Skip to main content

PdfToMarkdownPipeline

What it is​

  • A pipeline that converts PDF (application/pdf) files into Markdown text.
  • Implements conversion by writing the file bytes to a temporary file and using pymupdf4llm.to_markdown(...).

Public API​

  • PdfToMarkdownPipelineConfiguration

    • Extends ConvertToMarkdownBasePipelineConfiguration.
    • mime_type: str — defaults to "application/pdf".

  • PdfToMarkdownPipelineParameters

    • Extends ConvertToMarkdownBasePipelineParameters.
    • processor_iri: str — defaults to "http://ontology.naas.ai/abi/document/PDFToMarkdownProcessor".

  • PdfToMarkdownPipeline

    • Extends ConvertToMarkdownBasePipeline.
    • __init__(configuration: PdfToMarkdownPipelineConfiguration)
      • Initializes the base pipeline and retrieves the ABIModule singleton via ABIModule.get_instance().
    • convert_to_markdown(file: File) -> str
      • Reads file content (file.read()), writes it to a temporary file, converts it to Markdown via pymupdf4llm.to_markdown(temp_path), deletes the temp file, and returns the Markdown string.

Configuration/Dependencies​

  • Depends on:
    • pymupdf4llm for PDF-to-Markdown conversion.
    • naas_abi_marketplace.domains.document.ABIModule (singleton access).
    • File from the document ontology (file.read() must return bytes).
    • Base pipeline types:
      • ConvertToMarkdownBasePipeline
      • ConvertToMarkdownBasePipelineConfiguration
      • ConvertToMarkdownBasePipelineParameters
    • pydantic.Field / typing.Annotated for parameter metadata.

Usage​

from naas_abi_marketplace.domains.document.pipelines.PdfToMarkdownPipeline import (
    PdfToMarkdownPipeline,
    PdfToMarkdownPipelineConfiguration,
)

# 'file' must be an instance of the ontology File class with a .read() -> bytes method.
pipeline = PdfToMarkdownPipeline(PdfToMarkdownPipelineConfiguration())
markdown_text = pipeline.convert_to_markdown(file)
print(markdown_text)

Caveats​

  • The conversion writes the PDF bytes to a temporary file on disk (tempfile.NamedTemporaryFile(delete=False)).
  • Temporary file cleanup is performed after conversion (os.unlink); if conversion raises an exception before cleanup, the temporary file may remain on disk.