Bump version.

* feat: issue where inherited function from `markdownify.MarkdownConverter` doesn't have `current_tags` leading to error using `kwargs`, also set default value for `convert_as_inline`

Dockerfile

@@ -17,6 +17,7 @@ RUN pip install markitdown
 # Default USERID and GROUPID
 ARG USERID=10000
 ARG GROUPID=10000
+
 USER $USERID:$GROUPID
 
 ENTRYPOINT [ "markitdown" ]

README.md

@@ -33,12 +33,20 @@ Or use `-o` to specify the output file:
 markitdown path-to-file.pdf -o document.md
 ```
 
+To use Document Intelligence conversion:
+
+```bash
+markitdown path-to-file.pdf -o document.md -d -e "<document_intelligence_endpoint>"
+```
+
 You can also pipe content:
 
 ```bash
 cat path-to-file.pdf | markitdown
 ```
 
+More information about how to set up an Azure Document Intelligence Resource can be found [here](https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/how-to-guides/create-document-intelligence-resource?view=doc-intel-4.0.0)
+
 ### Python API
 
 Basic usage in Python:
@@ -51,6 +59,16 @@ result = md.convert("test.xlsx")
 print(result.text_content)
 ```
 
+Document Intelligence conversion in Python:
+
+```python
+from markitdown import MarkItDown
+
+md = MarkItDown(docintel_endpoint="<document_intelligence_endpoint>")
+result = md.convert("test.pdf")
+print(result.text_content)
+```
+
 To use Large Language Models for image descriptions, provide `llm_client` and `llm_model`:
 
 ```python

pyproject.toml

@@ -42,6 +42,8 @@ dependencies = [
   "pathvalidate",
   "charset-normalizer",
   "openai",
+  "azure-ai-documentintelligence",
+  "azure-identity"
 ]
 
 [project.urls]

src/markitdown/__about__.py

@@ -1,4 +1,4 @@
 # SPDX-FileCopyrightText: 2024-present Adam Fourney <adamfo@microsoft.com>
 #
 # SPDX-License-Identifier: MIT
-__version__ = "0.0.1a3"
+__version__ = "0.0.1a5"

src/markitdown/__main__.py

@@ -3,6 +3,7 @@
 # SPDX-License-Identifier: MIT
 import argparse
 import sys
+import shutil
 from textwrap import dedent
 from .__about__ import __version__
 from ._markitdown import MarkItDown, DocumentConverterResult
@@ -51,22 +52,50 @@ def main():
         help="show the version number and exit",
     )
 
-    parser.add_argument("filename", nargs="?")
     parser.add_argument(
         "-o",
         "--output",
         help="Output file name. If not provided, output is written to stdout.",
     )
+
+    parser.add_argument(
+        "-d",
+        "--use-docintel",
+        action="store_true",
+        help="Use Document Intelligence to extract text instead of offline conversion. Requires a valid Document Intelligence Endpoint.",
+    )
+
+    parser.add_argument(
+        "-e",
+        "--endpoint",
+        type=str,
+        help="Document Intelligence Endpoint. Required if using Document Intelligence.",
+    )
+
+    parser.add_argument("filename", nargs="?")
     args = parser.parse_args()
 
-    if args.filename is None:
+    which_exiftool = shutil.which("exiftool")
-        markitdown = MarkItDown()
+
-        result = markitdown.convert_stream(sys.stdin.buffer)
+    if args.use_docintel:
-        _handle_output(args, result)
+        if args.endpoint is None:
+            raise ValueError(
+                "Document Intelligence Endpoint is required when using Document Intelligence."
+            )
+        elif args.filename is None:
+            raise ValueError("Filename is required when using Document Intelligence.")
+        markitdown = MarkItDown(
+            exiftool_path=which_exiftool, docintel_endpoint=args.endpoint
+        )
+    else:
+        markitdown = MarkItDown(exiftool_path=which_exiftool)
+
+    if args.filename is None:
+        result = markitdown.convert_stream(sys.stdin.buffer)
     else:
-        markitdown = MarkItDown()
         result = markitdown.convert(args.filename)
-        _handle_output(args, result)
+
+    _handle_output(args, result)
 
 
 def _handle_output(args, result: DocumentConverterResult):

src/markitdown/_markitdown.py

@@ -33,6 +33,22 @@ import requests
 from bs4 import BeautifulSoup
 from charset_normalizer import from_path
 
+# Azure imports
+from azure.ai.documentintelligence import DocumentIntelligenceClient
+from azure.ai.documentintelligence.models import (
+    AnalyzeDocumentRequest,
+    AnalyzeResult,
+    DocumentAnalysisFeature,
+)
+from azure.identity import DefaultAzureCredential
+
+# TODO: currently, there is a bug in the document intelligence SDK with importing the "ContentFormat" enum.
+# This constant is a temporary fix until the bug is resolved.
+CONTENT_FORMAT = "markdown"
+
+# Override mimetype for csv to fix issue on windows
+mimetypes.add_type("text/csv", ".csv")
+
 # Optional Transcription support
 IS_AUDIO_TRANSCRIPTION_CAPABLE = False
