pyxtxt 0.3.4__tar.gz → 0.3.4.2__tar.gz

{pyxtxt-0.3.4/src/pyxtxt.egg-info → pyxtxt-0.3.4.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pyxtxt
-Version: 0.3.4
+Version: 0.3.4.2
 Summary: A Python library for extracting text from different types of files (PDF, DOCX, PPTX, XLSX, ODT, ecc.).
 Author-email: Giuseppe Levi <giuseppe.levi@gmail.com>
 License: MIT License
@@ -57,7 +57,6 @@ Provides-Extra: html
 Requires-Dist: beautifulsoup4; extra == "html"
 Requires-Dist: lxml; extra == "html"
 Provides-Extra: doc
-Requires-Dist: textract; extra == "doc"
 Provides-Extra: markdown
 Requires-Dist: markdown; extra == "markdown"
 Requires-Dist: beautifulsoup4; extra == "markdown"
@@ -82,7 +81,6 @@ Provides-Extra: ocr-ollama
 Requires-Dist: ollama; extra == "ocr-ollama"
 Requires-Dist: pillow; extra == "ocr-ollama"
 Provides-Extra: all
-Requires-Dist: textract; extra == "all"
 Requires-Dist: PyMuPDF; extra == "all"
 Requires-Dist: python-docx; extra == "all"
 Requires-Dist: python-pptx; extra == "all"
@@ -196,6 +194,33 @@ Use python-magic-bin instead of python-magic for easier installation.
 
 Dependencies are automatically installed based on selected optional groups.
 
+### System Dependencies
+Some extractors require system-level tools to be installed:
+
+- **Legacy DOC files**: `antiword` - Install via your package manager:
+  ```bash
+  # Ubuntu/Debian
+  sudo apt install antiword
+  
+  # macOS
+  brew install antiword
+  
+  # CentOS/RHEL
+  sudo yum install antiword
+  ```
+
+- **Audio/Video transcription**: `ffmpeg` - Required for audio preprocessing:
+  ```bash
+  # Ubuntu/Debian
+  sudo apt install ffmpeg
+  
+  # macOS
+  brew install ffmpeg
+  
+  # Windows
+  # Download from https://ffmpeg.org/download.html
+  ```
+
 ## 📚 Usage Examples
 
 ### Basic Usage

{pyxtxt-0.3.4 → pyxtxt-0.3.4.2}/README.md

@@ -92,6 +92,33 @@ Use python-magic-bin instead of python-magic for easier installation.
 
 Dependencies are automatically installed based on selected optional groups.
 
+### System Dependencies
+Some extractors require system-level tools to be installed:
+
+- **Legacy DOC files**: `antiword` - Install via your package manager:
+  ```bash
+  # Ubuntu/Debian
+  sudo apt install antiword
+  
+  # macOS
+  brew install antiword
+  
+  # CentOS/RHEL
+  sudo yum install antiword
+  ```
+
+- **Audio/Video transcription**: `ffmpeg` - Required for audio preprocessing:
+  ```bash
+  # Ubuntu/Debian
+  sudo apt install ffmpeg
+  
+  # macOS
+  brew install ffmpeg
+  
+  # Windows
+  # Download from https://ffmpeg.org/download.html
+  ```
+
 ## 📚 Usage Examples
 
 ### Basic Usage

{pyxtxt-0.3.4 → pyxtxt-0.3.4.2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "pyxtxt"
-version = "0.3.4"
+version = "0.3.4.2"
 description = "A Python library for extracting text from different types of files (PDF, DOCX, PPTX, XLSX, ODT, ecc.)."
 classifiers = [
   "Development Status :: 4 - Beta",
@@ -50,7 +50,6 @@ html = [
   "lxml",
 ]
 doc = [
-"textract",
 ]
 markdown = [
   "markdown",
@@ -85,7 +84,6 @@ ocr-ollama = [
   "pillow",
 ]
 all = [
-  "textract",
   "PyMuPDF",
   "python-docx",
   "python-pptx",

