Implementation:BerriAI Litellm Create File

Knowledge Sources	Domains	Last Updated
BerriAI/litellm	File Management, API Integration, Cloud Storage	2026-02-15

Overview

Concrete tool for uploading training and other data files to LLM providers provided by LiteLLM.

Description

The create_file function in LiteLLM provides a unified interface for uploading files to multiple LLM providers including OpenAI, Azure OpenAI, Vertex AI, Gemini, Bedrock, hosted vLLM, and Manus. It accepts a file object, a purpose tag (such as "fine-tune"), and a provider specification, then routes the upload to the appropriate provider-specific handler. The function handles API base resolution, credential lookup, timeout configuration, and response normalization, returning a standardized OpenAIFileObject regardless of the underlying provider. Both synchronous and asynchronous execution modes are supported.

Usage

Use create_file when:

• Uploading a JSONL training file before creating a fine-tuning job.
• Uploading files for the Assistants API or Batch API.
• Needing a single function call that works across OpenAI, Azure, Vertex AI, Bedrock, and other supported providers.

Code Reference

Source Location

litellm/files/main.py (lines 60-292)

Signature

# Async version
async def acreate_file(
    file: FileTypes,
    purpose: Literal["assistants", "batch", "fine-tune"],
    expires_after: Optional[FileExpiresAfter] = None,
    custom_llm_provider: Literal[
        "openai", "azure", "gemini", "vertex_ai", "bedrock", "hosted_vllm", "manus"
    ] = "openai",
    extra_headers: Optional[Dict[str, str]] = None,
    extra_body: Optional[Dict[str, str]] = None,
    **kwargs,
) -> OpenAIFileObject:

# Sync version
def create_file(
    file: FileTypes,
    purpose: Literal["assistants", "batch", "fine-tune"],
    expires_after: Optional[FileExpiresAfter] = None,
    custom_llm_provider: Optional[Literal[
        "openai", "azure", "gemini", "vertex_ai", "bedrock", "hosted_vllm", "manus"
    ]] = None,
    extra_headers: Optional[Dict[str, str]] = None,
    extra_body: Optional[Dict[str, str]] = None,
    **kwargs,
) -> Union[OpenAIFileObject, Coroutine[Any, Any, OpenAIFileObject]]:

Import

from litellm.files.main import create_file, acreate_file

I/O Contract

Inputs

Parameter	Type	Required	Description
file	FileTypes	Yes	The file content to upload. Accepts file-like objects, tuples of (filename, content), or bytes.
purpose	Literal["assistants", "batch", "fine-tune"]	Yes	The intended purpose of the file. Use "fine-tune" for training data.
expires_after	Optional[FileExpiresAfter]	No	Expiration policy for the uploaded file.
custom_llm_provider	Optional[Literal["openai", "azure", "gemini", "vertex_ai", "bedrock", "hosted_vllm", "manus"]]	No	The LLM provider to upload to. Defaults to "openai".
extra_headers	Optional[Dict[str, str]]	No	Additional HTTP headers to include in the upload request.
extra_body	Optional[Dict[str, str]]	No	Additional fields to include in the request body.
**kwargs	various	No	Additional parameters including api_key, api_base, timeout, and provider-specific options.

Outputs

Return Type	Description
OpenAIFileObject	A normalized file object containing the file ID, filename, purpose, bytes count, creation timestamp, and status. For sync calls, may return a Coroutine if called from an async context internally.

Usage Examples

Upload a training file to OpenAI

from litellm.files.main import create_file

# Upload a JSONL training file
file_object = create_file(
    file=open("training_data.jsonl", "rb"),
    purpose="fine-tune",
    custom_llm_provider="openai",
)

print(file_object.id)        # e.g., "file-abc123"
print(file_object.purpose)   # "fine-tune"
print(file_object.filename)  # "training_data.jsonl"

Async upload to Azure OpenAI

import asyncio
from litellm.files.main import acreate_file

async def upload_training_data():
    file_object = await acreate_file(
        file=open("training_data.jsonl", "rb"),
        purpose="fine-tune",
        custom_llm_provider="azure",
        api_base="https://my-resource.openai.azure.com/",
        api_key="my-azure-api-key",
        api_version="2024-02-01",
    )
    return file_object

result = asyncio.run(upload_training_data())
print(result.id)

Upload to Vertex AI

from litellm.files.main import create_file

file_object = create_file(
    file=open("training_data.jsonl", "rb"),
    purpose="fine-tune",
    custom_llm_provider="vertex_ai",
    vertex_project="my-gcp-project",
    vertex_location="us-central1",
)
print(file_object.id)

Related Pages

• Principle:BerriAI_Litellm_Training_File_Upload