supervertaler 1.9.163__py3-none-any.whl

modules/tracked_changes.py

@@ -0,0 +1,900 @@
+"""
+Tracked Changes Management Module
+
+This module handles tracked changes from DOCX files or TSV files.
+Provides AI with examples of preferred editing patterns to learn translator style.
+
+Classes:
+    - TrackedChangesAgent: Manages tracked changes data and provides search/filtering
+    - TrackedChangesBrowser: UI window for browsing and analyzing tracked changes
+"""
+
+import tkinter as tk
+from tkinter import ttk, filedialog, messagebox
+import os
+import re
+import queue
+from datetime import datetime
+from typing import List, Tuple, Dict, Optional
+
+# Import parse_docx_pairs from the parent modules (it's a standalone function)
+# Note: This will be imported from the main file
+# For now, we'll assume it's available in the calling context
+
+
+class TrackedChangesAgent:
+    """
+    Manages tracked changes from DOCX files or TSV files.
+    Provides AI with examples of preferred editing patterns to learn translator style.
+    """
+    def __init__(self, log_callback=None):
+        self.change_data = []  # List of (original_text, final_text) tuples
+        self.files_loaded = []  # Track which files have been loaded
+        self.log_callback = log_callback or print
+    
+    def log(self, message):
+        """Log a message"""
+        if callable(self.log_callback):
+            self.log_callback(message)
+    
+    def load_docx_changes(self, docx_path, parse_docx_pairs_func):
+        """Load tracked changes from a DOCX file
+        
+        Args:
+            docx_path: Path to DOCX file
+            parse_docx_pairs_func: Function to parse DOCX and extract change pairs
+        """
+        if not docx_path:
+            return False
+            
+        self.log(f"[Tracked Changes] Loading changes from: {os.path.basename(docx_path)}")
+        
+        try:
+            new_changes = parse_docx_pairs_func(docx_path)
+            
+            # Clear existing changes to prevent duplicates
+            self.change_data.clear()
+            self.files_loaded.clear()
+            
+            # Add new changes
+            self.change_data.extend(new_changes)
+            self.files_loaded.append(os.path.basename(docx_path))
+            
+            self.log(f"[Tracked Changes] Loaded {len(new_changes)} change pairs from {os.path.basename(docx_path)}")
+            self.log(f"[Tracked Changes] Total change pairs available: {len(self.change_data)}")
+            
+            return True
+        except Exception as e:
+            self.log(f"[Tracked Changes] Error loading {docx_path}: {e}")
+            messagebox.showerror("Tracked Changes Error", 
+                               f"Failed to load tracked changes from {os.path.basename(docx_path)}:\\n{e}")
+            return False
+    
+    def load_tsv_changes(self, tsv_path):
+        """Load tracked changes from a TSV file (original_text<tab>final_text format)"""
+        if not tsv_path:
+            return False
+            
+        self.log(f"[Tracked Changes] Loading changes from: {os.path.basename(tsv_path)}")
+        
+        try:
+            new_changes = []
+            with open(tsv_path, 'r', encoding='utf-8') as f:
+                for line_num, line in enumerate(f, 1):
+                    line = line.rstrip('\n\r')
+                    if not line.strip():
+                        continue
+                    
+                    # Skip header line if it looks like one
+                    if line_num == 1 and ('original' in line.lower() and 'final' in line.lower()):
+                        continue
+                    
+                    parts = line.split('\t')
+                    if len(parts) >= 2:
+                        original = parts[0].strip()
+                        final = parts[1].strip()
+                        if original and final and original != final:  # Only add if actually different
+                            new_changes.append((original, final))
+                    else:
+                        self.log(f"[Tracked Changes] Skipping line {line_num}: insufficient columns")
+            
+            # Add to existing changes
+            self.change_data.extend(new_changes)
+            self.files_loaded.append(os.path.basename(tsv_path))
+            
+            self.log(f"[Tracked Changes] Loaded {len(new_changes)} change pairs from {os.path.basename(tsv_path)}")
+            self.log(f"[Tracked Changes] Total change pairs available: {len(self.change_data)}")
+            
+            return True
+        except Exception as e:
+            self.log(f"[Tracked Changes] Error loading {tsv_path}: {e}")
+            messagebox.showerror("Tracked Changes Error", 
+                               f"Failed to load tracked changes from {os.path.basename(tsv_path)}:\\n{e}")
+            return False
+    
+    def clear_changes(self):
+        """Clear all loaded tracked changes"""
+        self.change_data.clear()
+        self.files_loaded.clear()
+        self.log("[Tracked Changes] All tracked changes cleared")
+    
+    def search_changes(self, search_text, exact_match=False):
+        """Search for changes containing the search text"""
+        if not search_text.strip():
+            return self.change_data
+        
+        search_lower = search_text.lower()
+        results = []
+        
+        for original, final in self.change_data:
+            if exact_match:
+                if search_text == original or search_text == final:
+                    results.append((original, final))
+            else:
+                if (search_lower in original.lower() or 
+                    search_lower in final.lower()):
+                    results.append((original, final))
+        
+        return results
+
+    def find_relevant_changes(self, source_segments, max_changes=10):
+        """
+        Find tracked changes relevant to the current source segments being processed.
