Floot turns, MCP bridges, and secret requests do not survive daemon restart #30
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Applying an Endo release restarts the daemon. Existing Floot CLI sessions preserve their workspace and backend conversation history, but the live turn above them is only in memory.
Observed in session mt9qcahn-oldcn9:
This is distinct from #23, which covers Claude Code background-task completion records, and from #22/#25, which add secret ingestion but keep the request broker in memory.
Architecture finding
Claude and Codex already have the same lifecycle shape: a formula-backed client, durable workspace plus backend conversation state, lazy sandbox recreation, queued one-process-per-turn execution, an MCP socket mount, and nearly duplicated mount-recreate, abort, stream, and status machinery. Their real differences are backend adapters: argv construction, event translation, and resume identity (Claude transcript UUID versus Codex thread ID). Floot session-turn status, the MCP socket server, and secret-request broker remain ephemeral closures.
Proposed factoring
Acceptance criteria