Add deployment config field to SagemakerEndpoint.

libs/aws/langchain_aws/llms/sagemaker_endpoint.py

@@ -133,9 +133,15 @@ class LLMContentHandler(ContentHandlerBase[str, str]):
 class SagemakerEndpoint(LLM):
     """Sagemaker Inference Endpoint models.
 
-    To use, you must supply the endpoint name from your deployed
+    To use with a pre-deployed SageMaker endpoint or inference component, you must 
+    supply the endpoint name and optional inference component name from your deployed
     Sagemaker model & the region where it is deployed.
 
+    To use with undeployed SageMaker resources, you can supply an endpoint name, 
+    optional inference component name, and deployment configuration which defines
+    the endpoint and model configs. This construct can then be used by the SageMaker
+    PythonSDK ModelBuilder class to deploy a Sagemaker model on the desired compute.
+
     To authenticate, the AWS client uses the following methods to
     automatically load credentials:
     https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html
@@ -191,8 +197,7 @@ class SagemakerEndpoint(LLM):
                 region_name=region_name,
                 credentials_profile_name=credentials_profile_name
             )
-
-        #Use with boto3 client
+            # Usage with boto3 client
             client = boto3.client(
                         "sagemaker-runtime",
                         region_name=region_name
@@ -208,7 +213,7 @@ class SagemakerEndpoint(LLM):
     """Boto3 client for sagemaker runtime"""
 
     endpoint_name: str = ""
-    """The name of the endpoint from the deployed Sagemaker model.
+    """The name of the endpoint created from a Sagemaker model.
     Must be unique within an AWS Region."""
 
     inference_component_name: Optional[str] = None
@@ -263,6 +268,33 @@ def transform_output(self, output: bytes) -> str:
     .. _boto3: <https://boto3.amazonaws.com/v1/documentation/api/latest/index.html>
     """
 
+    deployment_config: Optional[Dict] = None
+    """The deployment configuration for an undeployed endpoint or inference component 
+    which can be deployed through the Sagemaker Python SDK ModelBuilder class. 
+    Comprises two sub-dictionaries model_config and endpoint_config.
+    """
+
+    """
+     Schema:
+        .. code-block:: python
+        deployment_config = {
+            "model_config": {
+                "model": Optional[str],
+                "model_path": Optional[str],
+                "image_uri": Optional[str],
+                "model_server": Optional[str],
+                "content_type": Optional[str],
+                "accept_type": Optional[str]
+            },
+            "endpoint_config": {
+                "resources": Optional[Dict[str, int]],
+                "instance_type": Optional[str],
+                "initial_instance_count": Optional[int]
+            },
+            "tags": Optional[List[Dict]]
+        }
+    """
+
     model_config = ConfigDict(
         extra="forbid",
     )