What version of the IDE extension are you using?
openai.chatgpt-0.5.77-linux-x64 (Codex CLI reported by extension: 0.105.0)
What subscription do you have?
ChatGPT Team
Which IDE are you using?
VS Code (Remote - SSH)
What platform is your computer?
Linux 6.8.0-90-generic x86_64 x86_64 (remote host where extension host runs)
What issue are you seeing?
On a long-running Codex conversation, every new prompt fails before response generation with:
Error running remote compact task: unexpected status 413 Payload Too Large ... url: https://chatgpt.com/backend-api/codex/responses/compact
From extension log (Codex.log), for example on 2026-02-26:
compact_remote: remote compaction failed
failing_compaction_request_model_visible_bytes=1524260 (also observed much larger in other attempts)
cf-ray: 9d3d48a5f8da61ae-ORD
The extension keeps auto-resuming the same large conversation, so the failure repeats.
What steps can reproduce the bug?
Use Codex IDE extension in VS Code Remote-SSH.
Continue the same conversation for many turns (especially with many pasted images and large tool outputs).
Close/reopen VS Code (or reconnect Remote-SSH) so extension resumes prior conversation.
Send a short prompt (even “hi”).
Observe immediate failure with remote compact task and 413 Payload Too Large
What is the expected behavior?
The extension should handle oversized history gracefully instead of failing the turn, for example by:
Auto-pruning/chunking compaction payload when request size approaches gateway limits.
Excluding or replacing old inline image blobs from compaction input.
Auto-forking to a new thread (with summary) if compaction cannot be sent.
Showing a clear UI recovery action instead of hard-failing every prompt.
Additional information
No response
What version of the IDE extension are you using?
openai.chatgpt-0.5.77-linux-x64 (Codex CLI reported by extension: 0.105.0)
What subscription do you have?
ChatGPT Team
Which IDE are you using?
VS Code (Remote - SSH)
What platform is your computer?
Linux 6.8.0-90-generic x86_64 x86_64 (remote host where extension host runs)
What issue are you seeing?
On a long-running Codex conversation, every new prompt fails before response generation with:
Error running remote compact task: unexpected status 413 Payload Too Large ... url: https://chatgpt.com/backend-api/codex/responses/compact
From extension log (Codex.log), for example on 2026-02-26:
compact_remote: remote compaction failed
failing_compaction_request_model_visible_bytes=1524260 (also observed much larger in other attempts)
cf-ray: 9d3d48a5f8da61ae-ORD
The extension keeps auto-resuming the same large conversation, so the failure repeats.
What steps can reproduce the bug?
Use Codex IDE extension in VS Code Remote-SSH.
Continue the same conversation for many turns (especially with many pasted images and large tool outputs).
Close/reopen VS Code (or reconnect Remote-SSH) so extension resumes prior conversation.
Send a short prompt (even “hi”).
Observe immediate failure with remote compact task and 413 Payload Too Large
What is the expected behavior?
The extension should handle oversized history gracefully instead of failing the turn, for example by:
Auto-pruning/chunking compaction payload when request size approaches gateway limits.
Excluding or replacing old inline image blobs from compaction input.
Auto-forking to a new thread (with summary) if compaction cannot be sent.
Showing a clear UI recovery action instead of hard-failing every prompt.
Additional information
No response