Fix: Proper WMF/EMF image handling in PPTXReader

llama-index-integrations/readers/llama-index-readers-file/llama_index/readers/file/slides/base.py

@@ -56,9 +56,56 @@ def __init__(self) -> None:
             "tokenizer": tokenizer,
         }
 
+    def convert_wmf_to_png(self, input_path: str) -> str:
+        """Convert WMF/EMF to PNG using LibreOffice, handling both Windows and Linux/macOS."""
+        import sys
+        import shutil
+        import subprocess
+        from pathlib import Path
+
+        file_path = Path(input_path)
+        output_path = file_path.with_suffix(".png")
+
+        if sys.platform == "win32":
+            libreoffice_path = r"C:\Program Files\LibreOffice\program\soffice.exe"
+            command = [
+                libreoffice_path,
+                "--headless",
+                "--convert-to",
+                "png",
+                str(file_path),
+            ]
+        else:
+            libreoffice_path = shutil.which("libreoffice") or shutil.which("soffice")
+            command = [
+                "libreoffice",
+                "--headless",
+                "--convert-to",
+                "png",
+                "--outdir",
+                str(file_path.parent),
+                str(file_path),
+            ]
+
+        if not libreoffice_path or not Path(libreoffice_path).exists():
+            raise RuntimeError(
+                f"LibreOffice is not installed. Cannot convert {file_path}."
+            )
+
+        subprocess.run(
+            command, check=True, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL
+        )
+
+        if not output_path.exists():
+            raise RuntimeError(
+                f"Conversion failed: {file_path} → {output_path} does not exist."
+            )
+
+        return str(output_path)
+
     def caption_image(self, tmp_image_file: str) -> str:
         """Generate text caption of image."""
-        from PIL import Image
+        from PIL import Image, UnidentifiedImageError
 
         model = self.parser_config["model"]
         feature_extractor = self.parser_config["feature_extractor"]
@@ -71,7 +118,20 @@ def caption_image(self, tmp_image_file: str) -> str:
         num_beams = 4
         gen_kwargs = {"max_length": max_length, "num_beams": num_beams}
 
-        i_image = Image.open(tmp_image_file)
+        try:
+            i_image = Image.open(tmp_image_file)
+            image_format = i_image.format
+        except UnidentifiedImageError:
+            return "Error opening image file."
+
+        if image_format in ["WMF", "EMF"]:
+            try:
+                converted_path = self.convert_wmf_to_png(tmp_image_file)
+                i_image = Image.open(converted_path)
+            except Exception as e:
+                print(f"Error converting WMF/EMF image: {e}")
+                return f"Error converting WMF/EMF image"
+
         if i_image.mode != "RGB":
             i_image = i_image.convert(mode="RGB")