markitdown-pro 1.3.6__tar.gz → 1.3.7__tar.gz

{markitdown_pro-1.3.6 → markitdown_pro-1.3.7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: markitdown-pro
-Version: 1.3.6
+Version: 1.3.7
 Summary: A package that converts almost any file format to Markdown.
 Author: Developer
 Classifier: Programming Language :: Python :: 3

{markitdown_pro-1.3.6 → markitdown_pro-1.3.7}/markitdown_pro/converters/gpt_vision_wrapper.py

@@ -63,15 +63,27 @@ class GPTVisionWrapper(ConverterWrapper):
         model_name: str = "gpt-4.1-mini",
         api_version: str = "2024-09-01-preview",
         completion_tokens: int = 32000,
+        request_timeout_s: float = 20.0,
+        acquire_timeout_s: float = 20.0,
+        page_timeout_s: float = 20.0,
     ):
         super().__init__(model_name)
         self.gpt_vision = GPTVision(
             model_name=model_name,
             api_version=api_version,
             completion_tokens=completion_tokens,
+            request_timeout_s=request_timeout_s,
+            acquire_timeout_s=acquire_timeout_s,
+            page_timeout_s=page_timeout_s,
         )
 
-    async def convert(self, file_path: str) -> Optional[str]:
+    async def convert(
+        self,
+        file_path: str,
+        max_retries: int = 6,
+        base_delay: float = 0.5,
+        max_delay: float = 20.0,
+    ) -> Optional[str]:
         """
         Convert a file to Markdown using GPT-Vision.
 
@@ -95,7 +107,12 @@ class GPTVisionWrapper(ConverterWrapper):
             # Use the concurrent OCR path for PDFs (page-by-page, parallelized).
             return await self.gpt_vision.process_scanned_pdf_concurrent(file_path)
         # Otherwise, assume it's an image and run single-image OCR/analysis.
-        return await self.gpt_vision.process_image(file_path)
+        return await self.gpt_vision.process_image(
+            file_or_url=file_path,
+            max_retries=max_retries,
+            base_delay=base_delay,
+            max_delay=max_delay,
+        )
 
     async def aclose(self) -> None:
         await self.gpt_vision.aclose()

markitdown_pro-1.3.7/markitdown_pro/handlers/image_handler.py

@@ -0,0 +1,84 @@
+import contextlib
+from pathlib import Path
+from typing import List, Optional
+
+from ..common.logger import logger
+from ..common.utils import detect_extension
+from ..converters.gpt_vision_wrapper import GPTVisionWrapper
+from .base_handler import BaseHandler
+
+
+class ImageHandler(BaseHandler):
+    SUPPORTED_EXTENSIONS = GPTVisionWrapper.SUPPORTED_EXTENSIONS
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.gpt_vision_4_1_mini = GPTVisionWrapper(model_name="gpt-4.1-mini")
+        self.gpt_vision_4_1 = GPTVisionWrapper(model_name="gpt-4.1")
+
+        self.pipeline: List[GPTVisionWrapper] = [self.gpt_vision_4_1_mini, self.gpt_vision_4_1]
+
+        self._max_retries = kwargs.get("max_retries", 2)
+        self._base_delay = kwargs.get("base_delay", 0.5)
+        self._max_delay = kwargs.get("max_delay", 10.0)
+
+    async def handle(self, file_path, *args, **kwargs) -> Optional[str]:
+        """
+        Handles image files by converting them to Markdown using GPT Vision.
+
+        Args:
+            file_path: Path to the image file.
+
+        Returns:
+            Markdown string representing the image content (OCR and analysis),
+            or an error message.
+        """
+        for handler in self.pipeline:
+            try:
+                logger.info(f"ImageHandler: Trying handler {handler.name} for file '{file_path}'")
+                md_content = await handler.convert(
+                    file_path,
+                    max_retries=self._max_retries,
+                    base_delay=self._base_delay,
+                    max_delay=self._max_delay,
+                )
+                if md_content:
+                    return md_content
+            except Exception as e:
+                logger.error(
+                    f"ImageHandler: Handler {handler.name} failed for file '{file_path}': {e}"
+                )
+            finally:
+                # Prevent "Event loop is closed" from httpx finalizers during pytest teardown
+                with contextlib.suppress(Exception):
+                    if hasattr(handler, "gpt_vision"):
+                        await handler.gpt_vision.aclose()
+
+        logger.error(f"ImageHandler: All handlers failed for file '{file_path}'")
+        return None
+
+    async def get_page_count(self, file_path: str | Path) -> Optional[int]:
+        """
+        Determines the number of pages in the given image file.
+        """
+        p = Path(file_path)
+        ext = detect_extension(str(p.absolute()))
+        if ext not in self.SUPPORTED_EXTENSIONS:
+            logger.warning(f"ImageHandler: Unsupported file format: {ext}")
+            return None
+
+        if ext in {".tif", ".tiff"}:
+            try:
+                from PIL import Image, ImageSequence  # optional dependency
+
+                with Image.open(str(p)) as im:
+                    return sum(1 for _ in ImageSequence.Iterator(im)) or 1
+            except Exception as e_tiff:
+                logger.warning(f"ImageHandler: local TIFF page count failed for '{p}': {e_tiff}")
+        else:
+            return 1
+
+    async def aclose(self) -> None:
+        for handler in self.pipeline:
+            if hasattr(handler, "gpt_vision") and handler.gpt_vision:
+                await handler.gpt_vision.aclose()