+        Uses two-pass algorithm: exact matches first, then partial word overlap.
+        """
+        if not self.change_data or not source_segments:
+            return []
+        
+        relevant_changes = []
+        
+        # First pass: exact matches
+        for segment in source_segments:
+            segment_lower = segment.lower().strip()
+            for original, final in self.change_data:
+                original_lower = original.lower().strip()
+                if segment_lower == original_lower and (original, final) not in relevant_changes:
+                    relevant_changes.append((original, final))
+                    if len(relevant_changes) >= max_changes:
+                        return relevant_changes
+        
+        # Second pass: partial matches (word overlap)
+        if len(relevant_changes) < max_changes:
+            for segment in source_segments:
+                segment_words = set(word.lower() for word in segment.split() if len(word) > 3)
+                for original, final in self.change_data:
+                    if (original, final) in relevant_changes:
+                        continue
+                    
+                    original_words = set(word.lower() for word in original.split() if len(word) > 3)
+                    # Check if there's significant word overlap
+                    if segment_words and original_words:
+                        overlap = len(segment_words.intersection(original_words))
+                        min_overlap = min(2, len(segment_words) // 2)
+                        if overlap >= min_overlap:
+                            relevant_changes.append((original, final))
+                            if len(relevant_changes) >= max_changes:
+                                return relevant_changes
+        
+        return relevant_changes
+    
+    def get_entry_count(self):
+        """Get number of loaded change pairs"""
+        return len(self.change_data)
+
+
+class TrackedChangesBrowser:
+    """Browser UI for viewing and searching tracked changes"""
+    
+    def __init__(self, parent, tracked_changes_agent, parent_app=None, log_queue=None, 
+                 gemini_available=False, anthropic_available=False, openai_available=False, app_version="3.6.0"):
+        self.parent = parent
+        self.tracked_changes_agent = tracked_changes_agent
+        self.parent_app = parent_app  # Reference to main app for AI settings
+        self.log_queue = log_queue if log_queue else queue.Queue()
+        self.window = None
+        
+        # AI availability flags
+        self.GEMINI_AVAILABLE = gemini_available
+        self.ANTHROPIC_AVAILABLE = anthropic_available
+        self.OPENAI_AVAILABLE = openai_available
+        self.APP_VERSION = app_version
+    
+    def show_browser(self):
+        """Show the tracked changes browser window"""
+        if not self.tracked_changes_agent.change_data:
+            messagebox.showinfo("No Changes", "No tracked changes loaded. Load a DOCX or TSV file with tracked changes first.")
