Update Responses API unsupported list with 2 edits

prompt not supported

Groq’s official docs state these as unsupported features of the Responses API.

But it seems prompt is also not supported, because it’s unique to OpenAI. From their docs, prompt is used for setting prompt templates on their platform! OpenAI Doc Link on Responses > prompt.

developer as role isn’t supported

 input=[
        {
            "role": "developer", #not supported for non gpt! use "system"
            "content": "Talk like a pirate."
        },
        {
            "role": "user",
            "content": "Are semicolons optional in JavaScript?"
        }
    ]

This could be dependent on model selection but for Llama 3.3 70b “developer” isn’t supported and it fails silently. This could be a feature of the API being OpenAI specific.

For anyone else looking for this in the future, if you are using response api then keep using “system” as the the main role for system prompts for models other than gpt I suppose.

1 Like

That’s a good point, will consider it for our documentation

1 Like