The composer
Typing, sending, attachments, the mic, the model picker, workspace chip, editing and the keyboard.
The composer is the input card at the bottom of every chat. It is one flat card (no outline) with the text field, a + button, the model picker, the mic and the send button (a bold up-arrow).
Parts of the composer#
| Control | What it does |
|---|---|
| + (Add to task) | Opens the Add to task sheet: attach, tools, skills. See next page. |
| Text field | Type your message. Supports @ mentions and / commands (see Mentions and commands). |
| Model picker | Shows the current model. Tap to change model or provider. |
| Orbit button | Shows connected Orbits and lets you add one to the task. |
| Mic | Speech-to-text into the field using your selected speech provider. Tap again to Stop. |
| Send / Stop | Sends; while CX is answering it becomes Stop ("Cancel response?"). |
| Workspace chip | When a workspace is attached: Terminal, Files, Change workspace, Disconnect workspace. |
Attachments#
Attached items appear as chips above the field; tap a chip to preview, or Remove attachment. Supported: photos (camera or gallery, optional crop unless Skip Image Editing is on), videos, audio, PDFs, documents, code and other files. Unsupported types show "Unsupported file type".
Long text paste: with Paste Long Text as File on (Settings → Chat and display → Chat), pasting text longer than the threshold turns it into a file attachment instead of a giant message.
Editing a sent message#
Long-press your own message → Edit. The composer switches to Editing mode; send to create a new branch, or Cancel Edit. Use the arrows on a message to move between branches.
Keyboard#
- Swipe down on the chat to hide the keyboard. On Android 11+ with Animations on, the keyboard follows your finger and settles by position and speed; otherwise it simply hides.
- Volume keys can scroll pages if Volume Key Page Scroll is enabled (handy on e-ink phones).
Voice from the drawer#
The mic in the drawer opens a full-screen voice mode: speak, and what you said opens as a new task. Microphone permission is required.
When no model is available#
If no provider is enabled you see "No enabled model. Enable a provider and select a model first." Open Settings → AI models → Providers.