Codex App loses thread history while local JSONL sessions remain intact
Nobody has claimed this yet.
- Dominant language
- Rust
- Stars
- 125k
- Forks
- 19.5k
- PR merge metrics
- PR metrics pending
Description
What version of the Codex App are you using (From “About Codex” dialog)?
Exact About-dialog version unavailable; bundled codex-cli: 0.155.0-alpha.9.2
What subscription do you have?
ChatGPT Plus
What platform is your computer?
Microsoft Windows NT 10.0.26200.0 x64
What issue are you seeing?
Multiple Codex App project threads no longer display their conversation history. The threads still exist, are not archived or deleted, and retain the correct project IDs, but thread/read returns turns: [].
The corresponding JSONL rollout files still exist under %USERPROFILE%.codex\sessions and contain substantial data (from ~321 KB to ~499 MB). This indicates that the conversation data remains locally available but is no longer attached to or indexed by the app.
This affects ten project threads, including TourYours, JoonDog SEO, VTC preparation, AfterTourTips, AI Agent learning, art history/VAE, legal/tax planning, and home training. The issue prevents ongoing work because the prior context is invisible in the desktop app.
What steps can reproduce the bug?
- Open the Codex desktop app on Windows.
- Open an affected project, for example the TourYours task “پروژه سایت جدید گردشگری با AI”.
- The task exists but its conversation history is not displayed.
- Query the same task using thread/read; it returns zero turns.
- Inspect %USERPROFILE%.codex\sessions. A JSONL file with the same thread ID exists and contains the original history.
Primary affected thread ID: 019fb833-3da5-75a3-a4d8-aefc1acc6bdc
Primary local rollout: rollout-2026-07-31T14-43-13-019fb833-3da5-75a3-a4d8-aefc1acc6bdc.jsonl
File size: 29,032,903 bytes; 3,460 JSONL lines.
In-app feedback/session ID: 019fb791-00f1-7ca2-bf8d-2b0d31ecef42
Moving the thread between sidebar sections, navigating directly to it, and checking archived threads did not restore the history. No affected session files were deleted or modified.
What is the expected behavior?
The Codex App should load and display all turns stored in the existing local JSONL rollout for each thread. Existing thread-to-project associations should remain intact. If the local index is damaged, the app should safely rebuild it from CODEX_HOME/sessions without deleting or duplicating threads.
Additional information
Ten affected thread IDs and session sizes:
- 019fb833-21c7-7891-8d7f-1043ea50fd43 — 191,482,600 bytes
- 019fd31e-2395-7ea0-ade5-a0f8d436ac80 — 498,793,765 bytes
- 019fb859-02b3-7983-819d-ad28c487cca4 — 2,882,642 bytes
- 01a04c68-28d4-7082-b72b-e88aa58b4db2 — 2,983,781 bytes
- 019fb833-3da5-75a3-a4d8-aefc1acc6bdc — 29,032,903 bytes
- 01a07176-7f85-7ca1-81aa-dfc1882ecd30 — 1,770,534 bytes
- 01a05c29-421a-74e0-9f83-6e124c47d6e7 — 493,267 bytes
- 019fffd0-90ea-7fb1-afb9-af9850afdbba — 7,370,880 bytes
- 019fb859-0ab8-7811-85d8-94f5eb53b4aa — 320,855 bytes
- 019fb833-2fd2-7721-8890-84cfa6609ca3 — 5,522,609 bytes
No JSONL files are attached publicly because they contain private conversation content. They can be provided privately to OpenAI engineers if required.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by reproducing thread/read for the primary thread ID and compare its empty turns with the matching rollout under %USERPROFILE%.codex\sessions. Read the local session and indexing path around CODEX_HOME/sessions and thread/read. Done means existing JSONL histories load into their correct project threads without deletion, duplication, or broken associations.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- rust
- Domain
- desktop-dev
- Issue type
- Bug
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100