IDE Extension goes blank

Open 💬 2 comments Opened Jul 19, 2026 by mistervaibhav
💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

What version of the IDE extension are you using?

1.129.1

What subscription do you have?

ChatGPT Pro

Which IDE are you using?

VS Code

What platform is your computer?

Darwin 25.5.0 arm64 arm

What issue are you seeing?

<img width="3024" height="1964" alt="Image" src="https://github.com/user-attachments/assets/74c68c40-1bc6-4750-9240-9a26f115872b" />

and after I restarted the IDE, i saw this at the end of the chat session i was working on

Invalid Value: 'tools'. Function 'collaboration.spawn_agent' is reserved for use by this model and must match the configured schema.

What steps can reproduce the bug?

Happens randomly for me, while working,

I told 5.6 Sol to spawn subagents running on GPT-5.3-Codex-Spark to write code instead of coding itself - this particular time, so I am suspecting this has to do with this blank screen, but this has happened with me before as well without any special instructions

What is the expected behavior?

It must not go blank, and maybe tell me if I did something wrong on my end

Additional information

_No response_

View original on GitHub ↗

2 Comments

github-actions[bot] contributor · 1 month ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

  • #34024

Powered by Codex Action

BrimborionInc · 18 days ago

Additional current reproduction from a long-running engineering task:

Environment

  • Codex IDE extension: openai.chatgpt-26.803.41515-linux-x64
  • VS Code: 1.132.0 (df53daabb18cd157bdb08c7f01c34df936cf12f4)
  • Host: WSL2, Linux 5.15.167.4-microsoft-standard-WSL2, x86_64

What happened
A single autonomous thread had become very large after an extended code review/test task involving many tool calls, repeated context compaction, several collaboration subagents, and long-running command polling. While the assistant was polling a release-qualification process, the Codex panel became a plain grey screen and was unresponsive. There was no visible error, recovery action, or usable completion state. The user had to abandon the thread and request a manually prepared context handoff for a new task thread.

This is not an extension-startup blank screen: the panel had been working for a long time and failed under sustained use in one large active conversation. The failure risks losing visibility into an in-flight command and its qualification evidence.

Approximate reproduction

  1. Start one long Codex IDE thread.
  2. Run an autonomous, multi-stage repository task for an extended period.
  3. Use multiple subagents and many tool calls, with the thread compacted as it grows.
  4. Keep a long-running terminal command active and poll it periodically.
  5. Eventually the panel becomes grey/unresponsive instead of rendering the active turn.

Expected
The thread should remain responsive. If the renderer/session reaches a size or resource limit, the extension should fail visibly and recoverably—e.g. compact/page old content, preserve the active process state, and offer Resume/New thread with generated handoff—instead of showing a blank grey panel.

This appears closely related to #8197, but it still reproduces on the current 26.803.41515 extension and the trigger here strongly correlates with a very large tool/subagent-heavy thread. No repository contents or private logs are included; redacted diagnostics can be supplied if requested.