Dispatch: add a "Log and Refresh Chat" button to checkpoint context before it gets foggy
Feature request
In long Dispatch conversations, the context can get "foggy" over time — as context compaction kicks in, early detail is lost and some user requests can silently drop. In a real power-user session (thousands of orchestrated turns spread across multiple days), this shows up as the assistant gradually losing track of earlier decisions and occasionally missing an instruction that scrolled out of the compacted window.
Request
Add a "Log and Refresh Chat" (a.k.a. "Log and Restart Chat") button in Dispatch that:
- Writes a durable log/summary of the current conversation state (decisions made, open threads, outstanding requests, key context) to persistent storage.
- Starts a fresh context that picks up from that log — so the session continues cleanly without dragging the entire foggy history along.
Why
Long orchestrated sessions are exactly where compaction hurts most, and they're also the sessions you least want to lose the thread on. A user-triggered checkpoint-and-restart gives an explicit, reliable way to reset the working context while preserving intent — rather than hoping automatic compaction keeps the right details.
Suggested behavior
- One-click button in the Dispatch UI.
- Produces a durable, reviewable log/summary (ideally something the user can read and edit).
- New session boots seeded from that log, continuing the work.
Thanks for considering this!
This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