The "Thinking" / collapsed-thought indicator is rendered inline at the very top of an assistant message. For long responses, the indicator scrolls off the top of the viewport, and the user has to scroll a long way back up to re-open it or to confirm whether thinking has finished.
Steps to reproduce
- Ask the agent a question that produces a long answer (one that requires multiple scroll heights to read in full).
- Scroll down through the response.
- Try to glance at the thinking indicator while reading the body.
Expected behavior
The thinking indicator should remain easily reachable while reading a long response. Suggested approach: place it at the bottom-left of the message bubble (or as a sticky / floating element) so it stays in view regardless of scroll position.
Actual behavior
The indicator is fixed to the top of the message body and scrolls away with it. The user must scroll back up to interact with it.
Screenshots
02-thinking-indicator-top.png — long assistant response where the thinking summary is only visible at the very top
Environment
- App: Finch
1.1.0 (build 174) — Desktop
- OS: macOS
26.4.1 (25E253)
Steps to reproduce
Expected behavior
The thinking indicator should remain easily reachable while reading a long response. Suggested approach: place it at the bottom-left of the message bubble (or as a sticky / floating element) so it stays in view regardless of scroll position.
Actual behavior
The indicator is fixed to the top of the message body and scrolls away with it. The user must scroll back up to interact with it.
Screenshots
02-thinking-indicator-top.png— long assistant response where the thinking summary is only visible at the very topEnvironment
1.1.0 (build 174)— Desktop26.4.1 (25E253)