Documents by URL - Spidra

Scrape, Batch scrape and Search’s page content all accept document links. Spidra detects the file type, reads it, and returns Markdown, tables included. Supported: PDF, DOC and DOCX, XLSX, PPT and PPTX, ODT, RTF, EPUB and CSV.

{ "urls": [{ "url": "https://arxiv.org/pdf/1706.03762" }] }

result.content (shortened):

#### Provided proper attribution is provided, Google hereby grants permission to

reproduce the tables and figures in this paper solely for use in journalistic or scholarly works.

## Attention Is All You Need

##### Ashish Vaswani
...

The whole 15-page paper came back in about 17 seconds for 1 credit.

Scanned pages with no text layer are read with AI, so you still get text.
A prompt or schema works the same as on a web page: the AI reads the document’s text. Long documents use more AI credits.
Raw documents cost 1 credit per URL, however many pages they have.

Scrape or Parse?

Use Parse when documents are the main job:

You want	Use
A document that sits among web pages you’re scraping	Scrape
To upload files that aren’t on the web	Parse
Page ranges, or each page on its own	Parse
The cost before you start	Parse (`dryRun`)
Many documents, or files over 50 MB	Parse