+            return
+        
+        # Create window if it doesn't exist
+        if self.window is None or not self.window.winfo_exists():
+            self.create_window()
+        else:
+            self.window.lift()
+    
+    def create_window(self):
+        """Create the browser window"""
+        self.window = tk.Toplevel(self.parent)
+        self.window.title(f"Tracked Changes Browser ({len(self.tracked_changes_agent.change_data)} pairs)")
+        self.window.geometry("900x700")  # Taller to accommodate detail view
+        
+        # Search frame
+        search_frame = tk.Frame(self.window)
+        search_frame.pack(fill=tk.X, padx=10, pady=5)
+        
+        tk.Label(search_frame, text="Search:").pack(side=tk.LEFT)
+        self.search_var = tk.StringVar()
+        search_entry = tk.Entry(search_frame, textvariable=self.search_var, width=40)
+        search_entry.pack(side=tk.LEFT, padx=(5,0))
+        search_entry.bind('<KeyRelease>', self.on_search)
+        
+        self.exact_match_var = tk.BooleanVar()
+        tk.Checkbutton(search_frame, text="Exact match", variable=self.exact_match_var, 
+                      command=self.on_search).pack(side=tk.LEFT, padx=(10,0))
+        
+        tk.Button(search_frame, text="Clear", command=self.clear_search).pack(side=tk.LEFT, padx=(10,0))
+        
+        # Results info
+        self.results_label = tk.Label(self.window, text="")
+        self.results_label.pack(pady=2)
+        
+        # Main content frame (results + detail)
+        main_frame = tk.Frame(self.window)
+        main_frame.pack(fill=tk.BOTH, expand=True, padx=10, pady=5)
+        
+        # Results frame with scrollbar (top half)
+        results_frame = tk.Frame(main_frame)
+        results_frame.pack(fill=tk.BOTH, expand=True)
+        
+        # Create Treeview for displaying changes
+        columns = ('Original', 'Final')
+        self.tree = ttk.Treeview(results_frame, columns=columns, show='headings', height=12)
+        
+        # Define headings
+        self.tree.heading('Original', text='Original Text')
+        self.tree.heading('Final', text='Final Text')
+        
+        # Configure column widths
+        self.tree.column('Original', width=400)
+        self.tree.column('Final', width=400)
+        
+        # Add scrollbars
+        v_scrollbar = ttk.Scrollbar(results_frame, orient=tk.VERTICAL, command=self.tree.yview)
+        h_scrollbar = ttk.Scrollbar(results_frame, orient=tk.HORIZONTAL, command=self.tree.xview)
+        self.tree.configure(yscrollcommand=v_scrollbar.set, xscrollcommand=h_scrollbar.set)
+        
+        # Pack tree and scrollbars
+        self.tree.grid(row=0, column=0, sticky="nsew")
+        v_scrollbar.grid(row=0, column=1, sticky="ns")
+        h_scrollbar.grid(row=1, column=0, sticky="ew")
+        
+        results_frame.grid_rowconfigure(0, weight=1)
+        results_frame.grid_columnconfigure(0, weight=1)
+        
+        # Detail view frame (bottom half)
+        detail_frame = tk.LabelFrame(main_frame, text="Selected Change Details", padx=5, pady=5)
+        detail_frame.pack(fill=tk.BOTH, expand=False, pady=(10,0))
+        
+        # Original text display
+        tk.Label(detail_frame, text="Original Text:", font=("Segoe UI", 10, "bold")).pack(anchor="w")
+        self.original_text = tk.Text(detail_frame, height=4, wrap=tk.WORD, state="disabled", 
+                                    bg="#f8f8f8", relief="solid", borderwidth=1)
+        self.original_text.pack(fill=tk.X, pady=(2,5))
+        
+        # Final text display
+        tk.Label(detail_frame, text="Final Text:", font=("Segoe UI", 10, "bold")).pack(anchor="w")
+        self.final_text = tk.Text(detail_frame, height=4, wrap=tk.WORD, state="disabled",
+                                 bg="#f0f8ff", relief="solid", borderwidth=1)
+        self.final_text.pack(fill=tk.X, pady=(2,0))
+        
+        # Bind selection event
+        self.tree.bind('<<TreeviewSelect>>', self.on_selection_change)
+        
+        # Context menu for copying
+        self.context_menu = tk.Menu(self.window, tearoff=0)
+        self.context_menu.add_command(label="Copy original", command=self.copy_original)
+        self.context_menu.add_command(label="Copy final", command=self.copy_final)
+        self.context_menu.add_command(label="Copy both", command=self.copy_both)
+        
+        self.tree.bind("<Button-3>", self.show_context_menu)  # Right click
+        
+        # Export button frame
+        export_frame = tk.Frame(self.window)
+        export_frame.pack(fill=tk.X, padx=10, pady=5)
+        
+        tk.Button(export_frame, text="📊 Export Report (MD)", command=self.export_to_md_report,
+                 bg="#4CAF50", fg="white", font=("Segoe UI", 10, "bold"),
+                 relief="raised", padx=10, pady=5).pack(side=tk.LEFT)
+        
+        tk.Label(export_frame, text="Export tracked changes report with AI-powered change analysis",
+                fg="gray").pack(side=tk.LEFT, padx=(10,0))
+        
+        # Status bar
+        status_frame = tk.Frame(self.window)
