openai / openai/codex

[Bug] Mid-Conversation switch references latest selected agent instead of the last agent used for code generation

Open
#44,397 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

app bug
Dominant language
Rust
Stars
125k
Forks
19.4k
PR merge metrics
PR metrics pending

Description

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

26.903.61454

What subscription do you have?

Plus

What platform is your computer?

Windows

What issue are you seeing?

The "degraded performance" warning dialog tracks model state against the latest selected LLM in the picker rather than the last LLM that actually executed a generation. When switching from LLM A (which just generated code) to LLM B, the warning dialog appears. When the user immediately switches back to LLM A to avoid the issue, the dialog triggers again warning that switching back will cause degraded performance—even though LLM A is the model that just generated the previous response.

Impact

  • Misleading UI warnings that confuse the user.
  • Forces unnecessary warning prompts when attempting to undo an accidental or unexecuted model selection.
What steps can reproduce the bug?
  1. Generate an output using LLM A.
  2. Select LLM B in the model picker mid-thread.
  3. Observe the warning dialog indicating that switching to LLM B may degrade performance.
  4. Without generating any output with LLM B, select LLM A again to return to the active context.
  • The degraded performance warning dialog appears again when switching back to LLM A.
  • The warning logic compares against the latest chosen model in the picker (LLM B) rather than verifying which model actually ran the last generation cycle.
What is the expected behavior?
  • Switching back to the model that produced the last generation (LLM A) should be recognized as returning to the active model state.
  • The warning dialog should evaluate against the last LLM used for generation, suppressing false-positive warnings when the user reverts an uncommitted picker change.
Additional information

No response

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start at the model picker and the degraded performance warning dialog, then trace how the selected model and the last generation model are tracked during a conversation. Reproduce the A/B switch sequence described in the issue and verify that returning to A after no generation with B no longer shows the warning.

Written by the indexing model from the issue text.

Assessment

Tech stack
rust
Domain
desktop
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Active
Clarity
Mostly clear
Newbie friendliness
55/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.