{pyxtxt-0.3.4 → pyxtxt-0.3.4.2}/src/pyxtxt/__init__.py

@@ -5,13 +5,15 @@ try:
     from .estrattori.ocr_ollama import (
         set_ollama_model, get_ollama_model, xtxt_image_describe,
         set_ollama_config, get_ollama_config, reset_ollama_config,
-        xtxt_image_with_confidence
+        xtxt_image_with_confidence, configure_for_medical_images,
+        configure_for_xray_images, quick_xray_analysis
     )
     __all__ = [
         "xtxt", "extxt_available_formats", "xtxt_from_url", 
         "set_ollama_model", "get_ollama_model", "xtxt_image_describe",
         "set_ollama_config", "get_ollama_config", "reset_ollama_config",
-        "xtxt_image_with_confidence"
+        "xtxt_image_with_confidence", "configure_for_medical_images",
+        "configure_for_xray_images", "quick_xray_analysis"
     ]
 except ImportError:
     __all__ = ["xtxt", "extxt_available_formats", "xtxt_from_url"]

{pyxtxt-0.3.4 → pyxtxt-0.3.4.2}/src/pyxtxt/estrattori/ocr_ollama.py

@@ -19,7 +19,12 @@ OLLAMA_CONFIG = {
     'temperature': 0.1,  # Response creativity (0.0-1.0)
     'max_tokens': 1500,  # Maximum response length
     'confidence_threshold': 0.7,  # Minimum confidence for text extraction
-    'context': 'general'  # Context hint: general, cookbook, document, diagram, etc.
+    'context': 'general',  # Context hint: general, cookbook, document, diagram, medical, xray
+    'enhance_image': True,  # Apply image enhancement preprocessing
+    'min_size': 800,  # Minimum image size for processing (upscale if smaller)
+    'max_size': 2048,   # Maximum image size (downscale if larger)
+    'auto_fallback': True,  # Enable automatic model fallback for better results
+    'fallback_models': ['gemma3:27b', 'gemma3:12b', 'llava:13b']  # Models to try if primary fails
 }
 
 def set_ollama_model(model_name: str):
@@ -49,7 +54,12 @@ def set_ollama_config(**kwargs):
     - language: Language hint ('auto', 'italian', 'english', 'spanish', etc.)
     - caption_length: Caption length ('short', 'medium', 'long')
     - style: Caption style ('descriptive', 'technical', 'simple', 'detailed')
-    - context: Content context ('general', 'cookbook', 'document', 'handwriting', 'technical')
+    - context: Content context ('general', 'cookbook', 'document', 'handwriting', 'technical', 'medical', 'xray')
+    - enhance_image: Apply image enhancement preprocessing (default: True)
+    - min_size: Minimum image size for processing - upscale if smaller (default: 800)
+    - max_size: Maximum image size - downscale if larger (default: 2048)
+    - auto_fallback: Enable automatic model fallback for better results (default: True)
+    - fallback_models: List of models to try if primary fails (default: ['gemma3:27b', 'gemma3:12b', 'llava:13b'])
     - temperature: Response creativity 0.0-1.0 (default: 0.1)
     - max_tokens: Maximum response length (default: 1500)
     - confidence_threshold: Text extraction confidence 0.0-1.0 (default: 0.7)
@@ -58,6 +68,10 @@ def set_ollama_config(**kwargs):
         set_ollama_config(language='italian', style='detailed')
         set_ollama_config(context='document', caption_length='long')
         set_ollama_config(context='handwriting', temperature=0.2)
+        set_ollama_config(context='medical', enhance_image=True, min_size=1024)
+        set_ollama_config(context='xray', style='technical', confidence_threshold=0.8)
+        set_ollama_config(auto_fallback=False)  # Disable fallback for faster processing
+        set_ollama_config(fallback_models=['llava:13b', 'gemma3:27b'])  # Custom fallback sequence
     """
     global OLLAMA_CONFIG
     for key, value in kwargs.items():
@@ -71,6 +85,105 @@ def get_ollama_config():
     """Get current Ollama configuration"""
     return OLLAMA_CONFIG.copy()
 
