{ "error": { "message": "litellm.BadRequestError: OpenrouterException - {"error":{"message":"3.type: Invalid input: expected \"function\"; 3.function: Invalid input: expected object, received undefined","code":400,"metadata":{"provider_name":"OpenAI"}},"user_id":"org_..."}. Received Model Group=gpt-5.4\nAvailable Model Group Fallbacks=None", "type": null, "param": null, "code": "400" } }

Fix Plan

To resolve the issue with LiteLLM Proxy and OpenRouter, we need to ensure that the tools array is correctly serialized when proxying to OpenRouter. The error indicates that the 4th tool in the array (index 3) has a malformed entry.

Here are the steps to fix the issue:

• Check the tool definitions in the tools array and ensure that each entry has a type field set to "function" and a valid function object.
• Verify that the tools array is not being modified or transformed incorrectly when routing through OpenRouter.
• Update the LiteLLM Proxy code to correctly serialize the tools array when proxying to OpenRouter.

Example code snippet to validate the tools array:

def validate_tools_array(tools):
    for i, tool in enumerate(tools):
        if tool['type'] != 'function' or not tool.get('function'):
            raise ValueError(f"Malformed tool entry at index {i}: {tool}")
    return tools

• Use this function to validate the tools array before proxying the request to OpenRouter.

Verification

To verify that the fix worked, send a request with the updated tools array and check that the response does not contain a 400 error. You can also log the tools array before and after serialization to ensure that it is being correctly transformed.

Extra Tips

• Ensure that the LiteLLM Proxy version is up-to-date, as this issue may have been fixed in a later version.
• If the issue persists, try logging the entire request and response to identify any other potential issues with the serialization or deserialization of the tools array.