library/pdf-inspector

• wasi:filesystem Filesystem read-only

  ** (ro)

act:tools interface — expand a tool for its input schema.

• to_markdown Convert a PDF to Markdown, preserving headings, lists and tables. Returns the detected PDF type alongside the Markdown — check `pages_needing_ocr` and `has_encoding_issues` to know whether the text is trustworthy.

  Convert a PDF to Markdown, preserving headings, lists and tables. Returns the detected PDF type alongside the Markdown — check `pages_needing_ocr` and `has_encoding_issues` to know whether the text is trustworthy.

  Parameter	Type	Description
  data	object | null	Inline PDF bytes, as a CBOR byte string — or the canonical `{"$bytes": "<base64>"}` envelope over JSON transports.
  path	string | null	Path to a PDF file on the host. Requires a `wasi:filesystem` read grant covering this path.
  pages	array | null	1-indexed page numbers to extract. Omit to extract every page.
  profile	variant	`fidelity` (default) stays close to the source; `compact` emits fewer tokens.
  password	string | null	Password for an encrypted PDF.
  include_images	boolean | null	Include image placeholders in the Markdown.
  include_page_markers	boolean | null	Insert `<!-- Page N -->` markers between pages.

  Returns content.

• detect Detect whether a PDF is text-based or scanned, with page count, title and layout complexity. Does not extract text, so it is much cheaper than to_markdown.

  Detect whether a PDF is text-based or scanned, with page count, title and layout complexity. Does not extract text, so it is much cheaper than to_markdown.

  Parameter	Type	Description
  data	object | null	Inline PDF bytes, as a CBOR byte string — or the canonical `{"$bytes": "<base64>"}` envelope over JSON transports.
  path	string | null	Path to a PDF file on the host. Requires a `wasi:filesystem` read grant covering this path.
  password	string | null	Password for an encrypted PDF.

  Returns content.

• classify Cheaply classify a PDF (typically 10-50ms): type, page count, which pages need OCR, and confidence. Call this before to_markdown to avoid spending tokens on a scanned document.

  Cheaply classify a PDF (typically 10-50ms): type, page count, which pages need OCR, and confidence. Call this before to_markdown to avoid spending tokens on a scanned document.

  Parameter	Type	Description
  data	object | null	Inline PDF bytes, as a CBOR byte string — or the canonical `{"$bytes": "<base64>"}` envelope over JSON transports.
  path	string | null	Path to a PDF file on the host. Requires a `wasi:filesystem` read grant covering this path.

  Returns content.

• extract_text Extract plain text from a PDF with no Markdown formatting, one line per detected text line.

  Extract plain text from a PDF with no Markdown formatting, one line per detected text line.

  Parameter	Type	Description
  data	object | null	Inline PDF bytes, as a CBOR byte string — or the canonical `{"$bytes": "<base64>"}` envelope over JSON transports.
  path	string | null	Path to a PDF file on the host. Requires a `wasi:filesystem` read grant covering this path.

  Returns content.