litellm - 💡(How to fix) Fix Responses API response has wrong object field: "chat.completion" instead of "response"

Description

The /v1/responses endpoint returns responses with "object": "chat.completion" instead of the OpenAI Responses API spec value "object": "response".

Root Cause

In litellm/responses/litellm_completion_transformation/transformation.py line 1645:

object=chat_completion_response.object,

This directly copies the internal Chat Completions response's object field rather than setting it to "response" for the Responses API response envelope.

Expected Behavior

OpenAI's Responses API returns "object": "response" in the top-level response. The LiteLLM Responses API endpoint should match this.

Actual Behavior

"object": "chat.completion" is returned, which breaks strict clients (e.g., OpenAI Codex CLI with wire_api = "responses") that validate the response envelope.

Steps to Reproduce

1. Start LiteLLM proxy with any model
2. Send a request to /v1/responses
3. Observe "object": "chat.completion" in the response

Fix

- object=chat_completion_response.object,
+ object="response",

Also need to clear __pycache__ for the change to take effect.

Related

Also noted in issue #11204 but was not fixed by the associated PR.

Environment

• LiteLLM version: main-latest (ghcr.io/berriai/litellm:main-latest)
• Python: 3.13