How big can a document be?

Uploads up to 50 MB in one request, or up to 500 MB through a prepared upload. Up to 2,000 pages per document and 50 documents per batch.

What happens to the documents I upload?

They're kept for 24 hours so you can parse other pages or ask another question without sending them again, then deleted. You can delete a file yourself at any time.

How much does a document cost?

1 credit per page, scanned pages included, plus AI tokens when you use a prompt or schema. For formats without pages, every 3,000 characters of output counts as a page. Documents that fail cost nothing.

Reports and documents

PDFs and reports into fields you can query

Find documents with search, check what a parse will cost, then turn PDF, Word, Excel and PowerPoint files into Markdown or JSON. Scanned pages are read too.

You send a document

slicedinvoices.com/pdf/wordpress-pdf-invoice-plugin-sample.pdf

Prompt: Extract the invoice fields

invoice_numberinvoice_datedue_datetotalcurrencyline_items

Parse

You get

1 page · 2 credits

INV-3337

Issued January 25, 2016, due January 31, 2016

total 93.5currency $

Web Design

Line item

amount 85

Plus the whole invoice as Markdown

Find, price, parse, check

Four steps from a PDF to a checked record

Real numbers from real runs: a PDF search, a free estimate, an invoice parsed into fields, and one line of code that catches the odd one out.

Find

3 sustainability reports with a PDF search

1 credit

Price

15 pages in the file, 2 to parse, 2 credits

free estimate

Parse

Invoice INV-3337 into six fields

2 credits

Check

Line items 85, total 93.5: flagged for review

your code

Trust, then verify

Extraction reports what the document says. A one-line check catches the documents worth a second look; in the real invoice run, the total was higher than the only line item because tax was added.

check.py

items = sum(i["amount"] for i in doc.json["line_items"])   # 85
if abs(items - doc.json["total"]) > 0.01:                      # total 93.5
    flag(doc, "total differs from the line items")             # tax, here

How it works

Find it, price it, then read it

Search for documents on the web, see the cost before you pay, and parse links or uploads into your fields.

find_pdfs.py

1import os
2from spidra import Spidra
3 
4client = Spidra(api_key=os.environ["SPIDRA_API_KEY"])
5 
6pdfs = client.search("sustainability report 2025", sources=["web"], filetype="pdf", limit=3)
7 
8for r in pdfs.data.web:
9    print(r.title, "|", r.url)

PDF search works best on topical queries. A search that finds nothing costs nothing.

Plan it

What it costs, and what to know first

Typical jobs, roughly

1,000 one-page invoices into fieldsabout 2,000 credits
100 fifty-page reports as Markdown5,000 credits
Estimates before any of itfree

1 credit per page, scanned pages included, plus AI tokens when you ask for fields. Documents that fail cost nothing.

Know before you build

Dates come back as written
January 25, 2016 stays that way. Convert to ISO in your code.
Scans take longer
Pages without a text layer are read by AI, so they're slower than text pages.
Files are deleted after 24 hours
Re-parse within a day by fileId, or delete a file sooner.
Formats without pages
Word, spreadsheets, slides and the rest count each 3,000 characters of output as a page.

Typing it in

Accurate on a good day, slow, and costly at hundreds of documents a week.

A plain OCR tool

Gets you text, but not the fields, and leaves the parsing to you.

Spidra

Markdown and your fields from one request, a free estimate first, failed documents free.

What you get

Documents as Markdown and JSON

Every parse gives you the document as Markdown. Add a prompt or schema and you also get your fields.

The invoice, in the schema above

invoice_numberstringAs printed on the document
invoice_datestringAs written, like January 25, 2016
due_datestring | nullnull when the document has none
totalnumberA number you can add up
currencystring | nullThe symbol or code shown
line_itemsarrayOne object per line, with description and amount

Dates come back as the document writes them. Convert them in your code if you need ISO dates.

Where it works

Invoices and receipts

Numbers, dates, totals and line items into the same fields every time.

Reports and filings

Annual, sustainability and research reports, found with PDF search.

Spreadsheets and slides

Excel, CSV and PowerPoint files, as Markdown tables and text.

Scanned pages

Pages without a text layer are read by AI, at the same 1 credit per page.

FAQ

Questions, answered

Reports and documents

PDF, Word (DOCX and DOC), Excel (XLSX), PowerPoint (PPTX and PPT), OpenDocument files, RTF, EPUB, CSV and HTML. The format is detected from the file's content, not its extension.