The Converter Framework in markitdown establishes a robust and extensible architecture for handling diverse document conversion operations. It defines the core interfaces, manages the lifecycle of converters, and provides the necessary context for successful transformations into Markdown.
DocumentConverter Expand
This is the foundational abstract base class that defines the contract for all document converters. It mandates two essential methods: accepts(), which determines if a converter can process a given input stream based on its metadata, and convert(), which performs the actual transformation of the input stream into Markdown, returning a DocumentConverterResult. This abstract interface ensures consistency and extensibility across all concrete converter implementations.
Related Classes/Methods:
DocumentConverter(1:1)DocumentConverter:accepts(1:1)DocumentConverter:convert(1:1)
MarkItDown Expand
Serving as the central orchestrator of the markitdown library, MarkItDown is responsible for discovering, registering, and intelligently selecting the most appropriate DocumentConverter for a given input. It handles various input sources (local files, URIs, HTTP responses, binary streams) and prepares the StreamInfo object that provides context to the converters. It then delegates the actual conversion task to the chosen converter, managing the overall conversion workflow.
Related Classes/Methods:
MarkItDown(1:1)
This Python package acts as the primary repository for all concrete implementations of the DocumentConverter abstract base class. Each module within this directory (e.g., _docx_converter.py, _html_converter.py) encapsulates the specific logic required to convert a particular document type into Markdown. This modular organization facilitates the easy addition of new document type support, promoting extensibility and maintainability.
Related Classes/Methods:
markitdown.converters(1:1)
HtmlConverter Expand
A concrete implementation of DocumentConverter specifically designed to transform HTML content into Markdown. This converter is highly versatile and frequently serves as a crucial intermediate step for other converters (e.g., DocxConverter, EpubConverter) that first convert their input to HTML before generating the final Markdown output. It handles parsing HTML, removing unwanted elements like scripts and styles, and converting the remaining content to Markdown.
Related Classes/Methods:
HtmlConverter(1:1)
DocxConverter Expand
A concrete implementation of DocumentConverter tailored for converting Microsoft Word (.docx) files into Markdown. It leverages external libraries (like mammoth) to convert DOCX to HTML, and then utilizes an internal instance of HtmlConverter to perform the final HTML-to-Markdown transformation. It also incorporates pre-processing steps for specific DOCX elements, such as mathematical equations, ensuring accurate representation in the output.
Related Classes/Methods:
DocxConverter(1:1)
This data structure (defined within the _stream_info module) encapsulates comprehensive metadata about the input stream being processed. This includes vital information such as MIME type, file extension, character set, filename, local path, and URL. It provides a standardized and rich context that the MarkItDown orchestrator passes to DocumentConverter instances, enabling converters to make informed decisions about whether they can accept() a stream and how to convert() it effectively.
Related Classes/Methods:
_StreamInfo(1:1)