+def configure_for_medical_images():
+    """
+    Quick configuration setup for optimal medical image processing.
+    Configures model, enhancement, and context for X-rays and medical scans.
+    """
+    print("🏥 Configuring for medical image processing...")
+    
+    # Optimal medical settings  
+    global OLLAMA_CONFIG
+    OLLAMA_CONFIG.update({
+        'context': 'medical',
+        'enhance_image': True,
+        'min_size': 1024,  # Higher resolution for medical details
+        'max_size': 2048,  # Keep high quality
+        'style': 'technical',
+        'confidence_threshold': 0.8,  # Higher threshold for medical accuracy
+        'temperature': 0.05,  # Very low temperature for precision
+        'auto_fallback': True,
+        'fallback_models': ['gemma3:27b', 'llava:13b', 'gemma3:12b']  # Best models first
+    })
+    
+    # Suggest high-quality model if current is default
+    current_model = get_ollama_model()
+    if current_model == "gemma3:4b":
+        print("💡 Consider upgrading to gemma3:27b for better medical image recognition")
+        print("   Run: set_ollama_model('gemma3:27b')")
+    
+    print("✅ Medical configuration applied:")
+    print(f"   - Enhanced image processing: {OLLAMA_CONFIG['enhance_image']}")
+    print(f"   - Minimum resolution: {OLLAMA_CONFIG['min_size']}px")
+    print(f"   - Confidence threshold: {OLLAMA_CONFIG['confidence_threshold']}")
+    print(f"   - Fallback models: {len(OLLAMA_CONFIG['fallback_models'])} configured")
+
+def configure_for_xray_images():
+    """
+    Specialized configuration for X-ray and radiological image processing.
+    Optimized for detecting small text, markers, and technical annotations.
+    """
+    print("📷 Configuring for X-ray image processing...")
+    
+    # X-ray specific settings
+    global OLLAMA_CONFIG
+    OLLAMA_CONFIG.update({
+        'context': 'xray',
+        'enhance_image': True,
+        'min_size': 1200,  # Even higher for X-ray details
+        'max_size': 2048,
+        'style': 'technical',
+        'confidence_threshold': 0.85,  # Very high threshold for X-ray accuracy
+        'temperature': 0.02,  # Minimal creativity for technical precision
+        'auto_fallback': True,
+        'fallback_models': ['gemma3:27b', 'llava:13b']  # Only best models
+    })
+    
+    # Recommend best model for X-rays
+    current_model = get_ollama_model()
+    if current_model != "gemma3:27b":
+        print("🎯 For best X-ray results, use gemma3:27b model")
+        print("   Run: set_ollama_model('gemma3:27b')")
+    
+    print("✅ X-ray configuration applied:")
+    print(f"   - Context: Medical X-ray specialization")
+    print(f"   - Enhanced processing: Advanced medical filters")
+    print(f"   - High precision mode: {OLLAMA_CONFIG['confidence_threshold']} threshold")
+    print(f"   - Temperature: {OLLAMA_CONFIG['temperature']} (maximum precision)")
+
+def quick_xray_analysis(image_path: str) -> str:
+    """
+    Convenient one-function X-ray analysis with optimal settings.
+    Automatically configures system for X-ray processing and returns detailed analysis.
+    
+    Args:
+        image_path: Path to X-ray image file
+        
+    Returns:
+        str: Detailed text + description analysis
+    """
+    # Save current config
+    global OLLAMA_CONFIG
+    original_config = OLLAMA_CONFIG.copy()
+    original_model = get_ollama_model()
+    
+    try:
+        # Apply X-ray configuration  
+        configure_for_xray_images()
+        set_ollama_model('gemma3:27b')  # Use best model
+        
+        # Perform analysis
+        print("🔍 Analyzing X-ray image...")
+        result = xtxt_image_describe(image_path)
+        
+        return result
+        
+    finally:
+        # Restore original configuration
+        OLLAMA_CONFIG = original_config
+        set_ollama_model(original_model)
+        print("🔄 Configuration restored")
+
 def reset_ollama_config():
     """Reset Ollama configuration to defaults"""
     global OLLAMA_CONFIG
@@ -81,10 +194,125 @@ def reset_ollama_config():
         'temperature': 0.1,
         'max_tokens': 1500,
         'confidence_threshold': 0.7,
-        'context': 'general'
+        'context': 'general',
+        'enhance_image': True,
+        'min_size': 800,
+        'max_size': 2048,
+        'auto_fallback': True,
+        'fallback_models': ['gemma3:27b', 'gemma3:12b', 'llava:13b']
     }
     print("✅ Ollama configuration reset to defaults")
 
