Feature request: Custom conversation titles + checkpoint rollback / branching in Codex Extension chat

What version of the VS Code extension are you using?

0.4.66

What subscription do you have?

plus

Which IDE are you using?

cursor

What platform is your computer?

wsl

What issue are you seeing?

Hi team — thank you for building the Codex IDE extension. I’m using it heavily for multi-step coding tasks and would like to request two UX/workflow improvements that would significantly increase productivity when managing many sessions.

1) Custom conversation titles (rename chat sessions)

Current behavior
In the extension’s conversation list, a session is effectively labeled by the first message content (or an auto-derived snippet). There’s currently no way to rename a conversation.

Why this matters
When working across multiple tasks/projects (or multiple experiments within the same repo), it becomes hard to quickly locate the right session, and the list gets cluttered with similar-looking snippets.

Proposed solution

• Add “Rename” for conversations (e.g., right-click/context menu or an inline edit icon).
• Persist the custom title across restarts.
• Optional: allow auto-suggested titles but always editable by the user.
• Optional: enable search/filter by title.

Acceptance criteria

• I can rename an existing conversation at any time.
• The renamed title appears in the session list and remains after reload/restart.
• Renaming does not affect conversation content/history.

---

2) “Regenerate from here” + rollback checkpoints / branching for chat-driven code changes

Current behavior
The conversation is linear. If a response or code change goes in the wrong direction, the only option is to continue prompting forward and manually undo changes (or rely on external Git operations). There’s no built-in way to regenerate from a previous node, rollback the session state, or create a new branch of the conversation.

Why this matters
For iterative coding, it’s common to explore multiple approaches. Having “checkpoint rollback” and “branching” (similar to Cursor agent workflows or ChatGPT’s regenerate/backtrack patterns) makes experimentation safer and faster, especially when the assistant edits multiple files.

Proposed solution (UI + behavior)

• On any assistant message (or a “step”), provide actions like:

• Regenerate from here (keep prior context up to this node, discard later turns)
• Rollback to this point (conversation timeline rollback)
• Create branch from here (fork the conversation so I can explore an alternative path)
• Clearly indicate branches (e.g., a branch selector or a small tree/timeline UI).
• Optional (high value): integrate with Git to make rollback safer:

• create a git branch automatically when branching, or
• stash/apply changes per step, or
• show file diffs per step.

Acceptance criteria

• I can choose a previous turn and regenerate the assistant output from that point.
• I can rollback the conversation to an earlier node (with later turns removed or archived).
• I can fork/branch a conversation from a chosen node and switch between branches.
• The UI makes it clear which branch I’m on and what the active timeline is.

What steps can reproduce the bug?

null

What is the expected behavior?

_No response_

Additional information

_No response_