Documents & archives

Document readers expose useful text and structure without forcing binary containers into model context.

Documents

read_document extracts supported PDF, office, and notebook content. Use selectors where available to keep output bounded.

Archives

read_archive lists or reads a named member such as release.zip:manifest.json. write_archive updates a member under write approval and revision guards.

Artifacts

Large extracted content spills to a session artifact. Follow the returned artifact:// reference with narrow reads instead of reloading the entire document.