+def _enhance_image(image: Image.Image, context: str, min_size: int, max_size: int) -> Image.Image:
+    """
+    Apply image enhancement preprocessing for better OCR results.
+    
+    Args:
+        image: PIL Image object
+        context: Content context hint (medical, xray, document, etc.)
+        min_size: Minimum size threshold for upscaling
+        max_size: Maximum size threshold for downscaling
+    
+    Returns:
+        Enhanced PIL Image
+    """
+    try:
+        from PIL import ImageEnhance, ImageFilter, ImageOps
+    except ImportError:
+        # If PIL enhancements not available, return original
+        return image
+    
+    enhanced = image.copy()
+    
+    # Get current dimensions
+    width, height = enhanced.size
+    max_dimension = max(width, height)
+    
+    # Resize if needed (quality improvement for small images, memory management for large)
+    if max_dimension < min_size:
+        # Upscale small images for better model processing
+        scale_factor = min_size / max_dimension
+        new_width = int(width * scale_factor)
+        new_height = int(height * scale_factor) 
+        enhanced = enhanced.resize((new_width, new_height), Image.Resampling.LANCZOS)
+        print(f"📈 Image upscaled from {width}x{height} to {new_width}x{new_height}")
+    elif max_dimension > max_size:
+        # Downscale large images to manageable size
+        scale_factor = max_size / max_dimension
+        new_width = int(width * scale_factor)
+        new_height = int(height * scale_factor)
+        enhanced = enhanced.resize((new_width, new_height), Image.Resampling.LANCZOS)
+        print(f"📉 Image downscaled from {width}x{height} to {new_width}x{new_height}")
+    
+    # Context-specific enhancements
+    if context in ['medical', 'xray']:
+        print("🏥 Applying medical image enhancements...")
+        # Medical images often benefit from:
+        # 1. Contrast enhancement to bring out subtle details
+        contrast = ImageEnhance.Contrast(enhanced)
+        enhanced = contrast.enhance(1.3)
+        
+        # 2. Sharpening to improve edge definition 
+        enhanced = enhanced.filter(ImageFilter.UnsharpMask(radius=1, percent=120, threshold=3))
+        
+        # 3. Brightness adjustment for dark X-rays
+        brightness = ImageEnhance.Brightness(enhanced)
+        enhanced = brightness.enhance(1.1)
+        
+    elif context in ['document', 'handwriting', 'technical']:
+        print("📄 Applying document enhancement...")
+        # Documents benefit from:
+        # 1. Moderate sharpening for text clarity
+        enhanced = enhanced.filter(ImageFilter.UnsharpMask(radius=0.5, percent=150, threshold=3))
+        
+        # 2. Contrast boost for faded text
+        contrast = ImageEnhance.Contrast(enhanced)
+        enhanced = contrast.enhance(1.2)
+        
+    elif context == 'general':
+        print("🔧 Applying general enhancements...")
+        # General purpose light enhancement
+        # 1. Slight contrast improvement
+        contrast = ImageEnhance.Contrast(enhanced)
+        enhanced = contrast.enhance(1.1)
+        
+        # 2. Subtle sharpening
+        enhanced = enhanced.filter(ImageFilter.UnsharpMask(radius=0.5, percent=110, threshold=3))
+    
+    return enhanced
+
+def _try_ollama_request(prompt: str, img_base64: str, current_model: str, config: dict) -> tuple:
+    """
+    Attempt Ollama request with a specific model.
+    
+    Returns:
+        tuple: (success: bool, response: str, confidence: float)
+    """
+    try:
+        print(f"🤖 Trying model: {current_model}")
+        
+        response = ollama.generate(
+            model=current_model,
+            prompt=prompt,
+            images=[img_base64],
+            options={
+                'temperature': config['temperature'],
+                'top_p': 0.9,
+                'num_predict': config['max_tokens']
+            }
+        )
+        
+        extracted_content = response.get('response', '').strip()
+        
+        # Calculate confidence without printing warnings here (let parent handle it)
+        confidence_score = _calculate_confidence_score(extracted_content, "ocr" if "Extracted text:" in prompt else "describe")
+        
+        return True, extracted_content, confidence_score
+        
+    except Exception as e:
+        print(f"❌ Model {current_model} failed: {e}")
+        return False, "", 0.0
+
 def _calculate_confidence_score(content: str, mode: str) -> float:
     """
     Calculate confidence score (0.0-1.0) for OCR/caption quality.
@@ -261,14 +489,19 @@ if ollama and Image:
             if image.mode != 'RGB':
                 image = image.convert('RGB')
             
