pyxtxt 0.2.4__tar.gz → 0.3__tar.gz

{pyxtxt-0.2.4/src/pyxtxt.egg-info → pyxtxt-0.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pyxtxt
-Version: 0.2.4
+Version: 0.3
 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
@@ -78,6 +78,9 @@ Requires-Dist: openai-whisper; extra == "audio"
 Provides-Extra: ocr
 Requires-Dist: easyocr; extra == "ocr"
 Requires-Dist: pillow; extra == "ocr"
+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"
@@ -96,6 +99,7 @@ Requires-Dist: pylatexenc; extra == "all"
 Requires-Dist: openai-whisper; extra == "all"
 Requires-Dist: easyocr; extra == "all"
 Requires-Dist: pillow; extra == "all"
+Requires-Dist: ollama; extra == "all"
 Dynamic: license-file
 
 # PyxTxt
@@ -141,10 +145,13 @@ pip install pyxtxt[pdf,docx,presentation,spreadsheet,html,markdown,epub,email]
 # Audio transcription (~2GB download for Whisper models)
 pip install pyxtxt[audio]
 
-# OCR from images (~1GB download for EasyOCR models)
+# Traditional OCR from images (~1GB download for EasyOCR models)
 pip install pyxtxt[ocr]
 
-# Both audio and OCR
+# AI-powered OCR via Ollama (requires local Ollama + gemma3:4b model)
+pip install pyxtxt[ocr-ollama]
+
+# Both audio and traditional OCR
 pip install pyxtxt[audio,ocr]
 ```
 
@@ -185,6 +192,7 @@ Use python-magic-bin instead of python-magic for easier installation.
 - **LaTeX**: pylatexenc
 - **Audio**: openai-whisper (heavy ~2GB models)
 - **OCR**: easyocr, pillow (heavy ~1GB models)
+- **OCR-Ollama**: ollama, pillow (requires local Ollama server)
 
 Dependencies are automatically installed based on selected optional groups.
 
@@ -249,11 +257,25 @@ text = xtxt(video_response.content)
 ```python
 from pyxtxt import xtxt
 
-# Extract text from images
+# Traditional OCR with EasyOCR (install with: pip install pyxtxt[ocr])
 text = xtxt("scanned_document.png")
 text = xtxt("screenshot.jpg")
 text = xtxt("invoice.tiff")
 
+# AI-powered OCR with Ollama (install with: pip install pyxtxt[ocr-ollama])
+# Requires: ollama server running + gemma3:4b model
+from pyxtxt.estrattori.ocr_ollama import set_ollama_model, xtxt_image_describe
+
+# Configure model (optional, default is gemma3:4b)
+set_ollama_model("gemma3:12b")  # or llava:7b, llava:13b
+
+# Extract only text (OCR mode)
+text = xtxt("complex_document.png")
+
+# Extract text + image description
+description = xtxt_image_describe(open("diagram.png", "rb"))
+# Output: "TEXT: Chart Title: Sales Report 2024\nDESCRIPTION: Bar chart showing quarterly sales data with blue bars"
+
 # From web images
 import requests
 image_response = requests.get("https://example.com/document.png")

{pyxtxt-0.2.4 → pyxtxt-0.3}/README.md

@@ -41,10 +41,13 @@ pip install pyxtxt[pdf,docx,presentation,spreadsheet,html,markdown,epub,email]
 # Audio transcription (~2GB download for Whisper models)
 pip install pyxtxt[audio]
 
-# OCR from images (~1GB download for EasyOCR models)
+# Traditional OCR from images (~1GB download for EasyOCR models)
 pip install pyxtxt[ocr]
 
