fix(completion): transform image content in tool results for Responses API

[Chesars]

Relevant issues

Fixes #17762

Pre-Submission checklist

• I have Added testing in the tests/litellm/ directory
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible, it only solves 1 specific problem

Type

🐛 Bug Fix

Changes

When using litellm.completion() with model="openai/responses/...", images in tool message content were not being transformed from Chat Completion format to Responses API format.

Example Request

import litellm

messages = [
    {"role": "user", "content": "Fetch the image"},
    {
        "role": "assistant",
        "content": None,
        "tool_calls": [
            {
                "id": "call_abc123",
                "type": "function",
                "function": {
                    "name": "fetch_image",
                    "arguments": '{"url": "https://example.com/image.png"}'
                }
            }
        ]
    },
    {
        "role": "tool",
        "tool_call_id": "call_abc123",
        "content": [
            {
                "type": "image_url",
                "image_url": {
                    "url": "data:image/png;base64,iVBORw0KGgo..."
                }
            }
        ]
    },
    {"role": "user", "content": "What color is the image?"}
]

response = litellm.completion(
    model="openai/responses/gpt-4.1",
    messages=messages,
    tools=[{
        "type": "function",
        "function": {
            "name": "fetch_image",
            "parameters": {"type": "object", "properties": {"url": {"type": "string"}}}
        }
    }]
)

The tool message content was being passed directly without transformation:

Chat Completion format (what user sends):

{"type": "image_url", "image_url": {"url": "data:image/png;base64,..."}}

Responses API format (what OpenAI expects):

{"type": "input_image", "image_url": "data:image/png;base64,..."}

OpenAI reject the request with error 400:

Invalid value: 'image_url'. Supported values are: 'input_text', 'input_image', 'output_text', ...

Fix

Transform tool message content using _convert_content_to_responses_format() when the content is a list (multimodal content with images).

Tests Added

File: tests/test_litellm/completion_extras/litellm_responses_transformation/test_completion_extras_litellm_responses_transformation_transformation.py

Added test_convert_chat_completion_messages_to_responses_api_tool_result_with_image() which verifies:

• Tool messages with image_url content are correctly transformed
• Output type is input_image (not image_url)
• Output image_url is a flat string (not nested object)
• detail field is preserved"

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
litellm	Error			Dec 11, 2025 0:56am