{markitdown_pro-1.3.6 → markitdown_pro-1.3.7}/markitdown_pro/services/openai_services.py

@@ -71,9 +71,9 @@ class GPTVision:
         completion_tokens: int,
         max_concurrency: int = 150,
         *,
-        request_timeout_s: float = 60.0,
-        acquire_timeout_s: float = 60.0,
-        page_timeout_s: float = 120.0,
+        request_timeout_s: float = 20.0,
+        acquire_timeout_s: float = 20.0,
+        page_timeout_s: float = 20.0,
     ):
         """
         Parameters
@@ -326,7 +326,14 @@ class GPTVision:
     # Public OCR APIs
     # -------------------------
 
-    async def process_image(self, file_or_url: str, prompt: Optional[str] = None) -> Optional[str]:
+    async def process_image(
+        self,
+        file_or_url: str,
+        prompt: Optional[str] = None,
+        max_retries: int = 6,
+        base_delay: float = 0.5,
+        max_delay: float = 20.0,
+    ) -> Optional[str]:
         """
         OCR a single image (local path or URL) with the vision model.
 
@@ -363,7 +370,11 @@ class GPTVision:
             try:
                 # The LLM call itself has its own hard timeout & retry
                 response = await self._ainvoke_with_retry(
-                    file_or_url, [system_message, human_message]
+                    file_or_url,
+                    messages=[system_message, human_message],
+                    max_retries=max_retries,
+                    base_delay=base_delay,
+                    max_delay=max_delay,
                 )
             finally:
                 with suppress(Exception):
@@ -378,7 +389,7 @@ class GPTVision:
 
         except asyncio.TimeoutError:
             logger.error(f"GPTVision: process_image: Timed out for {file_or_url}")
-            return "[Timed out]"
+            return None
         except Exception as e:
             logger.error(
                 f"GPTVision: process_image: Error during GPT-vision image OCR for {file_or_url}: {e}"
@@ -466,7 +477,7 @@ class GPTVision:
                                 logger.error(
                                     f"GPTVission: ocr_page: {file_stem} - page {page_index} timed out after {self.page_timeout_s}s"
                                 )
-                                partial_md = "[Timed out]"
+                                partial_md = None
 
                             if partial_md:
                                 logger.info(

{markitdown_pro-1.3.6 → markitdown_pro-1.3.7}/markitdown_pro.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: markitdown-pro
-Version: 1.3.6
+Version: 1.3.7
 Summary: A package that converts almost any file format to Markdown.
 Author: Developer
 Classifier: Programming Language :: Python :: 3

{markitdown_pro-1.3.6 → markitdown_pro-1.3.7}/setup.py

@@ -2,7 +2,7 @@ from setuptools import find_packages, setup
 
 setup(
     name="markitdown-pro",
-    version="1.3.6",
+    version="1.3.7",
     author="Developer",
     description="A package that converts almost any file format to Markdown.",
     long_description=open("README.md").read(),

markitdown_pro-1.3.6/markitdown_pro/handlers/image_handler.py

@@ -1,65 +0,0 @@
-import contextlib
-from pathlib import Path
-from typing import Optional
-
-from ..common.logger import logger
-from ..common.utils import detect_extension
-from ..converters.gpt_vision_wrapper import GPTVisionWrapper
-from .base_handler import BaseHandler
-
-
-class ImageHandler(BaseHandler):
-    SUPPORTED_EXTENSIONS = GPTVisionWrapper.SUPPORTED_EXTENSIONS
-
-    def __init__(self, *args, **kwargs):
-        super().__init__(*args, **kwargs)
-        self.gpt_vision = GPTVisionWrapper()
-
-    async def handle(self, file_path, *args, **kwargs) -> Optional[str]:
-        """
-        Handles image files by converting them to Markdown using GPT Vision.
-
-        Args:
-            file_path: Path to the image file.
-
-        Returns:
-            Markdown string representing the image content (OCR and analysis),
-            or an error message.
-        """
-        logger.info(f"Processing image file: {file_path}")
-        try:
-            md_content = await self.gpt_vision.convert(file_path)
-            return md_content
-        except Exception as e:
-            logger.error(f"ImageHandler: Error handling image file '{file_path}': {e}")
-            return None
-        finally:
-            # Prevent "Event loop is closed" from httpx finalizers during pytest teardown
-            with contextlib.suppress(Exception):
-                if hasattr(self.gpt_vision, "gpt_vision"):
-                    await self.gpt_vision.gpt_vision.aclose()
-
-    async def get_page_count(self, file_path: str | Path) -> Optional[int]:
-        """
-        Determines the number of pages in the given image file.
-        """
-        p = Path(file_path)
-        ext = detect_extension(str(p.absolute()))
-        if ext not in self.SUPPORTED_EXTENSIONS:
-            logger.warning(f"ImageHandler: Unsupported file format: {ext}")
-            return None
-
-        if ext in {".tif", ".tiff"}:
-            try:
-                from PIL import Image, ImageSequence  # optional dependency
-
-                with Image.open(str(p)) as im:
-                    return sum(1 for _ in ImageSequence.Iterator(im)) or 1
-            except Exception as e_tiff:
-                logger.warning(f"ImageHandler: local TIFF page count failed for '{p}': {e_tiff}")
-        else:
-            return 1
-
-    async def aclose(self) -> None:
-        if hasattr(self, "gpt_vision") and self.gpt_vision:
-            await self.gpt_vision.aclose()