-# Both audio and OCR
+# AI-powered OCR via Ollama (requires local Ollama + gemma3:4b model)
+pip install pyxtxt[ocr-ollama]
+
+# Both audio and traditional OCR
 pip install pyxtxt[audio,ocr]
 ```
 
@@ -85,6 +88,7 @@ Use python-magic-bin instead of python-magic for easier installation.
 - **LaTeX**: pylatexenc
 - **Audio**: openai-whisper (heavy ~2GB models)
 - **OCR**: easyocr, pillow (heavy ~1GB models)
+- **OCR-Ollama**: ollama, pillow (requires local Ollama server)
 
 Dependencies are automatically installed based on selected optional groups.
 
@@ -149,11 +153,25 @@ text = xtxt(video_response.content)
 ```python
 from pyxtxt import xtxt
 
-# Extract text from images
+# Traditional OCR with EasyOCR (install with: pip install pyxtxt[ocr])
 text = xtxt("scanned_document.png")
 text = xtxt("screenshot.jpg")
 text = xtxt("invoice.tiff")
 
+# AI-powered OCR with Ollama (install with: pip install pyxtxt[ocr-ollama])
+# Requires: ollama server running + gemma3:4b model
+from pyxtxt.estrattori.ocr_ollama import set_ollama_model, xtxt_image_describe
+
+# Configure model (optional, default is gemma3:4b)
+set_ollama_model("gemma3:12b")  # or llava:7b, llava:13b
+
+# Extract only text (OCR mode)
+text = xtxt("complex_document.png")
+
+# Extract text + image description
+description = xtxt_image_describe(open("diagram.png", "rb"))
+# Output: "TEXT: Chart Title: Sales Report 2024\nDESCRIPTION: Bar chart showing quarterly sales data with blue bars"
+
 # From web images
 import requests
 image_response = requests.get("https://example.com/document.png")

{pyxtxt-0.2.4 → pyxtxt-0.3}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "pyxtxt"
-version = "0.2.4"
+version = "0.3"
 description = "A Python library for extracting text from different types of files (PDF, DOCX, PPTX, XLSX, ODT, ecc.)."
 classifiers = [
   "Development Status :: 4 - Beta",
@@ -80,6 +80,10 @@ ocr = [
   "easyocr",
   "pillow",
 ]
+ocr-ollama = [
+  "ollama",
+  "pillow",
+]
 all = [
   "textract",
   "PyMuPDF",
@@ -98,6 +102,7 @@ all = [
   "openai-whisper",
   "easyocr",
   "pillow",
+  "ollama",
 ]
 
 [build-system]

pyxtxt-0.3/src/pyxtxt/__init__.py

@@ -0,0 +1,8 @@
+from .core import xtxt, extxt_available_formats, xtxt_from_url
+
+# Import OCR-Ollama functions if available
+try:
+    from .estrattori.ocr_ollama import set_ollama_model, get_ollama_model, xtxt_image_describe
+    __all__ = ["xtxt", "extxt_available_formats", "xtxt_from_url", "set_ollama_model", "get_ollama_model", "xtxt_image_describe"]
+except ImportError:
+    __all__ = ["xtxt", "extxt_available_formats", "xtxt_from_url"]

