Azure OpenAI Data Files and Regions

Question

Azure OpenAI Data Files and Regions

Siry, Gaetan 280

Hello,

As far as I understand, right now, it is not possible to reference files stored in OpenAI Data Files in Regions East US from a model deployed in Region EAST US 2 .

Can you please confirm ?

and if so what strategy have people adopted for that ?

Some of the new models like gpt-5.1 are not available in EAST US, but that is where all of our files are, so now we have to deploy gpt-5.1 to EAST US 2 ( where it is available ) but it will not have access to any of our files.

Thank you

Aryan Parashar 3,380 Microsoft External Staff Moderator

Hi Siry, Gaetan,

I hope you're doing well.

I wanted to confirm that it isn’t possible to reference files stored in OpenAI Data Files in East US from a model deployed in East US 2, so the files need to be transferred between the two resources. Below is the script that uploads the files from the source resource to the destination resource:

import io
from openai import OpenAI

# -----------------------------
# SOURCE RESOURCE (read files)
# -----------------------------
source_client = OpenAI(
    api_key = "<API_KEY>",
    base_url="https://<RESOURCE-NAME>.openai.azure.com/openai/v1/" 
)

# -----------------------------
# DESTINATION RESOURCE (upload)
# -----------------------------
dest_client = OpenAI(
    api_key = "<API-KEY>,
    base_url="https://<RESOURCE-NAME>.openai.azure.com/openai/v1/" 
)

PURPOSE_MAPPING = {
    # Valid new purposes
    "fine-tune": "fine-tune",
    "assistants": "assistants",
    
    # Legacy 'assistants'/'file-search' types
    "file-search": "assistants",
    "code_interpreter": "assistants",
    
    # Deprecated purposes that we map to 'assistants'
    # These were used for the old /answers and /search endpoints
    "answers": "assistants",
    "search": "assistants",
    
    # Legacy classification purpose
    "classifications": "fine-tune",
    "batch":"batch",
}

# ------------------------------------------------
# STEP 1: List files in the SOURCE resource
# ------------------------------------------------
source_files = source_client.files.list()
print("Found", len(source_files.data), "files in source resource")

uploaded_ids = []

# ------------------------------------------------
# STEP 2: For each source file → download in-memory
# ------------------------------------------------
for f in source_files.data:
    print(f"\nTransferring file: {f.filename} ({f.id}) with ORIGINAL purpose: {f.purpose}")

    # --- START: NEW MAPPING LOGIC ---
    if f.purpose not in PURPOSE_MAPPING:
        print(f"SKIPPING file: Purpose '{f.purpose}' is unknown or deprecated and cannot be mapped. {f.filename}")
        continue # Skip to the next file
        
    new_purpose = PURPOSE_MAPPING[f.purpose]
    # --- END: NEW MAPPING LOGIC ---

    try:
        # Download file bytes
        file_content_response = source_client.files.content(f.id)
        file_bytes = file_content_response.content 

        # Convert bytes to an in-memory buffer
        buffer = io.BytesIO(file_bytes)

        # Create the correct file upload tuple (filename, file_object)
        upload_file_tuple = (f.filename, buffer)

        # ------------------------------------------------
        # STEP 3: Upload to the destination using the *new* purpose
        # ------------------------------------------------
        uploaded = dest_client.files.create(
            file=upload_file_tuple,
            purpose=new_purpose  # Use the mapped purpose
        )

        uploaded_ids.append(uploaded.id)
        print(f"Uploaded file ID: {uploaded.id} (New Purpose: {new_purpose})")

    except Exception as e:
        print(f"FAILED to transfer file {f.filename} ({f.id}): {e}")


print("\nAll uploaded file IDs:", uploaded_ids)

Let me know if you have more queries.
Thank you for reaching out to The Microsoft Q&A Portal.

Aryan Parashar 3,380 Reputation points Microsoft External Staff Moderator

2025-11-20T05:14:46.3933333+00:00

Hi Siry, Gaetan,
Let me know if the provided solution worked for you.

Your answer

Aryan Parashar 3,380 Reputation points Microsoft External Staff Moderator

2025-11-20T05:14:46.3933333+00:00

Hi Siry, Gaetan,
Let me know if the provided solution worked for you.

Share via

Azure OpenAI Data Files and Regions

Your answer