+        status_frame.pack(fill=tk.X, padx=10, pady=2)
+        
+        files_text = f"Files loaded: {', '.join(self.tracked_changes_agent.files_loaded)}" if self.tracked_changes_agent.files_loaded else "No files loaded"
+        tk.Label(status_frame, text=files_text, anchor=tk.W).pack(fill=tk.X)
+        
+        # Load all changes initially
+        self.load_results(self.tracked_changes_agent.change_data)
+    
+    def on_selection_change(self, event=None):
+        """Handle selection change in the tree"""
+        selection = self.tree.selection()
+        if not selection:
+            # Clear detail view if no selection
+            self.original_text.config(state="normal")
+            self.original_text.delete(1.0, tk.END)
+            self.original_text.config(state="disabled")
+            self.final_text.config(state="normal")
+            self.final_text.delete(1.0, tk.END)
+            self.final_text.config(state="disabled")
+            return
+        
+        # Get the selected change pair
+        original, final = self.get_selected_change()
+        if original and final:
+            # Update original text display
+            self.original_text.config(state="normal")
+            self.original_text.delete(1.0, tk.END)
+            self.original_text.insert(1.0, original)
+            self.original_text.config(state="disabled")
+            
+            # Update final text display
+            self.final_text.config(state="normal")
+            self.final_text.delete(1.0, tk.END)
+            self.final_text.insert(1.0, final)
+            self.final_text.config(state="disabled")
+    
+    def on_search(self, event=None):
+        """Handle search input"""
+        search_text = self.search_var.get()
+        exact_match = self.exact_match_var.get()
+        
+        results = self.tracked_changes_agent.search_changes(search_text, exact_match)
+        self.load_results(results)
+    
+    def clear_search(self):
+        """Clear search and show all results"""
+        self.search_var.set("")
+        self.load_results(self.tracked_changes_agent.change_data)
+    
+    def load_results(self, results):
+        """Load results into the treeview"""
+        # Clear existing items
+        for item in self.tree.get_children():
+            self.tree.delete(item)
+        
+        # Add new items
+        for i, (original, final) in enumerate(results):
+            # Truncate long text for display
+            display_original = (original[:100] + "...") if len(original) > 100 else original
+            display_final = (final[:100] + "...") if len(final) > 100 else final
+            
+            self.tree.insert('', 'end', values=(display_original, display_final))
+        
+        # Update results label
+        total_changes = len(self.tracked_changes_agent.change_data)
+        showing = len(results)
+        if showing == total_changes:
+            self.results_label.config(text=f"Showing all {total_changes} change pairs")
+        else:
+            self.results_label.config(text=f"Showing {showing} of {total_changes} change pairs")
+    
+    def show_context_menu(self, event):
+        """Show context menu for copying"""
+        item = self.tree.identify_row(event.y)
+        if item:
+            self.tree.selection_set(item)
+            self.context_menu.post(event.x_root, event.y_root)
+    
+    def get_selected_change(self):
+        """Get the currently selected change pair"""
+        selection = self.tree.selection()
+        if not selection:
+            return None, None
+        
+        item = selection[0]
+        index = self.tree.index(item)
+        
+        # Get current results (might be filtered)
+        search_text = self.search_var.get()
+        exact_match = self.exact_match_var.get()
+        current_results = self.tracked_changes_agent.search_changes(search_text, exact_match)
+        
+        if 0 <= index < len(current_results):
+            return current_results[index]
+        return None, None
+    
+    def copy_original(self):
+        """Copy original text to clipboard"""
+        original, _ = self.get_selected_change()
+        if original:
+            self.window.clipboard_clear()
+            self.window.clipboard_append(original)
+    
+    def copy_final(self):
+        """Copy final text to clipboard"""
+        _, final = self.get_selected_change()
+        if final:
+            self.window.clipboard_clear()
+            self.window.clipboard_append(final)
+    
+    def copy_both(self):
+        """Copy both texts to clipboard"""
+        original, final = self.get_selected_change()
+        if original and final:
+            both_text = f"Original: {original}\n\nFinal: {final}"
+            self.window.clipboard_clear()
+            self.window.clipboard_append(both_text)
+
+    
+    def export_to_md_report(self):
+        """Export tracked changes to a Markdown report with AI-powered change analysis"""
+        if not self.tracked_changes_agent.change_data:
+            messagebox.showwarning("No Data", "No tracked changes available to export.")
