What version of the Codex App are you using (From “About Codex” dialog)?

I updated Codex Desktop to the latest version available to me on Windows before filing this report, but the issue still reproduces. Inside rollout/session metadata on this machine I also see cli_version: 0.115.0-alpha.27.

What subscription do you have?

ChatGPT Plus / Pro-equivalent individual account

What platform is your computer?

Windows host with WSL workspace

What issue are you seeing?

Codex Desktop automations create automation runs in local state, but the corresponding rollout/session file never materializes. The UI then fails with:

Failed to resume task
no rollout found for thread id <thread-id>

This is not limited to one automation. I have reproduced it on at least two different automations with different prompts and different working directories:

• rechka-next-watcher
• codex-catalog-sync

This makes it look like an app-level automation lifecycle/state bug rather than a prompt-specific issue.

What steps can reproduce the bug?

1. Configure a desktop automation that runs in execution_environment = "worktree".
2. Make sure the target prompt file exists and the worktree path is valid.
3. Trigger the automation from Codex Desktop.
4. Observe that the automation run is registered locally.
5. The UI then attempts to resume the task/thread.
6. Resume fails with:

Failed to resume task
no rollout found for thread id <thread-id>

What is the expected behavior?

When an automation starts:

• a valid rollout/session file should be created;
• the automation run thread id should map to a real rollout/session artifact;
• the run should either complete normally or fail with a prompt/runtime error;
• the UI should not try to resume a nonexistent rollout.

Actual behavior

The automation run is recorded in local automation state, but the rollout/session file is missing.

What I verified locally:

• New rows appear in the local automation DB under automation_runs.
• Example observed run/thread ids:
  • 019d66a2-30e5-7241-b457-c9f45b8acc28
  • 019d66d4-03c3-7041-9ed7-9491815da5ad
• The runs can remain IN_PROGRESS or become archived, but there is still no matching rollout file.
• The local inbox_items table remains empty for these runs.
• The corresponding rollout/session files do not appear under the local .codex/sessions/YYYY/MM/DD/ tree.

So the failure appears to happen between:

• automation run creation,
• thread id assignment,
• rollout/session materialization,
• UI resume.

Additional information

Things I already tried:

• full app restart;
• recreating the automation config;
• recreating the dedicated runtime worktree;
• cleaning stale session index entries;
• reinstalling/re-updating Codex Desktop;
• verifying that the prompt file and worktree path are valid.

None of these fixed the issue.

This looks related in symptom to other “no rollout found” / orphaned-thread reports, but my repro is specifically about Desktop automations creating automation_runs without a corresponding rollout/session artifact.

Potentially related issues:

• app-server websocket shared-thread turns complete but rollout never materializes, so sibling thread/resume fails with "no rollout found" #16872
• Codex desktop is showing stale/orphaned thread entries that cannot be reopened or archived. #14162
• Codex Desktop App — “failed to resume task” on all sessions after service incident #14396

If useful, I can also provide sanitized details from the local automation state store showing that automation_runs are written while no matching rollout files are created.