pyxtxt-0.3/src/pyxtxt/estrattori/ocr_ollama.py

@@ -0,0 +1,125 @@
+# pyxtxt/extractors/image_ocr_ollama.py
+from . import register_extractor
+from io import BytesIO
+import base64
+
+try:
+    import ollama
+    from PIL import Image
+except ImportError:
+    ollama = None
+    Image = None
+
+# Global configuration for Ollama model
+OLLAMA_MODEL = "gemma3:4b"  # Default multimodal model
+
+def set_ollama_model(model_name: str):
+    """
+    Set the Ollama model to use for OCR.
+    
+    Recommended multimodal models:
+    - gemma3:4b (default, balanced speed/quality)
+    - gemma3:12b (higher quality, slower)
+    - gemma3:27b (best quality, very slow)
+    - llava:7b (alternative vision model)
+    - llava:13b (higher quality LLAVA)
+    """
+    global OLLAMA_MODEL
+    OLLAMA_MODEL = model_name
+    print(f"✅ Ollama OCR model set to: {model_name}")
+
+def get_ollama_model():
+    """Get current Ollama model name"""
+    return OLLAMA_MODEL
+
+if ollama and Image:
+    def xtxt_image_ocr_ollama(file_buffer, mode="ocr", model=None):
+        """
+        Extract text from images using Ollama with multimodal models.
+        
+        Args:
+            file_buffer: Image file buffer
+            mode: "ocr" (text only) or "describe" (text + description)  
+            model: Override default model (optional)
+        """
+        try:
+            # Use specified model or global default
+            current_model = model or OLLAMA_MODEL
+            
+            # Convert buffer to PIL Image
+            image = Image.open(BytesIO(file_buffer.read()))
+            
+            # Convert to RGB if needed
+            if image.mode != 'RGB':
+                image = image.convert('RGB')
+            
+            # Convert image to base64
+            buffered = BytesIO()
+            image.save(buffered, format="PNG")
+            img_base64 = base64.b64encode(buffered.getvalue()).decode()
+            
+            # Different prompts based on mode
+            if mode == "ocr":
+                prompt = """Extract ALL visible text from this image exactly as it appears. 
+Rules:
+- Only return text that is actually written/printed in the image
+- Preserve reading order (left to right, top to bottom)  
+- Maintain line breaks and formatting
+- Include numbers, symbols, special characters
+- Do NOT add descriptions, interpretations, or context
+- If no text is visible, return 'NO_TEXT_FOUND'
+
+Extracted text:"""
+            
+            else:  # mode == "describe"
+                prompt = """Analyze this image and provide:
+1. All visible text exactly as written
+2. Brief description of the image content and context
+
+Format:
+TEXT: [all visible text here, or NO_TEXT_FOUND if none]
+DESCRIPTION: [brief image description and context]"""
+            
+            # Send request to Ollama
+            response = ollama.generate(
+                model=current_model,
+                prompt=prompt,
+                images=[img_base64],
+                options={
+                    'temperature': 0.1,  # Low temperature for accuracy
+                    'top_p': 0.9,
+                    'num_predict': 1500
+                }
+            )
+            
+            # Extract and clean response
+            extracted_content = response.get('response', '').strip()
+            
+            # Handle no-text case for OCR mode
+            if mode == "ocr" and ('NO_TEXT_FOUND' in extracted_content or len(extracted_content) < 3):
+                return ""
+            
+            return extracted_content
+            
+        except Exception as e:
+            print(f"⚠️ Error extracting from image with Ollama {current_model}: {e}")
+            return ""
+    
+    # Wrapper functions for each mode
+    def xtxt_image_ocr_only(file_buffer):
+        """Traditional OCR: extract only visible text using Ollama"""
+        return xtxt_image_ocr_ollama(file_buffer, mode="ocr")
+    
+    def xtxt_image_describe(file_buffer):
+        """OCR + Description: text + image context using Ollama"""
+        return xtxt_image_ocr_ollama(file_buffer, mode="describe")
+    
+    # Register OCR-only version as default
+    # Note: Will override traditional EasyOCR if both modules are loaded
+    image_formats = [
+        "image/jpeg", "image/jpg", "image/png", 
+        "image/bmp", "image/tiff", "image/webp"
+    ]
+    
+    for format_type in image_formats:
+        register_extractor(format_type, xtxt_image_ocr_only, name="OCR-Ollama")

{pyxtxt-0.2.4 → pyxtxt-0.3/src/pyxtxt.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pyxtxt
-Version: 0.2.4
+Version: 0.3
 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
@@ -78,6 +78,9 @@ Requires-Dist: openai-whisper; extra == "audio"
 Provides-Extra: ocr
 Requires-Dist: easyocr; extra == "ocr"
 Requires-Dist: pillow; extra == "ocr"
+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"
@@ -96,6 +99,7 @@ Requires-Dist: pylatexenc; extra == "all"
 Requires-Dist: openai-whisper; extra == "all"
 Requires-Dist: easyocr; extra == "all"
 Requires-Dist: pillow; extra == "all"
+Requires-Dist: ollama; extra == "all"
 Dynamic: license-file
 