+            return
+        
+        # Ask user whether to export all or filtered results
+        search_text = self.search_var.get()
+        if search_text:
+            # User has active search filter
+            result = messagebox.askyesnocancel(
+                "Export Scope",
+                f"You have an active search filter showing {len(self.tree.get_children())} of {len(self.tracked_changes_agent.change_data)} changes.\n\n"
+                "Yes = Export filtered results only\n"
+                "No = Export all tracked changes\n"
+                "Cancel = Cancel export"
+            )
+            if result is None:  # Cancel
+                return
+            export_filtered = result
+        else:
+            export_filtered = False
+        
+        # Get the data to export
+        if export_filtered:
+            exact_match = self.exact_match_var.get()
+            data_to_export = self.tracked_changes_agent.search_changes(search_text, exact_match)
+            default_filename = "tracked_changes_filtered_report.md"
+        else:
+            data_to_export = self.tracked_changes_agent.change_data
+            default_filename = "tracked_changes_report.md"
+        
+        # Ask for save location
+        filepath = filedialog.asksaveasfilename(
+            title="Export Tracked Changes Report",
+            defaultextension=".md",
+            filetypes=(("Markdown files", "*.md"), ("All files", "*.*")),
+            initialfile=default_filename
+        )
+        
+        if not filepath:
+            return
+        
+        # Ask if user wants AI analysis
+        ai_analysis = messagebox.askyesno(
+            "AI Analysis",
+            f"Generate AI-powered change summaries?\n\n"
+            f"This will analyze {len(data_to_export)} changes using the currently selected AI model.\n\n"
+            f"Note: This may take a few minutes and will use API credits.\n\n"
+            f"Click 'No' to export without AI analysis."
+        )
+        
+        # If AI analysis enabled, let user choose batch size
+        batch_size = 25  # Default
+        if ai_analysis:
+            batch_dialog = tk.Toplevel(self.window)
+            batch_dialog.title("Batch Size Configuration")
+            batch_dialog.geometry("450x280")
+            batch_dialog.transient(self.window)
+            batch_dialog.grab_set()
+            
+            tk.Label(batch_dialog, text="Configure Batch Processing", 
+                    font=("Segoe UI", 11, "bold")).pack(pady=10)
+            tk.Label(batch_dialog, 
+                    text=f"Choose how many segments to process per AI request\n"
+                         f"Larger batches = faster but more tokens per request",
+                    font=("Segoe UI", 9)).pack(pady=5)
+            
+            # Slider for batch size
+            batch_var = tk.IntVar(value=25)
+            
+            slider_frame = tk.Frame(batch_dialog)
+            slider_frame.pack(pady=10, fill='x', padx=20)
+            
+            tk.Label(slider_frame, text="Batch Size:", font=("Segoe UI", 9)).pack(side='left')
+            batch_label = tk.Label(slider_frame, text="25", font=("Segoe UI", 10, "bold"), fg="blue")
+            batch_label.pack(side='right')
+            
+            def update_label(val):
+                batch_label.config(text=str(int(float(val))))
+            
+            slider = tk.Scale(batch_dialog, from_=1, to=100, orient='horizontal',
+                            variable=batch_var, command=update_label, length=350)
+            slider.pack(pady=5)
+            
+            # Info label
+            info_label = tk.Label(batch_dialog, 
+                                text=f"Total changes: {len(data_to_export)} | "
+                                     f"Estimated batches at size 25: {(len(data_to_export) + 24) // 25}",
+                                font=("Segoe UI", 8), fg="gray")
+            info_label.pack(pady=5)
+            
+            def update_info(*args):
+                size = batch_var.get()
+                batches = (len(data_to_export) + size - 1) // size
+                info_label.config(text=f"Total changes: {len(data_to_export)} | "
+                                      f"Estimated batches at size {size}: {batches}")
+            
+            batch_var.trace('w', update_info)
+            
+            # OK button
+            def on_ok():
+                nonlocal batch_size
+                batch_size = batch_var.get()
+                batch_dialog.destroy()
+            
+            tk.Button(batch_dialog, text="OK", command=on_ok, 
+                     font=("Segoe UI", 10), width=15).pack(pady=10)
+            
+            # Wait for dialog to close
+            batch_dialog.wait_window()
+        
+        try:
+            # Prepare report content
+            timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+            
+            # Build AI prompt info for report header
+            ai_prompt_info = ""
+            if ai_analysis and hasattr(self, 'parent_app') and self.parent_app:
+                provider = self.parent_app.current_llm_provider
+                model = self.parent_app.current_llm_model
+                # Capitalize provider name for display
+                provider_display = provider.capitalize()
+                ai_prompt_info = f"""
+
+### AI Analysis Configuration
+
+**Provider:** {provider_display}  
+**Model:** {model}
+
+**Prompt Template Used:**
+```
+You are a precision editor analyzing changes between two versions of text.