+            # Get current configuration
+            config = OLLAMA_CONFIG
+            
+            # Apply image enhancement if enabled
+            if config.get('enhance_image', True):
+                print("⚡ Enhancing image for better OCR...")
+                image = _enhance_image(image, config['context'], config['min_size'], config['max_size'])
+            
             # Convert image to base64
             buffered = BytesIO()
             image.save(buffered, format="PNG")
             img_base64 = base64.b64encode(buffered.getvalue()).decode()
             
-            # Get current configuration
-            config = OLLAMA_CONFIG
-            
             # Build language hint
             lang_hint = ""
             if config['language'] != 'auto':
@@ -276,7 +509,43 @@ if ollama and Image:
             
             # Different prompts based on mode
             if mode == "ocr":
-                prompt = f"""Look carefully at this image and extract ALL text that you can see, including:
+                # Context-specific OCR prompts
+                if config.get('context') == 'xray':
+                    prompt = f"""This is a medical X-ray or radiological image. Look carefully and extract ALL visible text, including:
+- Patient identification numbers, names, or codes
+- Date and time stamps (exam dates, birth dates)
+- Anatomical position markers (L/R, LEFT/RIGHT, AP, LAT)
+- Measurement scales, rulers, or calibration marks
+- Technical annotations or radiologist markings
+- Equipment identifiers or hospital names
+- Any small text on borders or corners
+
+IMPORTANT:
+- Medical images often have small text around borders - examine carefully
+- {lang_hint}Look for technical markings that might be faint or small
+- Include any numbers that might be measurements or identifiers
+- If absolutely no readable text is visible, respond with 'NO_TEXT_FOUND'
+
+Extracted text:"""
+                elif config.get('context') == 'medical':
+                    prompt = f"""This appears to be a medical document or image. Look carefully and extract ALL text, including:
+- Patient information (names, IDs, dates of birth)
+- Medical terminology and diagnostic information
+- Dates, times, and timestamps
+- Measurements, values, and test results
+- Doctor names, hospital information, department names
+- Small print and technical annotations
+
+IMPORTANT:
+- Medical documents often contain critical small text - examine thoroughly
+- {lang_hint}Include all numerical values as they may be measurements
+- Preserve formatting for medical data accuracy
+- If no readable text is found, respond with 'NO_TEXT_FOUND'
+
+Extracted text:"""
+                else:
+                    # General OCR prompt
+                    prompt = f"""Look carefully at this image and extract ALL text that you can see, including:
 - Titles, headings, and main text content  
 - Small print, captions, labels, and annotations
 - Numbers, measurements, quantities, and symbols
@@ -329,6 +598,17 @@ Extracted text:"""
                     context_hint = """
    - Focus on technical elements: labels, measurements, specifications, diagrams
    - Include any mathematical formulas, technical symbols, or engineering notations"""
+                elif context == 'medical':
+                    context_hint = """
+   - Focus on medical content: patient data, measurements, anatomical labels, medical terminology
+   - Look for dates, patient IDs, measurement values, diagnostic information
+   - Note any visible text on medical equipment or instrumentation"""
+                elif context == 'xray':
+                    context_hint = """
+   - This appears to be a medical X-ray or radiological image
+   - Look for: anatomical markers, measurement scales, patient information, timestamps
+   - Focus on any visible text annotations, labels, or technical markings
+   - Note positioning indicators (L/R, anterior/posterior) or measurement rulers"""
 
