Skip to content

Provider API rejects audio/video content parts for multimodal models #906

Description

@jtoronto

Feature Description

I’m trying to call these models through the Provider API:

meta/muse-spark-1.3-contributor
Qwen/Qwen3.8-Omni-Flash
using POST https://api.commandcode.ai/provider/v1/chat/completions.

I need to submit:

an MP3 as input_audio
an MP4 as video_url with a Base64 data URL
Both models return:

{
"error": {
"message": "Invalid input",
"type": "invalid_request_error",
"param": "messages.1.content"
}
}
Your provider documentation says “Text and images. Audio, file, and document parts are rejected by the schema,” but Qwen 3.8 Omni Flash is advertised as omni-modal and supports audio/video through its native OpenAI-compatible API.

Use Case

In the case of Qwen 3.8 Omni Flash, media understanding is its main feature. To not support media input would be an oversight. While the Command Code harness itself may not support media, there are a lot of applications that can.

Additional Context

No response

How important is this to you?

Important for my workflow

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions