Parse options

This page focuses on the request parameters for the parser-style document parsing endpoint. The examples below keep the file upload flow only and use the current public parameter names.

Request parameters

Parameter	Location	Type	Required	Default	Description
file	form	file	Yes	—	Input document file
image_type	query	string	No	url	How images are embedded in Markdown: url or base64
content_filter	query	string	No	all	Keep only selected content block types
options_json	form	JSON string	No	Built-in defaults	Parser configuration merged with the server defaults

image_type

image_type controls how image content is represented in the Markdown result:

Value	Meaning
url	Embed image content as accessible URLs
base64	Embed image content inline as Base64

Use url for most frontend and knowledge-base integrations. Choose base64 when you need a fully self-contained Markdown artifact.

content_filter

content_filter narrows the result to selected block types. Common patterns:

Value	Meaning
all	Return all content blocks
text	Keep only text-related content
table	Keep only table-related content
image	Keep only image-related content

If your workflow only needs one category, filtering at request time is usually simpler than post-filtering in downstream code.

options_json

options_json is a JSON string that controls parsing behaviour. Typical options include:

• generating a document tree / catalog
• merging related table fragments
• re-levelling title hierarchy
• ignoring headers, footers, footnotes, and similar auxiliary content

Example:

{
  "applyDocumentTree": true,
  "mergeTables": true,
  "relevelTitles": true,
  "ignore_labels": [
    "number",
    "footnote",
    "header",
    "header_image",
    "footer",
    "footer_image",
    "aside_text"
  ]
}

ignore_labels

ignore_labels is typically passed inside options_json to suppress auxiliary block types in the parse output. The supported labels are:

• number
• footnote
• header
• header_image
• footer
• footer_image
• aside_text

To keep all supported auxiliary content, pass an empty array explicitly:

--form 'options_json={"ignore_labels":[]}'

Recommendations

• For real-time previews, prefer image_type=url to keep payloads smaller.
• For search, extraction, or RAG workflows, use content_filter=text or content_filter=table to reduce downstream processing.
• For layout-heavy documents, combine document-tree and table-merge options, then inspect the output with Response overview.