Completion Failure with all OpenAI models within the Conversation flow

[isenaymgsu-oss]

Checked for Existing Issues?

• Yes, I have checked existing issues and cannot find one related to this.

Software

Desktop Application

Operating System / Platform

Windows

Your Pieces OS Version

PiecesOS 12.4.1

Early Access Program

• Yes, this is related to an Early Access Program feature.

Kindly describe the bug and include as much detail as possible on what you were doing so we can reproduce the bug.

Error message: from OpenRouter/Azure
Completion failed

What to try: The model could not complete the request. This is usually transient — try again in a moment.

Try again: Send another message (for example, Keep going) to resume from where the agent left off — the provider may have been temporarily overloaded.

Technical detail: openrouter rejected (400): {
"error": {
"message": "Unknown parameter: 'usage'.",
"type": "invalid_request_error",
"param": "usage",
"code": "unknown_parameter"
}
} [upstream: Azure] [model=openai/gpt-5.4-nano, provider=openrouter]

and then on a separate frame:

openrouter rejected (400): {
"error": {
"message": "Unknown parameter: 'usage'.",
"type": "invalid_request_error",
"param": "usage",
"code": "unknown_parameter"
}
} [upstream: Azure] [model=openai/gpt-5.4-nano, provider=openrouter]

It happens with ChatGPT but not with Claude/Gemini
I've cleared cache and restarted

[mark-at-pieces]

Hey @isenaymgsu-oss

Thank you for reaching out to support.

I confirmed the same behavior on my system. We believe this is a result of https://status.openai.com/ having some issues with their API and experiencing an increased number of failures.

Thank you for pinging us on this. We will keep an eye on this throughout the day!

Additionally, confirmed Claude/Gemini models are functioning properly within the system, so we would leverage either of these providers temporarily until OpenAI is back up and running.

[mark-at-pieces]

Hey @isenaymgsu-oss — quick update on this.

We've elevated this issue internally and believe there may be a broader problem affecting OpenAI model completions beyond what was initially reported. We're looking into this deeper on our end and will report back once we have additional information.

In the meantime, Claude and Gemini models are unaffected — you can switch to either of those as a workaround while we resolve this.

Thank you for your patience!

[mark-at-pieces]

hey @isenaymgsu-oss

There was an issue w/ Openrouter, where they pushed a bug to prod on their end. Ultimately, this is a no-code change on our end and should now be fully functional.

Let us know if this is now fixed on your end!

Thanks again for reaching out!