Title: /conversations/{id}/items response format incompatible with OpenAI spec
The response structure from /conversations/{id}/items does not match the expected OpenAI-compatible format, causing parsing issues for clients expecting standard responses.
Issues
1. Nested text object in assistant responses
Currently returns:
{
"text": {
"text": "response content"
}
}
Expected:
{
"text": "response content"
}
2. Incorrect content type values
| Current |
Expected |
input_text |
text |
reasoning_content->reasoning_text |
text |
Impact
Clients built against the OpenAI spec need additional transformation logic to handle these differences, breaking drop-in compatibility.