diff --git a/analyzer.py b/analyzer.py index ea8ccdd..196e9e7 100644 --- a/analyzer.py +++ b/analyzer.py @@ -261,161 +261,6 @@ def debug_doctags(doctags_text, debug_mode=False): return doctags_text - -def create_html_document(image_path, doctags_text, pil_image, args): - """Create HTML directly from DocTags and image.""" - print("Creating HTML document directly...") - - # Get filename for document title - doc_name = Path(image_path).stem - - # Extract content from doctags - title_match = re.search(r'.*?>(.*?)', doctags_text) - title = title_match.group(1) if title_match else doc_name - - # Clean title by removing any location tags - title = re.sub(r'', '', title) - - # Extract text content - text_matches = re.findall(r'.*?>(.*?)', doctags_text) - paragraphs = [] - for text in text_matches: - # Clean text by removing any location tags - cleaned_text = re.sub(r'', '', text) - paragraphs.append(cleaned_text) - - # Extract list items - list_items = [] - item_matches = re.findall(r'.*?>(.*?)', doctags_text) - for item in item_matches: - # Clean item by removing any location tags - cleaned_item = re.sub(r'', '', item) - list_items.append(cleaned_item) - - # Extract footer - footer_match = re.search(r'.*?>(.*?)', doctags_text) - footer_text = "" - if footer_match: - footer_text = re.sub(r'', '', footer_match.group(1)) - - # Create image data URI - max_dimension = 1200 - img_for_embedding = pil_image - if max(pil_image.size) > max_dimension: - ratio = min(max_dimension / pil_image.size[0], max_dimension / pil_image.size[1]) - new_size = (int(pil_image.size[0] * ratio), int(pil_image.size[1] * ratio)) - img_for_embedding = pil_image.resize(new_size, Image.LANCZOS) - - buffered = BytesIO() - img_for_embedding.save(buffered, format="PNG", optimize=True, quality=90) - img_str = base64.b64encode(buffered.getvalue()).decode() - img_data_uri = f"data:image/png;base64,{img_str}" - - # Build HTML paragraphs - paragraphs_html = "".join([f"

{text}

\n" for text in paragraphs]) if paragraphs else "" - - # Build list HTML - list_html = "" - if list_items: - list_html = "
    \n" - for item in list_items: - list_html += f"
  • {item}
    • \n" - list_html += "
\n" - - # Create footer HTML - footer_html = f"
{footer_text}
\n" if footer_text else "" - - # Create HTML document - html_content = f''' - - - -{title} - - - -
-

{title}

-
- Document Image -
Page {args.page}
-
-{paragraphs_html} -{list_html} -{footer_html} -
- -''' - - return html_content - - -def create_markdown_document(image_path, doctags_text, args): - """Create Markdown document from DocTags.""" - print("Creating Markdown document...") - - # Get filename for document title - doc_name = Path(image_path).stem - - # Extract content from doctags - title_match = re.search(r'.*?>(.*?)', doctags_text) - title = title_match.group(1) if title_match else doc_name - - # Clean title by removing any location tags - title = re.sub(r'', '', title) - - # Extract text content - text_matches = re.findall(r'.*?>(.*?)', doctags_text) - paragraphs = [] - for text in text_matches: - # Clean text by removing any location tags - cleaned_text = re.sub(r'', '', text) - paragraphs.append(cleaned_text) - - # Extract list items - list_items = [] - item_matches = re.findall(r'.*?>(.*?)', doctags_text) - for item in item_matches: - # Clean item by removing any location tags - cleaned_item = re.sub(r'', '', item) - list_items.append(cleaned_item) - - # Extract footer - footer_match = re.search(r'.*?>(.*?)', doctags_text) - footer_text = "" - if footer_match: - footer_text = re.sub(r'', '', footer_match.group(1)) - - # Build markdown content - markdown = f"# {title}\n\n" - markdown += f"*Page {args.page}*\n\n" - - # Add paragraphs - for p in paragraphs: - markdown += f"{p}\n\n" - - # Add list - if list_items: - for item in list_items: - markdown += f"* {item}\n" - markdown += "\n" - - # Add footer - if footer_text: - markdown += f"---\n{footer_text}\n" - - return markdown - - def process_page(args, model, processor, config, image_path, pil_image, page_num=1): """Process a single page from a PDF or image file.""" from mlx_vlm.prompt_utils import apply_chat_template @@ -487,31 +332,6 @@ def process_page(args, model, processor, config, image_path, pil_image, page_num f.write(tags_analysis) print(f"DocTags analysis saved to: {tags_path}") - # If not in DocTags-only mode, create documents - if not args.doctags_only: - # Create markdown document - md_content = create_markdown_document(image_path, output, args) - md_path = results_dir / f"{output_path.stem}.md" - with open(md_path, 'w', encoding='utf-8') as f: - f.write(md_content) - print(f"Markdown saved to: {md_path}") - - # Create HTML document - html_content = create_html_document(image_path, output, pil_image, args) - html_path = results_dir / f"{output_path.stem}.html" - with open(html_path, 'w', encoding='utf-8') as f: - f.write(html_content) - print(f"HTML saved to: {html_path}") - - # Calculate final file size - html_size = os.path.getsize(html_path) - print(f"Final HTML file size: {html_size} bytes") - - # Open in browser if requested (only for the first page or single pages) - if args.show and html_size > 100 and page_num <= 1: - import webbrowser - webbrowser.open(f"file:///{str(html_path.resolve())}") - # Save a copy of the processed image for reference if in debug mode if args.debug: img_debug_path = results_dir / f"{output_path.stem}.debug.png"