API Overview

MaaS Standard API V2

Introduces the call specifications for the chat API, provides detailed descriptions of input and output parameters, and shows usage examples.

OpenAI-compatible APIs

Allows calling models via an OpenAI-compatible chat API. Provides detailed descriptions of input and output parameters and shows usage examples.

Anthropic-compatible APIs

Allows calling models via an Anthropic-compatible chat API. Provides detailed descriptions of input and output parameters and shows usage examples.

MaaS Standard API V1

Introduces the call specifications for the chat API (applicable to select legacy models), provides detailed descriptions of input and output parameters, and shows usage examples. This API is no longer evolving; prioritize the use of MaaS Standard API V2.

Creating an Endpoint

This API is used to establish independent call entry points, set rate-limiting rules, and achieve precise fee tracking based on endpoint names. This assists you in effectively managing inference service resources and optimizing usage costs.

Editing an Endpoint

This API is used to edit the throttling settings, content guard switch, endpoint name, and endpoint description of a specified endpoint.

Querying Custom Endpoints

This API is used to query the endpoint list.

Deleting an Endpoint

This API is used to delete a specified endpoint based on the input custom_endpoint_id.

Enabling or Disabling an Endpoint

This API is used to enable or disable an endpoint based on the input project-id and custom_endpoint_id.

Obtaining Region Information

Used to obtain the region information of ModelArts real-time services. When source is set to custom_from_modelarts_v2 during the creation of a custom endpoint, the region parameter in the request is obtained.

Obtaining Workspace Information

Used to obtain the workspace information of ModelArts real-time services. When source is set to custom_from_modelarts_v2 during the creation of a custom endpoint, the infer_service_id is obtained.

Obtaining Inference Service Information

Used to obtain the workspace information of ModelArts real-time services. When source is set to custom_from_modelarts_v2 during the creation of a custom endpoint, the inference service information is obtained.

Obtaining the Latest Content Guard Disclaimer

Used to obtain the content guard disclaimer of MaaS real-time inference. If moderation is set to false when you create an endpoint, you must sign the disclaimer of the latest version.

Obtaining Total Statistics

Used to query the aggregated call data of real-time inference services, including: total calls, total failed calls, total tokens, input tokens, output tokens, etc. Data is retained for 30 days only.

Obtaining the Service Statistics List

Used to retrieve statistics on two types of services: subscribed built-in services and created endpoints. It displays metrics for each service, such as call count, failure rate, total tokens, input tokens, output tokens, and E2E latency. Data is retained for 30 days only.

Obtaining Time-based Service Metric Statistics

Used to retrieve granular metric data for a service. This allows viewing chronological trends for metrics such as call count, failure rate, token volume, input token size, output token size, E2E latency, TPM, RPM, QPS, and average generation time. Data is retained for 30 days only.

Obtaining Service Error Details

Used to retrieve detailed error data for services to review failure information, such as error codes, occurrences, and error messages. Data is retained for 30 days only.

Obtaining the Service List

Used to retrieve the service name corresponding to a specific service ID.

Querying the IP Address List

Used to retrieve IP addresses.

Querying Resource Monitoring Metric Details

Used to query the resource monitoring metrics for the "My Services" category within the MaaS real-time inference module. Data is retained for 30 days only.

Querying the Calling Data of a Service Version

Used to query all versions of a service and their corresponding monitoring metric data. Data is retained for 30 days only.

Obtaining the Metrics Supported by Different Model Types

Used to retrieve the list of supported metrics corresponding to a specific service model type.

Obtaining Time-based Service Error Code Statistics

Used to display the chronological distribution of service error code statistics. Data is retained for 30 days only.