                 prompt = f"""Analyze this image and provide:
 1. All visible text exactly as written (preserve formatting, line breaks, bullet points)
@@ -341,23 +621,31 @@ Format:
 TEXT: [all visible text here, or NO_TEXT_FOUND if none]
 DESCRIPTION: [image description following the guidelines above]"""
             
-            # Send request to Ollama with configured parameters
-            response = ollama.generate(
-                model=current_model,
-                prompt=prompt,
-                images=[img_base64],
-                options={
-                    'temperature': config['temperature'],
-                    'top_p': 0.9,
-                    'num_predict': config['max_tokens']
-                }
-            )
-            
-            # Extract and clean response
-            extracted_content = response.get('response', '').strip()
+            # Try primary model first
+            success, extracted_content, confidence_score = _try_ollama_request(prompt, img_base64, current_model, config)
             
-            # Calculate confidence score based on response quality indicators
-            confidence_score = _calculate_confidence_score(extracted_content, mode)
+            # If primary model failed or confidence is very low, try fallback models
+            if config.get('auto_fallback', True) and (not success or confidence_score < 0.3):
+                print("🔄 Primary model result unsatisfactory, trying fallback models...")
+                
+                fallback_models = config.get('fallback_models', [])
+                best_result = (extracted_content, confidence_score) if success else ("", 0.0)
+                
+                for fallback_model in fallback_models:
+                    if fallback_model == current_model:
+                        continue  # Skip if same as primary
+                    
+                    success, fallback_content, fallback_confidence = _try_ollama_request(prompt, img_base64, fallback_model, config)
+                    
+                    if success and fallback_confidence > best_result[1]:
+                        print(f"✨ Better result from {fallback_model} (confidence: {fallback_confidence:.2f} vs {best_result[1]:.2f})")
+                        best_result = (fallback_content, fallback_confidence)
+                        
+                        # Stop if we found a good enough result
+                        if fallback_confidence >= config['confidence_threshold']:
+                            break
+                
+                extracted_content, confidence_score = best_result
             
             # Check confidence threshold
             if confidence_score < config['confidence_threshold']:
@@ -401,14 +689,19 @@ DESCRIPTION: [image description following the guidelines above]"""
             if image.mode != 'RGB':
                 image = image.convert('RGB')
             
+            # Get current configuration
+            config = OLLAMA_CONFIG
+            
+            # Apply image enhancement if enabled
+            if config.get('enhance_image', True):
+                print("⚡ Enhancing image for better OCR...")
+                image = _enhance_image(image, config['context'], config['min_size'], config['max_size'])
+            
             # Convert image to base64
             buffered = BytesIO()
             image.save(buffered, format="PNG")
             img_base64 = base64.b64encode(buffered.getvalue()).decode()
             
-            # Get current configuration
-            config = OLLAMA_CONFIG
-            
             # Build prompts (same logic as main function)
             lang_hint = ""
             if config['language'] != 'auto':
@@ -457,23 +750,26 @@ Format:
 TEXT: [all visible text here, or NO_TEXT_FOUND if none]
 DESCRIPTION: [image description following the guidelines above]"""
             
