Request Syntax URI Request Parameters Request Body Response Syntax Response Elements Errors See Also

InvokeEndpointWithResponseStream

Invokes a model at the specified endpoint to return the inference response as a stream. The inference stream provides the response payload incrementally as a series of parts. Before you can get an inference stream, you must have access to a model that's deployed using Amazon SageMaker hosting services, and the container for that model must support inference streaming.

For more information that can help you use this API, see the following sections in the Amazon SageMaker Developer Guide:

For information about how to add streaming support to a model, see How Containers Serve Requests.
For information about how to process the streaming response, see Invoke real-time endpoints.

Before you can use this operation, your IAM permissions must allow the sagemaker:InvokeEndpoint action. For more information about Amazon SageMaker actions for IAM policies, see Actions, resources, and condition keys for Amazon SageMaker in the IAM Service Authorization Reference.

Amazon SageMaker strips all POST headers except those supported by the API. Amazon SageMaker might add additional headers. You should not rely on the behavior of headers outside those enumerated in the request syntax.

Calls to InvokeEndpointWithResponseStream are authenticated by using AWS Signature Version 4. For information, see Authenticating Requests (AWS Signature Version 4) in the Amazon S3 API Reference.

Request Syntax


POST /endpoints/EndpointName/invocations-response-stream HTTP/1.1
Content-Type: ContentType
X-Amzn-SageMaker-Accept: Accept
X-Amzn-SageMaker-Custom-Attributes: CustomAttributes
X-Amzn-SageMaker-Target-Variant: TargetVariant
X-Amzn-SageMaker-Target-Container-Hostname: TargetContainerHostname
X-Amzn-SageMaker-Inference-Id: InferenceId
X-Amzn-SageMaker-Inference-Component: InferenceComponentName
X-Amzn-SageMaker-Session-Id: SessionId

Body

URI Request Parameters

The request uses the following URI parameters.

Accept

The desired MIME type of the inference response from the model container.

Length Constraints: Maximum length of 1024.

Pattern: \p{ASCII}*

ContentType

The MIME type of the input data in the request body.

Length Constraints: Maximum length of 1024.

Pattern: \p{ASCII}*

CustomAttributes

Provides additional information about a request for an inference submitted to a model hosted at an Amazon SageMaker endpoint. The information is an opaque value that is forwarded verbatim. You could use this value, for example, to provide an ID that you can use to track a request or to provide other metadata that a service endpoint was programmed to process. The value must consist of no more than 1024 visible US-ASCII characters as specified in Section 3.3.6. Field Value Components of the Hypertext Transfer Protocol (HTTP/1.1).

The code in your model is responsible for setting or updating any custom attributes in the response. If your code does not set this value in the response, an empty value is returned. For example, if a custom attribute represents the trace ID, your model can prepend the custom attribute with Trace ID: in your post-processing function.

This feature is currently supported in the AWS SDKs but not in the Amazon SageMaker Python SDK.

Length Constraints: Maximum length of 1024.

Pattern: \p{ASCII}*

EndpointName

The name of the endpoint that you specified when you created the endpoint using the CreateEndpoint API.

Length Constraints: Maximum length of 63.

Pattern: ^[a-zA-Z0-9](-*[a-zA-Z0-9])*

Required: Yes

InferenceComponentName

If the endpoint hosts one or more inference components, this parameter specifies the name of inference component to invoke for a streaming response.

Length Constraints: Maximum length of 63.

Pattern: ^[a-zA-Z0-9]([\-a-zA-Z0-9]*[a-zA-Z0-9])?$

InferenceId

An identifier that you assign to your request.

Length Constraints: Minimum length of 1. Maximum length of 64.

Pattern: \A\S[\p{Print}]*\z

SessionId

The ID of a stateful session to handle your request.

You can't create a stateful session by using the InvokeEndpointWithResponseStream action. Instead, you can create one by using the InvokeEndpoint action. In your request, you specify NEW_SESSION for the SessionId request parameter. The response to that request provides the session ID for the NewSessionId response parameter.

Length Constraints: Maximum length of 256.

