[Bug]: ModuleNotFoundError: No module named 'fastapi' in python SDK

[N-analyst]

What happened?

I've encountered a ModuleNotFoundError for fastapi when using what appears to be a core function of the litellm library.

It seems that some modules within the core library have a dependency on fastapi, but fastapi is not included as a core dependency of the package. This dependency seems to be required even when not using the proxy features, causing a crash in standard library usage.

I found the issue in v1.80.10, but I think the error occurred in v1.80.5 when update_responses_input_with_model_file_ids was added to the response() function.

• https://github.com/BerriAI/litellm/blob/v1.80.10.rc.4/litellm/responses/main.py#L584

When installing with the existing Python SDK, fastapi is optional and will not be installed.

• https://github.com/BerriAI/litellm/blob/v1.80.10.rc.4/pyproject.toml#L40

Relevant log output

### Traceback

Traceback (most recent call last):
  File "<venv>/lib/python3.12/site-packages/litellm/responses/main.py", line 584, in responses
    input = cast(Union[str, ResponseInputParam], update_responses_input_with_model_file_ids(input=input))
                                                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<venv>/lib/python3.12/site-packages/litellm/litellm_core_utils/prompt_templates/common_utils.py", line 450, in update_responses_input_with_model_file_ids
    from litellm.proxy.openai_files_endpoints.common_utils import (
  File "<venv>/lib/python3.12/site-packages/litellm/proxy/openai_files_endpoints/common_utils.py", line 7, in <module>
    from fastapi import Request
ModuleNotFoundError: No module named 'fastapi'
`

### Root Cause
The file litellm/proxy/openai_files_endpoints/common_utils.py imports fastapi at the module level:


from fastapi import Request  # Line 7

This module is imported by litellm/litellm_core_utils/prompt_templates/common_utils.py (line 450) during the execution of responses(), which is a core function that should work without proxy dependencies.

What part of LiteLLM is this about?

SDK (litellm Python package)

What LiteLLM version are you on ?

v1.80.10

Twitter / LinkedIn details

No response

[dinmukhamedm]

+1. I encountered this issue when using the responses and aresponses methods.

Minimal repro

import litellm

response = litellm.aresponses(
    model="openai/gpt-4.1-nano",
    input=[
        {
            "role": "user",
            "content": "What is the capital of France?",
        },
    ],
)

print(response.output[0].content[0].text)