 # PyxTxt
@@ -141,10 +145,13 @@ pip install pyxtxt[pdf,docx,presentation,spreadsheet,html,markdown,epub,email]
 # Audio transcription (~2GB download for Whisper models)
 pip install pyxtxt[audio]
 
-# OCR from images (~1GB download for EasyOCR models)
+# Traditional OCR from images (~1GB download for EasyOCR models)
 pip install pyxtxt[ocr]
 
-# Both audio and OCR
+# AI-powered OCR via Ollama (requires local Ollama + gemma3:4b model)
+pip install pyxtxt[ocr-ollama]
+
+# Both audio and traditional OCR
 pip install pyxtxt[audio,ocr]
 ```
 
@@ -185,6 +192,7 @@ Use python-magic-bin instead of python-magic for easier installation.
 - **LaTeX**: pylatexenc
 - **Audio**: openai-whisper (heavy ~2GB models)
 - **OCR**: easyocr, pillow (heavy ~1GB models)
+- **OCR-Ollama**: ollama, pillow (requires local Ollama server)
 
 Dependencies are automatically installed based on selected optional groups.
 
@@ -249,11 +257,25 @@ text = xtxt(video_response.content)
 ```python
 from pyxtxt import xtxt
 
-# Extract text from images
+# Traditional OCR with EasyOCR (install with: pip install pyxtxt[ocr])
 text = xtxt("scanned_document.png")
 text = xtxt("screenshot.jpg")
 text = xtxt("invoice.tiff")
 
+# AI-powered OCR with Ollama (install with: pip install pyxtxt[ocr-ollama])
+# Requires: ollama server running + gemma3:4b model
+from pyxtxt.estrattori.ocr_ollama import set_ollama_model, xtxt_image_describe
+
+# Configure model (optional, default is gemma3:4b)
+set_ollama_model("gemma3:12b")  # or llava:7b, llava:13b
+
+# Extract only text (OCR mode)
+text = xtxt("complex_document.png")
+
+# Extract text + image description
+description = xtxt_image_describe(open("diagram.png", "rb"))
+# Output: "TEXT: Chart Title: Sales Report 2024\nDESCRIPTION: Bar chart showing quarterly sales data with blue bars"
+
 # From web images
 import requests
 image_response = requests.get("https://example.com/document.png")

{pyxtxt-0.2.4 → pyxtxt-0.3}/src/pyxtxt.egg-info/SOURCES.txt

@@ -21,6 +21,7 @@ src/pyxtxt/estrattori/html.py
 src/pyxtxt/estrattori/md.py
 src/pyxtxt/estrattori/msg.py
 src/pyxtxt/estrattori/ocr.py
+src/pyxtxt/estrattori/ocr_ollama.py
 src/pyxtxt/estrattori/odt.py
 src/pyxtxt/estrattori/pdf.py
 src/pyxtxt/estrattori/pptx.py

{pyxtxt-0.2.4 → pyxtxt-0.3}/src/pyxtxt.egg-info/requires.txt

@@ -23,6 +23,7 @@ pylatexenc
 openai-whisper
 easyocr
 pillow
+ollama
 
 [audio]
 openai-whisper
@@ -55,6 +56,10 @@ beautifulsoup4
 easyocr
 pillow
 
+[ocr-ollama]
+ollama
+pillow
+
 [odf]
 odfpy
 

pyxtxt-0.2.4/src/pyxtxt/__init__.py

@@ -1 +0,0 @@
-from .core import xtxt, extxt_available_formats, xtxt_from_url