Pattern: ^[a-zA-Z0-9](-*[a-zA-Z0-9])*$

TargetContainerHostname

If the endpoint hosts multiple containers and is configured to use direct invocation, this parameter specifies the host name of the container to invoke.

Length Constraints: Maximum length of 63.

Pattern: ^[a-zA-Z0-9](-*[a-zA-Z0-9])*

TargetVariant

Specify the production variant to send the inference request to when invoking an endpoint that is running two or more variants. Note that this parameter overrides the default behavior for the endpoint, which is to distribute the invocation traffic based on the variant weights.

For information about how to use variant targeting to perform a/b testing, see Test models in production

Length Constraints: Maximum length of 63.

Pattern: ^[a-zA-Z0-9](-*[a-zA-Z0-9])*

Request Body

The request accepts the following binary data.

Body

Provides input data, in the format specified in the ContentType request header. Amazon SageMaker passes all of the data in the body to the model.

For information about the format of the request body, see Common Data Formats-Inference.

Length Constraints: Maximum length of 6291456.

Required: Yes

Response Syntax


HTTP/1.1 200
X-Amzn-SageMaker-Content-Type: ContentType
x-Amzn-Invoked-Production-Variant: InvokedProductionVariant
X-Amzn-SageMaker-Custom-Attributes: CustomAttributes
Content-type: application/json

{
   "InternalStreamFailure": { 
   },
   "ModelStreamError": { 
   },
   "PayloadPart": { 
      "Bytes": blob
   }
}

Response Elements

If the action is successful, the service sends back an HTTP 200 response.

The response returns the following HTTP headers.

ContentType

The MIME type of the inference returned from the model container.

Length Constraints: Maximum length of 1024.

Pattern: \p{ASCII}*

CustomAttributes

Provides additional information in the response about the inference returned by a model hosted at an Amazon SageMaker endpoint. The information is an opaque value that is forwarded verbatim. You could use this value, for example, to return an ID received in the CustomAttributes header of a request or other metadata that a service endpoint was programmed to produce. The value must consist of no more than 1024 visible US-ASCII characters as specified in Section 3.3.6. Field Value Components of the Hypertext Transfer Protocol (HTTP/1.1). If the customer wants the custom attribute returned, the model must set the custom attribute to be included on the way back.

This feature is currently supported in the AWS SDKs but not in the Amazon SageMaker Python SDK.

Length Constraints: Maximum length of 1024.

Pattern: \p{ASCII}*

InvokedProductionVariant

Identifies the production variant that was invoked.

Length Constraints: Maximum length of 1024.

Pattern: \p{ASCII}*

The following data is returned in JSON format by the service.

InternalStreamFailure

The stream processing failed because of an unknown error, exception or failure. Try your request again.

Type: Exception

HTTP Status Code:

ModelStreamError

An error occurred while streaming the response body. This error can have the following error codes:

ModelInvocationTimeExceeded: The model failed to finish sending the response within the timeout period allowed by Amazon SageMaker.
StreamBroken: The Transmission Control Protocol (TCP) connection between the client and the model was reset or closed.

Type: Exception

HTTP Status Code:

PayloadPart

A wrapper for pieces of the payload that's returned in response to a streaming inference request. A streaming inference response consists of one or more payload parts.

Type: PayloadPart object

Errors

For information about the errors that are common to all actions, see Common Errors.

InternalFailure

An internal failure occurred.

HTTP Status Code: 500

InternalStreamFailure

The stream processing failed because of an unknown error, exception or failure. Try your request again.

HTTP Status Code: 500

ModelError

Model (owned by the customer in the container) returned 4xx or 5xx error code.

HTTP Status Code: 424

ModelStreamError

An error occurred while streaming the response body. This error can have the following error codes:

ModelInvocationTimeExceeded: The model failed to finish sending the response within the timeout period allowed by Amazon SageMaker.
StreamBroken: The Transmission Control Protocol (TCP) connection between the client and the model was reset or closed.

HTTP Status Code: 400

ServiceUnavailable

The service is unavailable. Try your call again.

HTTP Status Code: 503

ValidationError

Inspect your request and try again.

HTTP Status Code: 400