You'll noticeTier: how much it should matter to you
2Useful: my rating, 1 to 5
2Signal: worth watching, 1 to 5
PDF ReadingArea: what it touches
ImprovementsKind: in v2.1.323,
ImprovementsSection of the release
Unclear It is unclear which tool returns these results, so it is not known where a reader would see them.
What
Results that Claude Code returns from documents and PDFs can now carry more:
Search results can say whether an image is attached and which page it is from. Page images for the hits are sent to the model as images after the main result.
A PDF page can carry its text layer, the text stored in the page. This only happens when no page range was asked for and the PDF was too large to read whole.
Pages taken from a PDF now also return a "Text layer of page N" block when the page has text.
Why
PDFs that are too large to read whole may now come back as text rather than failing.
How sure we are
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubtIt is unclear which tool returns these results, so it is not known where a reader would see them.
