Issue Station
Sign In
Tools hereToolbox

codex - 💡(How to fix) Fix Queued prompts behave like steer prompts; multiple queued prompts sent simultaneously [1 comments, 2 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
openai/codex#17285•Fetched 2026-04-10 03:43:03
View on GitHub
Comments
1
Participants
2
Timeline
4
Reactions
0
Author
diegocp01diegocp01
Participants
diegocp01diegocp01github-actions[bot]github-actions[bot]
Timeline (top)
labeled ×2commented ×1cross-referenced ×1
RAW_BUFFERClick to expand / collapse

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

26.406.31014

What subscription do you have?

Plus

What platform is your computer?

macOS Tahoe 26.5

What issue are you seeing?

Problem Queued prompts should wait until the full inference finishes; steer prompts should send after the last tool call. After the latest update, queued prompts are being sent like steer prompts (as soon as the tool finishes).

Steps to reproduce

  1. Start a new chat in the Codex app on macOS, and run a task where Codex alternates tool execution with text responses (tool -> text -> tool).
  2. While a tool is executing, queue a prompt.
  3. Observe it shows in the queue, but when the tool finishes Codex sends the queued prompt immediately instead of waiting for inference to end.

Additional issue

  • If you queue more than one prompt while a tool is executing, when the tool finishes Codex can send multiple queued prompts at the same time.

Expected behavior

  • Queued prompts should be sent only after the entire inference ends.
  • If multiple prompts are queued, they should be sent one at a time.

Actual behavior

  • Queued prompts are sent when the tool execution finishes (steer behavior).
  • Multiple queued prompts can be sent simultaneously after tool completion.

Environment

  • Codex App version: 26.406.31014
  • Platform: macOS Tahoe 26.5
  • Subscription: Plus

What steps can reproduce the bug?

  1. Start a new chat in the Codex app on macOS.
  2. Begin a run where the model executes a tool (tool call happens before the model finishes inference).
  3. While the tool is still running, queue one or more prompts.
  4. When the tool finishes, the queued prompts are sent immediately (as if they were steer prompts), instead of waiting for the full inference to complete.

Additional issue:

  • If you have more than one queued prompt, when the tool finishes Codex sends multiple queued prompts simultaneously instead of one at a time.

No code snippet needed—this is UI workflow behavior.

What is the expected behavior?

Queued prompts should send only after the full inference completes (not when the tool finishes), and queued prompts should be sent one at a time (not simultaneously).

Additional information

No response

extent analysis

TL;DR

The issue can be mitigated by modifying the prompt queuing logic to wait for the entire inference to end before sending queued prompts.

Guidance

  • Review the Codex app's logic for handling queued prompts and tool execution to identify where the queuing behavior is being triggered prematurely.
  • Verify that the inference completion event is correctly detected and used as a trigger for sending queued prompts.
  • Consider implementing a queue management system that ensures prompts are sent one at a time, even if multiple prompts are queued while a tool is executing.
  • Test the modified logic with multiple queued prompts to ensure they are sent sequentially after the inference completes.

Notes

The issue seems to be related to the timing of when queued prompts are sent in relation to tool execution and inference completion. Without access to the Codex app's codebase, it's difficult to provide a more specific solution.

Recommendation

Apply a workaround by modifying the prompt queuing logic to wait for the entire inference to end before sending queued prompts, as this directly addresses the reported issue.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#autograd error#model save/load#optimization#mixed precision#training loop

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING