ModelBuilder.deploy(update_endpoint=True) raises AttributeError on Session.create_endpoint_config()

[gakugaku]

PySDK Version

• PySDK V2 (2.x)
• PySDK V3 (3.x)

Describe the bug

ModelBuilder.deploy(update_endpoint=True) raises AttributeError: 'Session' object has no attribute 'create_endpoint_config'.

In the installed V3 packages, ModelBuilder._deploy_core_endpoint() calls self.sagemaker_session.create_endpoint_config(...) on the update path, but sagemaker.core.helper.session_helper.Session does not implement that method.

To reproduce

Run the following script:

from types import SimpleNamespace

import boto3
from sagemaker.core.helper.session_helper import Session
from sagemaker.serve.model_builder import ModelBuilder


builder = ModelBuilder(
    s3_model_data_url={
        "S3DataSource": {
            "S3Uri": "s3://example-bucket/model/",
            "S3DataType": "S3Prefix",
            "CompressionType": "None",
        }
    },
    image_uri="123456789012.dkr.ecr.us-west-2.amazonaws.com/example:latest",
    role_arn="arn:aws:iam::123456789012:role/SageMakerExecutionRole",
    sagemaker_session=Session(boto_session=boto3.Session(region_name="us-west-2")),
    instance_type="ml.g5.xlarge",
    env_vars={},
)

builder.model_name = "demo-model"
builder.region = "us-west-2"
builder.endpoint_name = "demo-endpoint"
builder.built_model = SimpleNamespace(model_name="demo-model")

builder._deploy_core_endpoint(
    endpoint_name="demo-endpoint",
    instance_type="ml.g5.xlarge",
    initial_instance_count=1,
    wait=False,
    update_endpoint=True,
)

Steps

1. Create a Python environment with the V3 SageMaker packages listed below.
2. Save the script above.
3. Run python minimal_repro_model_builder_update_endpoint.py.

Expected behavior

One of the following should happen:

• ModelBuilder.deploy(update_endpoint=True) should work with the default Session implementation used by V3 packages.
• Or ModelBuilder should use a session method that actually exists on sagemaker.core.helper.session_helper.Session.

Screenshots or logs

Actual error:

AttributeError: 'Session' object has no attribute 'create_endpoint_config'

I also confirmed locally that:

• hasattr(Session, "create_endpoint_config") == False
• ModelBuilder._deploy_core_endpoint() contains a call to self.sagemaker_session.create_endpoint_config(...)

In my real deployment code, this was triggered from ModelBuilder.build(...); ModelBuilder.deploy(..., update_endpoint=True) while updating an existing SageMaker endpoint.

System information

• SageMaker Python SDK version: V3 packages; sagemaker-core==2.5.0, sagemaker-serve==1.5.0
• Framework name (eg. PyTorch) or algorithm (eg. KMeans): N/A for repro; issue is in ModelBuilder update flow
• Framework version: N/A
• Python version: 3.14.3
• CPU or GPU: Repro itself is local and does not require either; production usage was GPU-backed SageMaker endpoint deployment
• Custom Docker image (Y/N): N

Additional context

• Avoiding update_endpoint=True and recreating the endpoint instead works around the issue.
• This looks related to the ongoing V3 API/session transition, and may be adjacent to issue AttributeError: module 'sagemaker' has no attribute 'Session' #5336, but this report is specifically about the missing create_endpoint_config() method on the session object used by ModelBuilder.