Issue
When switching chat models (e.g., from GPT-5.4 to a different version) in the Stept chat interface, the context pill at the top keeps displaying the old chat entry instead of reflecting the new context. This can be confusing for users, especially when the context should change along with the model.
Additionally, the displayed session times are inaccurate: two sessions currently show as "2h ago" but were actually started just minutes ago.
Questions/Considerations
- What should the behavior ideally be when switching models? Should a new session be started, or should the model continue within the same session?
- What are best practices here for context and session continuity from a UX perspective?
Steps to Reproduce
- Start a chat session in Stept.
- Switch model (e.g., between available GPT versions).
- Observe that the context pill does not update, and session times are inaccurate.
Expected Behavior
- The context pill updates to reflect the current session/model context.
- Session start times display accurately in the chat list.
Actual Behavior
- Context pill shows old chat entry after model switch.
- Session time information is outdated/incorrect (shows 2h when chat started minutes ago).
Additional Notes
- Improvements to session management may require UX design decisions.
- See attached screenshot for reference.
Issue
When switching chat models (e.g., from GPT-5.4 to a different version) in the Stept chat interface, the context pill at the top keeps displaying the old chat entry instead of reflecting the new context. This can be confusing for users, especially when the context should change along with the model.
Additionally, the displayed session times are inaccurate: two sessions currently show as "2h ago" but were actually started just minutes ago.
Questions/Considerations
Steps to Reproduce
Expected Behavior
Actual Behavior
Additional Notes