fix: transform tool output content types for Responses API bridge

[Chesars]

Relevant issues

Fixes #17507

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

Description

When using litellm.completion() with models that only support the Responses API (gpt-5.1-codex, o1-pro, o3-pro, etc.), LiteLLM automatically transforms the request from Chat Completions format to Responses API format.

The Responses API expects tool results in this format:

# Responses API format 
{
    "type": "function_call_output",
    "call_id": "call_abc123",
    "output": [
        {"type": "output_text", "text": "Screenshot captured"},
        {"type": "input_image", "image_url": "data:image/png;base64,..."}
    ]
}

But we were sending Chat Completions format:

# Chat Completions format (what we were sending)
{
    "type": "function_call_output",
    "call_id": "call_abc123", 
    "output": [
        {"type": "text", "text": "Screenshot captured"},
        {"type": "image_url", "image_url": {"url": "data:image/png;base64,..."}}
    ]
}

Fix

Updated litellm/completion_extras/litellm_responses_transformation/transformation.py to transform tool output content when it's a list:

• {"type": "text"} → {"type": "output_text"}
• {"type": "image_url", "image_url": {"url": "..."}} → {"type": "input_image", "image_url": "..."}

Tests

• test_convert_tool_output_to_responses_format
• test_convert_tool_output_string_content_unchanged

[vercel]

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

Project	Deployment	Preview	Comments	Updated (UTC)
litellm	Ready	Preview	Comment	Dec 5, 2025 6:44pm