Add system prompt, glossary, presets for Ollama/WebLLM, image translation support

translators/pptx_translator.py

@@ -9,6 +9,8 @@ from pptx.shapes.group import GroupShape
 from pptx.util import Inches, Pt
 from pptx.enum.shapes import MSO_SHAPE_TYPE
 from services.translation_service import translation_service
+import tempfile
+import os
 
 
 class PowerPointTranslator:
@@ -32,21 +34,23 @@ class PowerPointTranslator:
         presentation = Presentation(input_path)
         
         # Translate each slide
-        for slide in presentation.slides:
-            self._translate_slide(slide, target_language)
+        for slide_idx, slide in enumerate(presentation.slides):
+            self._translate_slide(slide, target_language, slide_idx + 1, input_path)
         
         # Save the translated presentation
         presentation.save(output_path)
         
         return output_path
     
-    def _translate_slide(self, slide, target_language: str):
+    def _translate_slide(self, slide, target_language: str, slide_num: int, input_path: Path):
         """
         Translate all text elements in a slide while preserving layout
         
         Args:
             slide: Slide to translate
             target_language: Target language code
+            slide_num: Slide number for reference
+            input_path: Path to source file for image extraction
         """
         # Translate notes (speaker notes)
         if slide.has_notes_slide:
@@ -56,15 +60,16 @@ class PowerPointTranslator:
         
         # Translate shapes in the slide
         for shape in slide.shapes:
-            self._translate_shape(shape, target_language)
+            self._translate_shape(shape, target_language, slide)
     
-    def _translate_shape(self, shape: BaseShape, target_language: str):
+    def _translate_shape(self, shape: BaseShape, target_language: str, slide=None):
         """
         Translate text in a shape based on its type
         
         Args:
             shape: Shape to translate
             target_language: Target language code
+            slide: Parent slide for adding image translations
         """
         # Handle text-containing shapes
         if shape.has_text_frame:
@@ -74,20 +79,72 @@ class PowerPointTranslator:
         if shape.shape_type == MSO_SHAPE_TYPE.TABLE:
             self._translate_table(shape.table, target_language)
         
+        # Handle pictures/images
+        if shape.shape_type == MSO_SHAPE_TYPE.PICTURE:
+            self._translate_image_shape(shape, target_language, slide)
+        
         # Handle group shapes (shapes within shapes)
         if shape.shape_type == MSO_SHAPE_TYPE.GROUP:
             for sub_shape in shape.shapes:
-                self._translate_shape(sub_shape, target_language)
+                self._translate_shape(sub_shape, target_language, slide)
         
         # Handle smart art (contains multiple shapes)
         # Smart art is complex, but we can try to translate text within it
         if hasattr(shape, 'shapes'):
             try:
                 for sub_shape in shape.shapes:
-                    self._translate_shape(sub_shape, target_language)
+                    self._translate_shape(sub_shape, target_language, slide)
             except:
                 pass  # Some shapes may not support iteration
     
+    def _translate_image_shape(self, shape, target_language: str, slide):
+        """
+        Translate text in an image using vision model and add as text box
+        """
+        if not getattr(self.translation_service, 'translate_images', False):
+            return
+        
+        from services.translation_service import OllamaTranslationProvider
+        
+        if not isinstance(self.translation_service.provider, OllamaTranslationProvider):
+            return
+        
+        try:
+            # Get image blob
+            image_blob = shape.image.blob
+            ext = shape.image.ext
+            
+            # Save to temp file
+            with tempfile.NamedTemporaryFile(suffix=f'.{ext}', delete=False) as tmp:
+                tmp.write(image_blob)
+                tmp_path = tmp.name
+            
+            # Translate with vision
+            translated_text = self.translation_service.provider.translate_image(tmp_path, target_language)
+            
+            # Clean up
+            os.unlink(tmp_path)
+            
+            if translated_text and translated_text.strip():
+                # Add text box below the image with translation
+                left = shape.left
+                top = shape.top + shape.height + Inches(0.1)
+                width = shape.width
+                height = Inches(0.5)
+                
+                # Add text box
+                textbox = slide.shapes.add_textbox(left, top, width, height)
+                tf = textbox.text_frame
+                p = tf.paragraphs[0]
+                p.text = f"[{translated_text}]"
+                p.font.size = Pt(10)
+                p.font.italic = True
+                
+                print(f"Added image translation: {translated_text[:50]}...")
+                
+        except Exception as e:
+            print(f"Error translating image: {e}")
+    
     def _translate_text_frame(self, text_frame, target_language: str):
         """
         Translate text within a text frame while preserving formatting