You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
Articles split across numbered pages (page 2, page 3, …) are now read to the end, up to five pages, and saved as one clip that says how many pages it covers.
The duplicate check also matches the main address a page names for itself (its canonical URL, on the same site only), so a tracking or share link to an article you've already clipped is caught. When the page says it changed after your clip, Gleaner tells you before asking whether to save an updated copy.
"Grab the part about X" starts from the page heading that matches X, when there is one.
A page's own schema.org type helps pick the clip's Type, such as tutorial for a how-to page.
These use a small read-only script, so they work in the Chrome side panel, the Claude Code desktop browser, Claude in Chrome, and ChatGPT's Work browser when it has a JavaScript tool. ChatGPT's connected tabs work as before.