Pre-checks
What problem does this solve?
Currently there are four UX issues with conversation management: 1) Historical conversations pile up indefinitely in the sidebar, making it increasingly difficult to find active ones; 2) Model selection is global, forcing users to manually switch back and forth when working across conversations with different purposes; 3) Switching between conversations causes unsent input content to be lost or overwritten, disrupting the workflow; 4) Conversations are not truly isolated — running multiple conversations simultaneously causes conflicts, making it impossible to maintain parallel dialogues.
Proposed solution
-
Auto-archive old conversations
- Add an "Auto-archive" setting with configurable threshold: 7 / 14 / 30 (default) / 60 / 90 days
- Conversations older than the threshold are automatically moved to an "Archived" section
- Archived conversations remain searchable and restorable, but do not clutter the main conversation list
- Support manual archive / un-archive actions
-
Per-conversation model selection
- Each conversation independently remembers its model choice, unaffected by others
- Users can use GPT-4 for coding in Conversation A and Claude for writing in Conversation B without constant switching
-
Independent input fields per conversation
- Each conversation maintains its own input state (unsent text, attachments, etc.)
- Switching away and back should fully preserve the input content
-
Full conversation isolation
- Allow multiple conversations to run simultaneously without interfering with each other
- Currently, running Conversation A and Conversation B in parallel causes conflicts where one conversation becomes unresponsive or messages get mixed
- Each conversation should be fully isolated — context, state, and message queue are independent
Willing to contribute?
Pre-checks
What problem does this solve?
Currently there are four UX issues with conversation management: 1) Historical conversations pile up indefinitely in the sidebar, making it increasingly difficult to find active ones; 2) Model selection is global, forcing users to manually switch back and forth when working across conversations with different purposes; 3) Switching between conversations causes unsent input content to be lost or overwritten, disrupting the workflow; 4) Conversations are not truly isolated — running multiple conversations simultaneously causes conflicts, making it impossible to maintain parallel dialogues.
Proposed solution
Auto-archive old conversations
Per-conversation model selection
Independent input fields per conversation
Full conversation isolation
Willing to contribute?