+Compare the original and revised text and identify EXACTLY what changed.
+
+CRITICAL INSTRUCTIONS:
+- Be extremely specific and precise
+- PAY SPECIAL ATTENTION to quote marks: " vs " vs " (curly vs straight)
+- Check for apostrophe changes: ' vs ' (curly vs straight)  
+- Check for dash changes: - vs – vs — (hyphen vs en-dash vs em-dash)
+- Quote the exact words/phrases that changed
+- Use this format: "X" → "Y"
+- For single word changes: quote both words
+- For multiple changes: put each on its own line
+- For punctuation/formatting: describe precisely
+- For additions: "Added: [exact text]"
+- For deletions: "Removed: [exact text]"
+- DO NOT say "No change" unless texts are 100% identical
+- DO NOT use vague terms like "clarified", "improved", "fixed"
+- DO quote the actual changed text
+
+Examples of single changes:
+✓ "pre-cut" → "incision"
+✓ Curly quotes → straight quotes: "word" → "word"
+✓ Curly apostrophe → straight: don't → don't
+✓ "package" → "packaging"
+
+Examples of multiple changes (one per line):
+✓ "split portions" → "divided portions"
+  "connected by a" → "connected, via a"
+  Curly quotes → straight quotes throughout
+✓ Added: "carefully"
+✓ "color" → "colour" (US to UK spelling)
+✗ Clarified terminology (too vague)
+✗ Fixed grammar (not specific)
+✗ Improved word choice (not helpful)
+```
+
+---
+
+"""
+            
+            md_content = f"""# Tracked Changes Analysis Report ([Supervertaler](https://github.com/michaelbeijer/Supervertaler) {self.APP_VERSION})
+
+## What is this report?
+
+This report analyzes the differences between AI-generated translations and your final edited versions exported from your CAT tool (memoQ, CafeTran, etc.). It shows exactly what you changed during post-editing, helping you review your editing decisions and track your translation workflow improvements.
+
+**Use case:** After completing a translation project in your CAT tool with tracked changes enabled, export the bilingual document and load it here to see a detailed breakdown of all modifications made to the AI-generated baseline.
+
+---
+
+**Generated:** {timestamp}  
+**Total Changes:** {len(data_to_export)}  
+**Filter Applied:** {"Yes - " + search_text if export_filtered else "No"}  
+**AI Analysis:** {"Enabled" if ai_analysis else "Disabled"}
+{ai_prompt_info}
+"""
+            
+            # Process changes with paragraph format
+            if ai_analysis:
+                # Show progress window
+                self.log_queue.put(f"[Export] Generating AI summaries for {len(data_to_export)} changes in batches...")
+                
+                progress_window = tk.Toplevel(self.window)
+                progress_window.title("Generating AI Analysis...")
+                progress_window.geometry("400x150")
+                progress_window.transient(self.window)
+                progress_window.grab_set()
+                
+                tk.Label(progress_window, text="Analyzing tracked changes with AI (batched)...", 
+                        font=("Segoe UI", 10)).pack(pady=10)
+                progress_label = tk.Label(progress_window, text="Processing batch 0 of 0")
+                progress_label.pack()
+                batch_info_label = tk.Label(progress_window, text="", font=("Segoe UI", 8), fg="gray")
+                batch_info_label.pack()
+                
+                # Process in batches (user-configured)
+                # batch_size already set from dialog above
+                total_batches = (len(data_to_export) + batch_size - 1) // batch_size
+                all_summaries = {}
+                
+                for batch_num in range(total_batches):
+                    start_idx = batch_num * batch_size
+                    end_idx = min(start_idx + batch_size, len(data_to_export))
+                    batch = data_to_export[start_idx:end_idx]
+                    
+                    progress_label.config(text=f"Processing batch {batch_num + 1} of {total_batches}")
+                    batch_info_label.config(text=f"Segments {start_idx + 1}-{end_idx} of {len(data_to_export)}")
+                    progress_window.update()
+                    
+                    # Generate AI summaries for this batch
+                    try:
+                        batch_summaries = self.get_ai_change_summaries_batch(batch, start_idx)
+                        all_summaries.update(batch_summaries)
+                        self.log_queue.put(f"[Export] Batch {batch_num + 1}/{total_batches} complete ({len(batch)} segments)")
+                    except Exception as e:
+                        self.log_queue.put(f"[Export] Error in batch {batch_num + 1}: {e}")