-            # Send request to Ollama
-            response = ollama.generate(
-                model=current_model,
-                prompt=prompt,
-                images=[img_base64],
-                options={
-                    'temperature': config['temperature'],
-                    'top_p': 0.9,
-                    'num_predict': config['max_tokens']
-                }
-            )
-            
-            # Extract response
-            extracted_content = response.get('response', '').strip()
+            # Try primary model first
+            success, extracted_content, confidence_score = _try_ollama_request(prompt, img_base64, current_model, config)
             
-            # Calculate confidence without threshold filtering
-            confidence_score = _calculate_confidence_score(extracted_content, mode)
+            # Try fallback if enabled and primary result is poor
+            if config.get('auto_fallback', True) and (not success or confidence_score < 0.3):
+                fallback_models = config.get('fallback_models', [])
+                best_result = (extracted_content, confidence_score) if success else ("", 0.0)
+                
+                for fallback_model in fallback_models:
+                    if fallback_model == current_model:
+                        continue
+                    
+                    success, fallback_content, fallback_confidence = _try_ollama_request(prompt, img_base64, fallback_model, config)
+                    
+                    if success and fallback_confidence > best_result[1]:
+                        best_result = (fallback_content, fallback_confidence)
+                        if fallback_confidence >= config['confidence_threshold']:
+                            break
+                
+                extracted_content, confidence_score = best_result
             
             # Return both text and confidence
             return extracted_content, confidence_score

{pyxtxt-0.3.4 → pyxtxt-0.3.4.2/src/pyxtxt.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pyxtxt
-Version: 0.3.4
+Version: 0.3.4.2
 Summary: A Python library for extracting text from different types of files (PDF, DOCX, PPTX, XLSX, ODT, ecc.).
 Author-email: Giuseppe Levi <giuseppe.levi@gmail.com>
 License: MIT License
@@ -57,7 +57,6 @@ Provides-Extra: html
 Requires-Dist: beautifulsoup4; extra == "html"
 Requires-Dist: lxml; extra == "html"
 Provides-Extra: doc
-Requires-Dist: textract; extra == "doc"
 Provides-Extra: markdown
 Requires-Dist: markdown; extra == "markdown"
 Requires-Dist: beautifulsoup4; extra == "markdown"
@@ -82,7 +81,6 @@ Provides-Extra: ocr-ollama
 Requires-Dist: ollama; extra == "ocr-ollama"
 Requires-Dist: pillow; extra == "ocr-ollama"
 Provides-Extra: all
-Requires-Dist: textract; extra == "all"
 Requires-Dist: PyMuPDF; extra == "all"
 Requires-Dist: python-docx; extra == "all"
 Requires-Dist: python-pptx; extra == "all"
@@ -196,6 +194,33 @@ Use python-magic-bin instead of python-magic for easier installation.
 
 Dependencies are automatically installed based on selected optional groups.
 
+### System Dependencies
+Some extractors require system-level tools to be installed:
+
+- **Legacy DOC files**: `antiword` - Install via your package manager:
+  ```bash
+  # Ubuntu/Debian
+  sudo apt install antiword
+  
+  # macOS
+  brew install antiword
+  
+  # CentOS/RHEL
+  sudo yum install antiword
+  ```
+
+- **Audio/Video transcription**: `ffmpeg` - Required for audio preprocessing:
+  ```bash
+  # Ubuntu/Debian
+  sudo apt install ffmpeg
+  
+  # macOS
+  brew install ffmpeg
+  
+  # Windows
+  # Download from https://ffmpeg.org/download.html
+  ```
+
 ## 📚 Usage Examples
 
 ### Basic Usage

{pyxtxt-0.3.4 → pyxtxt-0.3.4.2}/src/pyxtxt.egg-info/requires.txt

@@ -6,7 +6,6 @@ python-magic
 python-magic-bin
 
 [all]
-textract
 PyMuPDF
 python-docx
 python-pptx
@@ -29,7 +28,6 @@ ollama
 openai-whisper
 
 [doc]
-textract
 
 [docx]
 python-docx