You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Report Azure API Bug [Invalid 'input[2].id'] #9131
{
"error": {
"message": "Invalid 'input[2].id': 'review:rollout:user'. Expected an ID that contains letters, numbers, underscores, or dashes, but thisvalue contained additional characters.","type": "invalid_request_error",
"param": "input[2].id",
"code": "invalid_value"
}
}
What steps can reproduce the bug?
Open CODEX-CLI.
In a project, trigger a “Code review.”
After the review output finishes, continue in the same session with a follow-up prompt / next task (e.g., “upgrade this session’s work to the latest framework” / “continue with an upgrade plan based on the review”).
When CODEX sends the next request, it immediately errors out and stops (instead of continuing to generate output).
Could you /feedback on the session and send us the session ID please?
I’m sorry—I can’t find the specific conversation from when the bug occurred. But I can enable those two experimental parameters and keep using it; if I run into the bug again, I’ll report it to you.
Activity
[-]Report Azure API Bug[/-][+]Report Azure API Bug [Invalid 'input[2].id'][/+]etraut-openai commentedon Jan 13, 2026
Could you please post your config settings for your model provider? Are you using the "responses" wire_api?
LeXwDeX commentedon Jan 14, 2026
Hi,this is my config
etraut-openai commentedon Jan 14, 2026
You're using two experimental features:
unified_execandshell_snapshot. If you disable these, does the problem go away?LeXwDeX commentedon Jan 14, 2026
I’ve been using it continuously for three hours, and so far I haven’t seen that error anymore.
jif-oai commentedon Jan 14, 2026
Could you
/feedbackon the session and send us the session ID please?LeXwDeX commentedon Jan 15, 2026
I’m sorry—I can’t find the specific conversation from when the bug occurred. But I can enable those two experimental parameters and keep using it; if I run into the bug again, I’ll report it to you.
etraut-openai commentedon Jan 17, 2026
I think we understand this issue and have a proposed fix under review. Tracking in #9360, so I'm going to close this as duplicate.