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
Context Usage: the chat input shows how much of the model's context window the latest response used, with input/output tokens and how many messages the next Model Context includes. Workers AI context windows are read from the model catalog.
Retry: the latest response in a conversation, and the latest image in Image History, can be retried; the new response replaces the old one once it succeeds.
Reasoning shows how long the model thought for.
Changed
AI Elements: messages, bubbles, scrolling, reasoning and tool calls (Trace), attachments, code blocks and spinners now use shadcn AI Elements. Code blocks are highlighted with Shiki and can be copied.
Assistant responses have copy and Retry actions.
The model selector and context indicator use the primary colour; the chat input no longer looks disabled while empty.
Dependencies: added @shadcn/react and shiki; removed @streamdown/code.
Fixed
Workers AI responses were cut off at 256 tokens; requests now ask for up to 8,192 output tokens (at most a quarter of the context window).
The selected model and search toggle no longer revert after navigating between pages.