+                        # Fill in error messages for failed batch
+                        for i in range(start_idx, end_idx):
+                            all_summaries[i] = f"_Error generating summary: {str(e)}_"
+                
+                progress_window.destroy()
+                
+                # Now build the markdown content with the summaries
+                for i, (original, final) in enumerate(data_to_export):
+                    summary = all_summaries.get(i, "_No summary available_")
+                    
+                    # Add segment in paragraph format
+                    md_content += f"""### Segment {i + 1}
+
+**Target (Original):**  
+{original}
+
+**Target (Revised):**  
+{final}
+
+**Change Summary:**  
+{summary}
+
+---
+
+"""
+            else:
+                # No AI analysis - simpler paragraph format
+                for i, (original, final) in enumerate(data_to_export, 1):
+                    md_content += f"""### Segment {i}
+
+**Target (Original):**  
+{original}
+
+**Target (Revised):**  
+{final}
+
+---
+
+"""
+            
+            md_content += f"""
+
+---
+
+## Summary Statistics
+
+- **Total Segments Analyzed:** {len(data_to_export)}
+- **AI Analysis:** {"Enabled" if ai_analysis else "Disabled"}
+- **Export Type:** {"Filtered" if export_filtered else "Complete"}
+
+*This report was generated by [Supervertaler](https://github.com/michaelbeijer/Supervertaler) {self.APP_VERSION}*
+"""
+            
+            # Write to file
+            with open(filepath, 'w', encoding='utf-8') as f:
+                f.write(md_content)
+            
+            messagebox.showinfo(
+                "Export Successful",
+                f"Exported {len(data_to_export)} tracked changes to:\n{filepath}\n\n"
+                + ("AI change summaries included." if ai_analysis else "Export completed without AI analysis.")
+            )
+            
+            self.log_queue.put(f"[Export] Report saved to: {filepath}")
+            
+        except Exception as e:
+            messagebox.showerror(
+                "Export Error",
+                f"Failed to export tracked changes report:\n{str(e)}"
+            )
+            self.log_queue.put(f"[Export] Error: {e}")
+    
+    def get_ai_change_summaries_batch(self, changes_batch, start_index):
+        """Get AI summaries for a batch of changes - much faster than one-by-one"""
+        if not hasattr(self, 'parent_app') or not self.parent_app:
+            # Fallback for batch
+            return {i: "Modified text" for i in range(start_index, start_index + len(changes_batch))}
+        
+        try:
+            provider = self.parent_app.current_llm_provider
+            model_name = self.parent_app.current_llm_model
+            api_key = ""
+            
+            # Debug logging
+            self.log_queue.put(f"[Export] Using provider: {provider}, model: {model_name}")
+            
+            if provider == "claude":
+                api_key = self.parent_app.api_keys.get("claude", "")
+            elif provider == "gemini":
+                api_key = self.parent_app.api_keys.get("google", "")
+            elif provider == "openai":
+                api_key = self.parent_app.api_keys.get("openai", "")
+            
+            if not api_key:
+                self.log_queue.put(f"[Export] ERROR: No API key found for provider: {provider}")
+                return {i: "AI unavailable - no API key" for i in range(start_index, start_index + len(changes_batch))}
+            
+            self.log_queue.put(f"[Export] API key found, calling {provider}...")
+            
+            # Build batch prompt with all changes
+            batch_prompt = """You are a precision editor analyzing changes between multiple text versions.
+For each numbered pair below, identify EXACTLY what changed.
+
+CRITICAL INSTRUCTIONS:
+- Be extremely specific and precise
+- PAY SPECIAL ATTENTION to quote marks: " vs " vs " (curly vs straight)
+- Check for apostrophe changes: ' vs ' (curly vs straight)
+- Check for dash changes: - vs – vs — (hyphen vs en-dash vs em-dash)
+- Quote the exact words/phrases that changed
+- Use format: "X" → "Y"
+- For multiple changes in one segment: put each on its own line
+- For punctuation/formatting: describe precisely (e.g., 'Curly quotes → straight quotes: "word" → "word"')
+- DO NOT say "No change" unless texts are 100% identical (byte-for-byte)
+- DO NOT use vague terms like "clarified", "improved", "fixed"
+- DO quote the actual changed text
+
+IMPORTANT: If only punctuation changed (quotes, apostrophes, dashes), you MUST report it!
