[Bug] Extension shortcuts never register in daemon-backed interactive chats (shift+tab effort-cycle) #1516
stanleytejakusuma
started this conversation in
Bug reports
Replies: 1 comment 1 reply
|
I think that would be a great feature. |
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Summary
Extension-registered shortcuts never wire up in daemon-backed interactive chats. The motivating case: the effort-cycle extension binds
shift+tabto cycle the thinking level, but in a normal daemon-attached session/hotkeysshows no shortcut and the binding does nothing. The workaround is the/effortcommand.Environment
Symptoms
~/.prime/agent/extensions/that callspi.registerShortcut(...)insession_start. The effort-cycle extension is a real example: it registersshift+tabwith a handler that cycles the thinking level.shift+tab. Nothing happens./hotkeysdoes not list the shortcut.The same extension works in a non-daemon interactive session, so the extension code itself is valid.
Suspected trigger
The daemon path constructs
InteractiveModewithbindLocalSessionExtensions: falsehardcoded inmain.js, so the client never binds the discovered extension shortcuts.dist/modes/daemon/daemon-extension-bindingimplements the client-local UI callbacks as no-ops. The daemon worker owns the active extension runner, but the shortcut registration needs the live TUI client, which is exactly the gap described in discussion #1462.Expected behavior
registerShortcutwork in daemon-backed interactive chats, the same way they do in local sessions./hotkeysreflects the actual active bindings so users can discover what is wired.Related reports
keybindings.jsonmechanism exists for app-level bindings)Suggested fix direction
Tie this into the #1462 fix: when the client binds the local extension runner, register extension shortcuts against the live TUI keybindings manager and emit
/hotkeysentries from the same source.All reactions