0.9.5

Author: birdnamoo

CHANGELOG.md

@@ -8,6 +8,21 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ## [Unreleased]
 
 ### Added
+- **PDF Header/Footer Filter**: Repetitive line pattern detection and removal
+  - `PdfHeaderFooterFilter`: Detects lines appearing on multiple pages (configurable threshold)
+  - `RefineOptions.FilterPdfHeaderFooter`: Opt-in feature for PDF noise reduction
+  - `RefineOptions.PdfHeaderFooterThreshold`: Configurable ratio (default 0.5 = 50% of pages)
+- **DOCX Table Complexity Analyzer**: Phase 1 of complex table handling roadmap
+  - `DocxTableComplexityAnalyzer`: Detects merged cells, nested tables, irregular rows
+  - `TableAnalysisResult`: Comprehensive metrics with complexity scoring (0.0-1.0)
+  - `TableComplexityLevel`: Simple, Low, Medium, High, VeryHigh classification
+  - Generates warnings for Markdown-incompatible table structures
+- **PPTX Slide Title Extraction**: Enhanced title detection in presentations
+  - Improved title placeholder detection using shape type analysis
+  - Better handling of slides without explicit title shapes
+- **XLSX Markdown Table Output**: Enhanced spreadsheet to Markdown conversion
+  - Proper column alignment detection and formatting
+  - Improved cell content handling for complex data
 - **Auto Chunking Strategy Selection**: Document structure-based strategy selection
   - `SelectChunkingStrategy()`: Analyzes document for headings and numbered sections
   - `AnalyzeDocumentStructure()`: Detects Markdown headings, numbered patterns

Directory.Build.props

@@ -2,7 +2,7 @@
 <Project>
   <!-- Version Information (can be overridden by build scripts) -->
   <PropertyGroup>
-    <VersionPrefix>0.9.4</VersionPrefix>
+    <VersionPrefix>0.9.5</VersionPrefix>
     <AssemblyVersion>$(VersionPrefix)</AssemblyVersion>
     <FileVersion>$(VersionPrefix)</FileVersion>
     <PackageVersion Condition="'$(PackageVersion)' == ''">$(VersionPrefix)</PackageVersion>

src/FileFlux.Core/Contracts/IDocumentProcessor.cs

@@ -282,6 +282,19 @@ public class RefineOptions
     /// </summary>
     public int? MaxLlmTokens { get; set; }
 
+    /// <summary>
+    /// Enable PDF header/footer pattern detection and removal.
+    /// When enabled, repetitive lines appearing on multiple pages are identified and filtered.
+    /// Default: false (opt-in feature).
+    /// </summary>
+    public bool FilterPdfHeaderFooter { get; set; } = false;
+
+    /// <summary>
+    /// Minimum ratio of pages a line must appear on to be considered header/footer.
+    /// Range: 0.0-1.0. Default: 0.5 (50% of pages).
+    /// </summary>
+    public double PdfHeaderFooterThreshold { get; set; } = 0.5;
+
     /// <summary>
     /// Default refinement options.
     /// </summary>

src/FileFlux.Core/Infrastructure/Readers/ExcelDocumentReader.cs

@@ -427,7 +427,10 @@ private static (string Content, int RowCount, int CellCount) ExtractWorksheetCon
         try
         {
             var rows = worksheetPart.Worksheet.Descendants<Row>().ToList();
+            var processedRows = new List<List<string>>();
+            var maxColumns = 0;
 
+            // First pass: collect all rows and determine max columns
             foreach (var row in rows)
             {
                 cancellationToken.ThrowIfCancellationRequested();
@@ -439,7 +442,9 @@ private static (string Content, int RowCount, int CellCount) ExtractWorksheetCon
                 foreach (var cell in cells)
                 {
                     var cellValue = GetCellValue(cell, sharedStringTable);
-                    cellValues.Add(cellValue);
+                    // Escape pipe characters for Markdown table compatibility
+                    var escapedValue = cellValue.Replace("|", "\\|").Replace("\n", " ").Replace("\r", "");
+                    cellValues.Add(escapedValue);
 
                     if (!string.IsNullOrWhiteSpace(cellValue))
                     {
@@ -450,13 +455,33 @@ private static (string Content, int RowCount, int CellCount) ExtractWorksheetCon
 