+
+"""
+            
+            # Add all changes to the prompt
+            for i, (original, final) in enumerate(changes_batch):
+                batch_prompt += f"""
+[{i + 1}] ORIGINAL: {original}
+    REVISED: {final}
+
+"""
+            
+            batch_prompt += """
+Now provide the change summary for each segment, formatted as:
+
+[1] your precise summary here
+[2] your precise summary here
+[3] your precise summary here
+
+(etc. for all segments)"""
+            
+            # Call AI based on provider
+            self.log_queue.put(f"[Export] Checking provider condition: {provider} == gemini? {provider == 'gemini'}, GEMINI_AVAILABLE? {self.GEMINI_AVAILABLE}")
+            if provider == "gemini" and self.GEMINI_AVAILABLE:
+                import google.generativeai as genai
+                genai.configure(api_key=api_key)
+                model = genai.GenerativeModel(model_name)
+                
+                response = model.generate_content(batch_prompt)
+                response_text = response.text.strip()
+                
+            elif provider == "claude" and self.ANTHROPIC_AVAILABLE:
+                import anthropic
+                client = anthropic.Anthropic(api_key=api_key)
+                
+                message = client.messages.create(
+                    model=model_name,
+                    max_tokens=2000,  # Larger for batch
+                    messages=[{
+                        "role": "user",
+                        "content": batch_prompt
+                    }]
+                )
+                
+                response_text = message.content[0].text.strip()
+                
+            elif provider == "openai" and self.OPENAI_AVAILABLE:
+                import openai
+                client = openai.OpenAI(api_key=api_key)
+                
+                response = client.chat.completions.create(
+                    model=model_name,
+                    max_tokens=2000,  # Larger for batch
+                    messages=[{
+                        "role": "user",
+                        "content": batch_prompt
+                    }]
+                )
+                
+                response_text = response.choices[0].message.content.strip()
+            else:
+                self.log_queue.put(f"[Export] ERROR: No matching provider condition for {provider}")
+                return {i: "Provider not available" for i in range(start_index, start_index + len(changes_batch))}
+            
+            # Parse the response to extract individual summaries
+            summaries = {}
+            current_num = None
+            current_summary_lines = []
+            
+            for line in response_text.split('\n'):
+                line = line.strip()
+                if not line:
+                    continue
+                
+                # Check if line starts with [N]
+                match = re.match(r'^\[(\d+)\]\s*(.*)$', line)
+                if match:
+                    # Save previous summary if any
+                    if current_num is not None:
+                        summary_text = '\n'.join(current_summary_lines).strip()
+                        summaries[start_index + current_num - 1] = summary_text
+                    
+                    # Start new summary
+                    current_num = int(match.group(1))
+                    summary_start = match.group(2).strip()
+                    current_summary_lines = [summary_start] if summary_start else []
+                elif current_num is not None:
+                    # Continuation of current summary
+                    current_summary_lines.append(line)
+            
+            # Save last summary
+            if current_num is not None:
+                summary_text = '\n'.join(current_summary_lines).strip()
+                summaries[start_index + current_num - 1] = summary_text
+            
+            # Fill in any missing summaries
+            for i in range(len(changes_batch)):
+                if (start_index + i) not in summaries:
+                    summaries[start_index + i] = "_Summary not parsed correctly_"
+            
+            return summaries
+            
+        except Exception as e:
+            self.log_queue.put(f"[AI Batch] Error: {e}")
+            return {i: f"Analysis failed: {str(e)}" for i in range(start_index, start_index + len(changes_batch))}
+
+
+def format_tracked_changes_context(tracked_changes_list, max_length=1000):
+    """Format tracked changes for AI context, keeping within token limits"""
+    if not tracked_changes_list:
+        return ""
+    
+    context_parts = ["TRACKED CHANGES REFERENCE (Original→Final editing patterns):"]
+    current_length = len(context_parts[0])
+    
+    for i, (original, final) in enumerate(tracked_changes_list):
+        change_text = f"• \"{original}\" → \"{final}\""
+        if current_length + len(change_text) > max_length:
+            if i > 0:  # Only add if we have at least one example
+                context_parts.append("(Additional examples truncated to save space)")
+            break
+        context_parts.append(change_text)
+        current_length += len(change_text)
+    
+    return "\n".join(context_parts) + "\n"