All conversations show "Oops, an error occurred" after Codex Desktop update

Resolved 💬 9 comments Opened May 16, 2026 by Redjo Closed May 16, 2026
💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

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

26.513.3673.0

What subscription do you have?

chatGPT pro

What platform is your computer?

"$([Environment]::OSVersion | ForEach-Object VersionString) $((if ([Environment]::Is64BitOperatingSystem) { "x64" } else { "x86" }))"

What issue are you seeing?

After updating Codex Desktop on Windows, all existing conversations fail to open and show:

"Oops, an error occurred"

The app UI suggests checking for updates or retrying, but retrying does not fix it.

A feedback report was uploaded from the Codex app with this correlation ID:

019e30cd-2df5-7532-a313-d6f26cd79992

Versions:

  • Codex Desktop: 26.513.3673.0
  • codex-cli: 0.130.0-alpha.5
  • Platform: Windows x64

Relevant local logs:

  • Received turn/started for unknown conversation
  • thread/goal/get failed: goals feature is disabled

I tried restarting Codex and reinstalling the app, but the issue persists.

What steps can reproduce the bug?

Feedback ID: 019e30cd-2df5-7532-a313-d6f26cd79992

Maybe on chat with supabase extension connected

<img width="877" height="515" alt="Image" src="https://github.com/user-attachments/assets/68f74134-3195-4212-b2ca-de3523336fcf" />

What is the expected behavior?

_No response_

Additional information

_No response_

View original on GitHub ↗

9 Comments

github-actions[bot] contributor · 2 months ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

  • #22981
  • #22971
  • #22984
  • #23020
  • #23028

Powered by Codex Action

jatmn · 2 months ago

Yup same issue here.
Codex Desktop (windows 11) 26.513.31313

mbf2188-rgb · 2 months ago

I’m seeing what appears to be the same Codex Desktop Windows session-resume issue after updating to 26.513.3673.0.

Environment:

  • Windows x64
  • Codex Desktop: 26.513.3673.0
  • Issue started immediately after the Codex Desktop update/restart

Symptoms:

  • Recent Codex chats fail to open with “Oops, an error has occurred.”
  • Some older chats still load.
  • The affected chats may appear in search/sidebar, but opening them fails.
  • Killing Codex processes and reopening does not reliably fix it.
  • Moving/resetting .codex-global-state.json changed the sidebar/project list but did not fix the underlying thread resume issue.

More specific error:
Failed to resume chat
cannot resume running thread <REDACTED_THREAD_ID> with stale path:
requested \\?\C:\Users\<REDACTED_USER>\.codex\sessions\<DATE_PATH>\rollout-<REDACTED>.jsonl
active C:\Users\<REDACTED_USER>\.codex\sessions\<DATE_PATH>\rollout-<REDACTED>.jsonl

I also saw the reverse form earlier:
requested C:\Users\<REDACTED_USER>\.codex\sessions\<DATE_PATH>\rollout-<REDACTED>.jsonl
active \\?\C:\Users\<REDACTED_USER>\.codex\sessions\<DATE_PATH>\rollout-<REDACTED>.jsonl

This looks like a Windows path normalization issue where Codex treats \\?\C:\... and C:\... as different active session paths even though they refer to the same file.

Local files where the affected thread ID appeared:

  • .codex-global-state.json
  • .codex-global-state.json.bak
  • session_index.jsonl
  • state_5.sqlite
  • state_5.sqlite-wal
  • logs_2.sqlite
  • logs_2.sqlite-wal

I did not delete .codex\sessions. The repo itself appears fine; git commits and working tree are intact. The failure appears isolated to Codex Desktop session/thread resume state.

hinaoing · 2 months ago

Same issue on Windows 11.

Codex
26.513.31313

<img width="431" height="279" alt="Image" src="https://github.com/user-attachments/assets/e03aab65-e204-4668-a810-6fa3cc1b3ee2" />

luisnucg · 2 months ago

Same here on Windows 11

Environment

  • OS: Windows 11
  • Codex App package: OpenAI.Codex_26.513.3673.0_x64__2p2nqsd0c76g0
  • App release in logs: codex@26.513.31313
  • App server version: 0.131.0-alpha.9
  • Electron: 42.0.1

Issue
After installing the latest Codex App version, opening some previous threads shows the generic error screen:

Oops, an error has occurred

The thread does not render.

Root Cause Observed
The renderer crashes while parsing historical saved assistant messages that contain Codex git directives with Windows-style cwd paths, for example:

::git-create-branch{cwd="C:\Users\<user>\Documents\<repo>" branch="codex/example-branch"}
::git-stage{cwd="C:\Users\<user>\Documents\<repo>"}
::git-commit{cwd="C:\Users\<user>\Documents\<repo>"}

Representative log error with personal/repo details redacted:

error boundary ... name=LocalConversationPage
errorMessage="invalid syntax at line 1 col 5:

1  cwd=\"C:\\Users\\<user>\\Documents\\<repo>\" branch=\"codex/example-branch\"
       ^"

Component stack starts around:

at Br (app://-/assets/markdown-D32gobzN.js:8:5425)
at ... local-conversation-thread-*.js
at ... local-conversation-page-*.js

Steps To Reproduce

  1. On Windows, open Codex App 26.513.3673.0.
  2. Open a previous local thread whose saved final answer includes ::git-create-branch, ::git-stage, or ::git-commit directives with a Windows cwd path containing backslashes.
  3. The thread fails to render and shows the generic “Oops” error screen.

Expected
Threads should render even if historical assistant messages contain app directives. Invalid or legacy directives should be ignored, escaped, or rendered as text.

Actual
The markdown/directive parser throws, the React error boundary is hit, and the entire thread page fails to open.

Local Workaround
Escaping the historical directives in the saved JSONL thread files prevents the crash:

\::git-create-branch{...}
\::git-stage{...}
\::git-commit{...}

This confirms the trigger is directive parsing, not repo state or project code.

Notes
This appears to be a regression in the new renderer/parser, because these were previously saved by Codex itself as valid final-response metadata.

<img width="308" height="223" alt="Image" src="https://github.com/user-attachments/assets/4fe30b57-d730-4112-9785-5d4c4978d43c" />

jppolk · 2 months ago

<img width="553" height="115" alt="Image" src="https://github.com/user-attachments/assets/9f094536-50dd-434b-a960-fd51d3c747f9" />

thank you @luisnucg, this fixed my .jsonl file so codex could read it again! it worked for now so i can continue to work thanks to you. untill the real fix comes out. thanks a lot man!

<update: it doesn't completely work, just codex is able to read it but can't actually compact it>

luisnucg · 2 months ago

@jppolk Glad it helped, still it is just a bandaid, the bug will continue showing up until an update that fixes the root cause is shipped!

jppolk · 2 months ago

Still can't work with the modified .jsonl I get this error:

Context automatically compacted
Error running remote compact task: { "error": { "message": "Your input exceeds the context window of this model. Please adjust your input and try again.", "type": "invalid_request_error", "param": "input", "code": "context_length_exceeded" } }

etraut-openai contributor · 2 months ago

We've issued a hotfix to address this. Please update to the latest version.