Recreate deleted worktrees from their branch on restore

Archiving a Paseo-owned worktree deletes its directory, so reopening its
agent previously dead-ended even after the workspace record came back —
there was no folder to return to.

Restore now recreates the worktree at its original path from the kept
branch (the branch is never deleted on archive), re-points the same
workspace record, and clears archivedAt — preserving the workspaceId so
the agent stays linked. The directory is recreated before the un-archive
so the reconciler can't re-archive it for a missing directory. A deleted
or already-checked-out branch surfaces a typed failure and leaves the
workspace archived.

Gated behind a worktreeRestore capability flag; older hosts get an
explicit "update your host" state instead of a misleading failure.
This commit is contained in:
Mohamed Boudra
2026-06-18 20:44:57 +07:00
parent 31fcde2ea6
commit b3f81c0bf4
17 changed files with 670 additions and 24 deletions
@@ -7,8 +7,10 @@ import { navigateToPreparedWorkspaceTab } from "@/utils/workspace-navigation";
export type { NavigateToAgentInput } from "./resolve";
// Clears the transient restoring state if the daemon resolves refreshAgent without
// re-emitting a workspace_update (the directory-gone case), so the gate never spins forever.
const RESTORE_TIMEOUT_MS = 7000;
// re-emitting a workspace_update (the directory-gone case), so the gate never spins
// forever. Recreating a worktree can require a git fetch, so the budget is generous
// to avoid flashing a false "failed" on a capable daemon doing slow real work.
const RESTORE_TIMEOUT_MS = 30000;
function restoreArchivedWorkspace(serverId: string, agentId: string, workspaceId: string): void {
const snapshot = getHostRuntimeStore().getSnapshot(serverId);
@@ -33,6 +35,15 @@ function restoreArchivedWorkspace(serverId: string, agentId: string, workspaceId
return;
}
// COMPAT(worktreeRestore): added in v0.1.98, drop the gate when floor >= v0.1.98
// Single capability read for restore. An old daemon recreates nothing on
// refresh_agent, so a gone directory would spin then flash a misleading
// "couldn't restore". Surface an explicit "update your host" state instead.
if (session?.serverInfo?.features?.worktreeRestore !== true) {
store.setWorkspaceRestoreStatus(serverId, workspaceId, "needs-host-upgrade");
return;
}
store.setWorkspaceRestoreStatus(serverId, workspaceId, "restoring");
// The reducer guards "failed" so a late timeout after the descriptor lands is a no-op.
setTimeout(