 try:
@@ -75,7 +91,14 @@ class _CustomMarkdownify(markdownify.MarkdownConverter):
         # Explicitly cast options to the expected type if necessary
         super().__init__(**options)
 
-    def convert_hn(self, n: int, el: Any, text: str, convert_as_inline: bool) -> str:
+    def convert_hn(
+        self,
+        n: int,
+        el: Any,
+        text: str,
+        convert_as_inline: Optional[bool] = False,
+        **kwargs,
+    ) -> str:
         """Same as usual, but be sure to start with a new line"""
         if not convert_as_inline:
             if not re.search(r"^\n", text):
@@ -83,7 +106,9 @@ class _CustomMarkdownify(markdownify.MarkdownConverter):
 
         return super().convert_hn(n, el, text, convert_as_inline)  # type: ignore
 
-    def convert_a(self, el: Any, text: str, convert_as_inline: bool):
+    def convert_a(
+        self, el: Any, text: str, convert_as_inline: Optional[bool] = False, **kwargs
+    ):
         """Same as usual converter, but removes Javascript links and escapes URIs."""
         prefix, suffix, text = markdownify.chomp(text)  # type: ignore
         if not text:
@@ -119,7 +144,9 @@ class _CustomMarkdownify(markdownify.MarkdownConverter):
             else text
         )
 
-    def convert_img(self, el: Any, text: str, convert_as_inline: bool) -> str:
+    def convert_img(
+        self, el: Any, text: str, convert_as_inline: Optional[bool] = False, **kwargs
+    ) -> str:
         """Same as usual converter, but removes data URIs"""
 
         alt = el.attrs.get("alt", None) or ""
@@ -204,7 +231,7 @@ class HtmlConverter(DocumentConverter):
         return result
 
     def _convert(self, html_content: str) -> Union[None, DocumentConverterResult]:
-        """Helper function that converts and HTML string."""
+        """Helper function that converts an HTML string."""
 
         # Parse the string
         soup = BeautifulSoup(html_content, "html.parser")
@@ -223,6 +250,9 @@ class HtmlConverter(DocumentConverter):
 
         assert isinstance(webpage_text, str)
 
+        # remove leading and trailing \n
+        webpage_text = webpage_text.strip()
+
         return DocumentConverterResult(
             title=None if soup.title is None else soup.title.string,
             text_content=webpage_text,
@@ -771,6 +801,35 @@ class PptxConverter(HtmlConverter):
     Converts PPTX files to Markdown. Supports heading, tables and images with alt text.
     """
 
+    def _get_llm_description(
+        self, llm_client, llm_model, image_blob, content_type, prompt=None
+    ):
+        if prompt is None or prompt.strip() == "":
+            prompt = "Write a detailed alt text for this image with less than 50 words."
+
+        image_base64 = base64.b64encode(image_blob).decode("utf-8")
+        data_uri = f"data:{content_type};base64,{image_base64}"
+
+        messages = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "image_url",
+                        "image_url": {
+                            "url": data_uri,
+                        },
+                    },
+                    {"type": "text", "text": prompt},
+                ],
+            }
+        ]
+
+        response = llm_client.chat.completions.create(
+            model=llm_model, messages=messages
+        )
+        return response.choices[0].message.content
+
     def convert(self, local_path, **kwargs) -> Union[None, DocumentConverterResult]:
         # Bail if not a PPTX
         extension = kwargs.get("file_extension", "")
@@ -791,17 +850,38 @@ class PptxConverter(HtmlConverter):
                 # Pictures
                 if self._is_picture(shape):
                     # https://github.com/scanny/python-pptx/pull/512#issuecomment-1713100069
-                    alt_text = ""
+
-                    try:
+                    llm_description = None
-                        alt_text = shape._element._nvXxPr.cNvPr.attrib.get("descr", "")
+                    alt_text = None
-                    except Exception:
+
-                        pass
+                    llm_client = kwargs.get("llm_client")
+                    llm_model = kwargs.get("llm_model")
+                    if llm_client is not None and llm_model is not None:
+                        try:
+                            llm_description = self._get_llm_description(
+                                llm_client,
+                                llm_model,
+                                shape.image.blob,
+                                shape.image.content_type,
+                            )
+                        except Exception:
+                            # Unable to describe with LLM
