📝 docs: Anthropic card shows 7-day tokens and prompt caching Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
📝 docs: Anthropic spend shows $0 on the first UTC day of the month
📝 docs: a saved provider URL is not called with the server key
📝 docs: describe the Balance & usage Settings card Cover which providers show the card, the admin-key env switch, and that Codex and SuperGrok meters stay separate.
📝 docs: say a filtered search lists at most 30 topics
📝 docs: describe search filters for assistant and date range
📝 docs: a pasted topic reference stays for later replies in that topic
📝 docs: describe account topic search and pasted topic references
📝 docs: point Chinese xelatex at a font whose text can be copied
📝 docs: describe sandbox exportFile batches, subdirectory outputs, and image :2
📝 docs: explain Claude earlier-reasoning drops and readable background errors Fable 5.1 and the 5.5 models drop mismatched earlier thinking instead of failing; background-turn errors now show the provider message.
📝 docs: describe retired Claude cards and Haiku 5.5 long-prompt cost Assistants on Opus 5, Sonnet 5, Opus 4.8/4.7/4.6, and Sonnet 4.6 keep plugins, images, window, gear options, and cost; Haiku 5.5 cost switches to the higher rates over 100K prompt tokens.
📝 docs: list Claude 5.5 Anthropic cards Document Anthropic's current lineup, the always-on thinking and default effort per model, hidden legacy ids, and the Anthropic Compatible retarget to Sonnet 5.5 / Opus 5.5.
📝 docs: describe phone back from a provider page
📝 docs: include public storage links in the iOS save steps
📝 docs: explain saving sandbox files on iPhone and iPad
📝 docs: explain resume request-failed cards A background read can retry without a stack of cards. A failed save still shows one card.
📝 docs: point Sandbox setup at OpenSandbox only The Dify sidecar is no longer a backend. Operator pages now use the Sandbox tool and OPENSANDBOX_* settings.
📝 docs: say a policy change replaces the sandbox and old files stay put
📝 docs: publish the OpenSandbox per-conversation setup page The Dify page still told operators to expect no sandbox after each turn. Reuse, the expiry volume, and the create-timeout footgun now live on their own page.
📝 docs: add the OpenSandbox Compose setup used beside ChatHub
📝 docs: count sandbox basename order by content changes
📝 docs: say the last same-name sandbox write is the returned file
📝 docs: describe real absolute paths and capped sandbox downloads
📝 docs: document the OpenSandbox Code Interpreter backend
📝 docs: keep foreign citations and code out of PDF link rewrites
📝 docs: describe downloadable Code Interpreter PDF links Co-authored-by: Cursor <cursoragent@cursor.com>
📝 docs: describe the longer tool-round budget and closing reply Co-authored-by: Cursor <cursoragent@cursor.com>
📝 docs: explain the outdated Grok CLI version error The message quotes ChatHub's pinned client header. Updating ChatHub is the fix; grok update on the host is not.
📝 docs: skip calibration when skill text is unavailable A send still goes out when skill instructions were not already available; that reply is not used for the learned correction.