                 if (hasContent)
                 {
-                    // 테이블 형식으로 셀 값들을 연결 (파이프 구분자 사용)
-                    var rowText = string.Join(" | ", cellValues.Where(v => !string.IsNullOrWhiteSpace(v)));
-                    if (!string.IsNullOrWhiteSpace(rowText))
-                    {
-                        contentBuilder.AppendLine(rowText);
-                        rowCount++;
-                    }
+                    processedRows.Add(cellValues);
+                    maxColumns = Math.Max(maxColumns, cellValues.Count);
+                    rowCount++;
+                }
+            }
+
+            // Generate Markdown table
+            if (processedRows.Count > 0 && maxColumns > 0)
+            {
+                // Normalize all rows to have same column count
+                foreach (var row in processedRows)
+                {
+                    while (row.Count < maxColumns)
+                        row.Add(string.Empty);
+                }
+
+                // Header row (first row)
+                var headerRow = processedRows[0];
+                contentBuilder.AppendLine("| " + string.Join(" | ", headerRow) + " |");
+
+                // Separator row
+                contentBuilder.AppendLine("| " + string.Join(" | ", headerRow.Select(_ => "---")) + " |");
+
+                // Data rows (skip header)
+                for (int i = 1; i < processedRows.Count; i++)
+                {
+                    contentBuilder.AppendLine("| " + string.Join(" | ", processedRows[i]) + " |");
                 }
             }
         }

src/FileFlux.Core/Infrastructure/Readers/PowerPointDocumentReader.cs

@@ -404,14 +404,68 @@ private static RawContent ExtractPowerPointContentFromStream(Stream stream, stri
         }
     }
 
+    /// <summary>
+    /// Extracts the slide title from title placeholder shapes.
+    /// Falls back to first text content or slide number if no title found.
+    /// </summary>
+    private static string ExtractSlideTitle(Slide slide, int slideNumber)
+    {
+        var shapeTree = slide.CommonSlideData?.ShapeTree;
+        if (shapeTree == null)
+            return $"Slide {slideNumber}";
+
+        // 1. Try to find title placeholder (Title or CenteredTitle)
+        foreach (var shape in shapeTree.Elements<Shape>())
+        {
+            var placeholder = shape.NonVisualShapeProperties?
+                .ApplicationNonVisualDrawingProperties?
+                .GetFirstChild<PlaceholderShape>();
+
+            if (placeholder?.Type != null && placeholder.Type.HasValue)
+            {
+                var placeholderType = placeholder.Type.Value;
+                if (placeholderType == PlaceholderValues.Title ||
+                    placeholderType == PlaceholderValues.CenteredTitle)
+                {
+                    var titleText = ExtractTextFromShape(shape);
+                    if (!string.IsNullOrWhiteSpace(titleText))
+                    {
+                        // Clean up: remove newlines and trim
+                        return titleText.Replace("\r\n", " ").Replace("\n", " ").Trim();
+                    }
+                }
+            }
+        }
+
+        // 2. Fallback: try first text block if it looks like a title (short text)
+        foreach (var shape in shapeTree.Elements<Shape>())
+        {
+            var text = ExtractTextFromShape(shape);
+            if (!string.IsNullOrWhiteSpace(text))
+            {
+                var cleanText = text.Replace("\r\n", " ").Replace("\n", " ").Trim();
+                // Use as title only if reasonably short (under 100 chars)
+                if (cleanText.Length > 0 && cleanText.Length < 100)
+                {
+                    return cleanText;
+                }
+            }
+        }
+
+        // 3. Final fallback
+        return $"Slide {slideNumber}";
+    }
+
     private static (string Content, int ShapeCount) ExtractSlideContent(Slide slide, SlidePart slidePart, int slideNumber, List<string> warnings, CancellationToken cancellationToken)
     {
         var contentBuilder = new StringBuilder();
         var shapeCount = 0;
 
         try
         {
-            contentBuilder.AppendLine($"## Slide {slideNumber}");
+            // Extract actual slide title from title placeholder
+            var slideTitle = ExtractSlideTitle(slide, slideNumber);
+            contentBuilder.AppendLine($"## {slideTitle}");
             contentBuilder.AppendLine();
 
             var shapeTree = slide.CommonSlideData?.ShapeTree;