+                            pass
+
+                    if not llm_description:
+                        try:
+                            alt_text = shape._element._nvXxPr.cNvPr.attrib.get(
+                                "descr", ""
+                            )
+                        except Exception:
+                            # Unable to get alt text
+                            pass
 
                     # A placeholder name
                     filename = re.sub(r"\W", "", shape.name) + ".jpg"
                     md_content += (
                         "\n!["
-                        + (alt_text if alt_text else shape.name)
+                        + (llm_description or alt_text or shape.name)
                         + "]("
                         + filename
                         + ")\n"
@@ -1318,6 +1398,74 @@ class ZipConverter(DocumentConverter):
             )
 
 
+class DocumentIntelligenceConverter(DocumentConverter):
+    """Specialized DocumentConverter that uses Document Intelligence to extract text from documents."""
+
+    def __init__(
+        self,
+        endpoint: str,
+        api_version: str = "2024-07-31-preview",
+    ):
+        self.endpoint = endpoint
+        self.api_version = api_version
+        self.doc_intel_client = DocumentIntelligenceClient(
+            endpoint=self.endpoint,
+            api_version=self.api_version,
+            credential=DefaultAzureCredential(),
+        )
+
+    def convert(
+        self, local_path: str, **kwargs: Any
+    ) -> Union[None, DocumentConverterResult]:
+        # Bail if extension is not supported by Document Intelligence
+        extension = kwargs.get("file_extension", "")
+        docintel_extensions = [
+            ".pdf",
+            ".docx",
+            ".xlsx",
+            ".pptx",
+            ".html",
+            ".jpeg",
+            ".jpg",
+            ".png",
+            ".bmp",
+            ".tiff",
+            ".heif",
+        ]
+        if extension.lower() not in docintel_extensions:
+            return None
+
+        # Get the bytestring for the local path
+        with open(local_path, "rb") as f:
+            file_bytes = f.read()
+
+        # Certain document analysis features are not availiable for filetypes (.xlsx, .pptx, .html)
+        if extension.lower() in [".xlsx", ".pptx", ".html"]:
+            analysis_features = []
+        else:
+            analysis_features = [
+                DocumentAnalysisFeature.FORMULAS,  # enable formula extraction
+                DocumentAnalysisFeature.OCR_HIGH_RESOLUTION,  # enable high resolution OCR
+                DocumentAnalysisFeature.STYLE_FONT,  # enable font style extraction
+            ]
+
+        # Extract the text using Azure Document Intelligence
+        poller = self.doc_intel_client.begin_analyze_document(
+            model_id="prebuilt-layout",
+            body=AnalyzeDocumentRequest(bytes_source=file_bytes),
+            features=analysis_features,
+            output_content_format=CONTENT_FORMAT,  # TODO: replace with "ContentFormat.MARKDOWN" when the bug is fixed
+        )
+        result: AnalyzeResult = poller.result()
+
+        # remove comments from the markdown content generated by Doc Intelligence and append to markdown string
+        markdown_text = re.sub(r"<!--.*?-->", "", result.content, flags=re.DOTALL)
+        return DocumentConverterResult(
+            title=None,
+            text_content=markdown_text,
+        )
+
+
 class FileConversionException(BaseException):
     pass
 
@@ -1337,6 +1485,7 @@ class MarkItDown:
         llm_model: Optional[str] = None,
         style_map: Optional[str] = None,
         exiftool_path: Optional[str] = None,
+        docintel_endpoint: Optional[str] = None,
         # Deprecated
         mlm_client: Optional[Any] = None,
         mlm_model: Optional[str] = None,
@@ -1406,6 +1555,12 @@ class MarkItDown:
         self.register_page_converter(ZipConverter())
         self.register_page_converter(OutlookMsgConverter())
 
+        # Register Document Intelligence converter at the top of the stack if endpoint is provided
+        if docintel_endpoint is not None:
+            self.register_page_converter(
+                DocumentIntelligenceConverter(endpoint=docintel_endpoint)
+            )
+
     def convert(
         self, source: Union[str, requests.Response, Path], **kwargs: Any
     ) -> DocumentConverterResult:  # TODO: deal with kwargs
@@ -1608,6 +1763,8 @@ class MarkItDown:
         ext = ext.strip()
         if ext == "":
             return
+        if ext in extensions:
+            return
         # if ext not in extensions:
         extensions.append(ext)