2026-10-25: API 1.39

The read_pdf step returns the tables of a PDF as data, billed by the pages it reads.

#Added

• dsl: The read_pdf step reads the tables of a PDF the job is on or has downloaded: cells as verbatim strings, a table that continues over pages as one table, with max_pages, include_text and into: "vars". See documents and PDFs.
• pricing: A read_pdf costs one credit for every started 50 pages it read, counted as compute and shown as PDF reading in the job's cost lines. POST /v1/jobs/estimate includes it, and a replay of a stored run reads the PDF again at no charge.
• api: The error pdf_source_missing (422, not retryable): a read_pdf step had no PDF to read.
• api: The warning pdf_pages_truncated: a reading stopped at its max_pages; count and of say how many pages were read of how many.
• dsl: Lint warning W010: a read_pdf with no earlier goto or download that could produce a PDF.
• api: cache.max_age applies to a program that goes to a PDF and reads it with read_pdf.