Supported formats

docgate detects formats from a path extension, or accepts an explicit format argument. PDF paths and PDF bytes are supported. Other formats currently need a file path; non-PDF byte streams are reserved for future stream adapters.

Format	Logical pages	Extraction path	Extra install
PDF	PDF pages	Embedded text, then provider OCR/vision for pages without usable text	None
DOCX	One document	Paragraphs and table cells	docgate[office]
XLSX	One worksheet	Cell values	docgate[office]
PPTX	One slide	Text-bearing shapes	docgate[office]
TXT, Markdown	One file	Native text	None
HTML, HTM	One file	Basic tag-stripped text	None

plan = docgate.plan("notes.txt", format="txt")
result = docgate.execute(plan)

For non-PDF formats, all extraction is local. need and selection still affect which logical pages are returned, but docgate does not render or send Office, text, or HTML documents to providers.

PDF rendering safety

When a PDF page has no usable text layer, docgate renders it as PNG. Rendering is bounded by max_pixels (8,000,000 by default). If a requested image must be reduced, the receipt includes a Clamp; inspect receipt.clamps if image resolution is important to your use case.

Encrypted PDFs cannot be planned without a password and are not decrypted by docgate.