Codex Desktop 26.707.71524 loses all Workspace Tasks after upgrade while local sessions remain intact

Open 💬 1 comment Opened Jul 14, 2026 by wangyunchen8868

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

26.707.71524

What subscription do you have?

ChatGPT Plus

What platform is your computer?

Microsoft Windows NT 10.0.19045.0 x64

What issue are you seeing?

Summary

After upgrading to ChatGPT Desktop / Codex App 26.707.71524 on Windows, all Codex tasks disappeared from my Local Workspace.

The local workspace still exists and can be opened normally, but the task list shows:

No tasks

However, the local Codex session data still exists on disk.

This appears to be a Workspace-to-Session indexing issue rather than data loss.

---

Environment

  • OS: Windows 11 x64
  • Codex App: 26.707.71524
  • Subscription: ChatGPT Plus

---

Reproduction

  1. Open an existing Local Workspace.
  2. Upgrade ChatGPT Desktop to 26.707.71524.
  3. Open the same Local Workspace again.

Expected:

  • Previous Codex tasks should still be visible.

Actual:

  • Workspace opens normally.
  • Local Folder is still attached.
  • Task list displays "No tasks".

---

Investigation

I manually inspected the local Codex data.

The following files still exist:

C:\Users\<USER>\.codex\sessions\
C:\Users\<USER>\.codex\session_index.jsonl
C:\Users\<USER>\.codex\logs_2.sqlite

### What steps can reproduce the bug?

This does not appear to be actual data loss.

It appears the Desktop application no longer associates existing sessions with the Local Workspace after the upgrade.

If there is a way to rebuild the Workspace index from existing rollout sessions, that would recover all previous tasks.

I can provide additional logs if needed.

### What is the expected behavior?

_No response_

### Additional information

I have verified that both `session_index.jsonl` and the rollout session files are intact.

This looks like a regression in Workspace session indexing rather than actual session loss.

View original on GitHub ↗

This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