Legacy Codex local JSONL session import into cloud-synced Codex Desktop

Codex
, ,
1

Hello OpenAI/Codex Support,

I am requesting a supported migration/import path for legacy Codex local sessions into the new cloud-synced Codex Desktop app.

My current situation:

  • I have 269 legacy Codex sessions stored locally as .jsonl files.

  • Some individual session files are over 180MB.

  • I upgraded to the new Codex Desktop app, package com.codexorbit.app, version family v26.422+.

  • The new app appears to use cloud-synced thread history for the Desktop sidebar, with local SQLite caching in orbit.db, rather than directly listing legacy local .jsonl sessions.

Desired feature:

Please provide a supported way to import local legacy Codex .jsonl sessions so they appear as native, sidebar-visible Codex Desktop threads under my account.

Why current workarounds are insufficient:

  • SQLite patching is not viable. Injecting records into orbit.db does not create valid account-owned cloud thread IDs, and the app’s sync layer appears to purge or reject records that do not exist on the backend.

  • Drag-and-drop file attachment is useful for referencing old context, but it does not preserve 269 individual threads, titles, timestamps, project grouping, tool history, or sidebar continuity.

  • External local viewers can make the archive searchable, but they do not solve migration into the native Codex Desktop workflow.

A useful migration feature could be one of:

  • A Codex Desktop “Import legacy sessions” command.

  • A CLI migration tool that uploads local .jsonl sessions into my account as Codex cloud threads.

  • A documented admin/support-assisted import process.

  • A local-only legacy sidebar mode that indexes old .jsonl sessions without conflicting with cloud sync.

I am happy to provide sanitized schema samples, file-size distribution, and logs if helpful. The key requirement is a supported import path that does not require unsupported SQLite edits or private backend API manipulation.

Thank you.

2

Welcome to the forum!

You can discuss Codex-related issues here, but the official place to report them is the OpenAI Codex GitHub Issues.

I used ChatGPT to search for similar reports and found several related issues. However, without first-hand experience of your specific case, I won’t point to a particular one.

It’s worth reviewing existing issues to see if any align with your situation. If none do, consider opening a new report with clear reproduction steps. If you do find a match, adding a :+1: reaction can help increase its visibility to the developers.