AiSdkModel

Wraps a model from the AI SDK that adheres to the LanguageModelV1 spec to be used used as a model in the OpenAI Agents SDK to use other models.

While you can use this with the OpenAI models, it is recommended to use the default OpenAI model provider instead.

If tracing is enabled, the model will send generation spans to your traces processor.

import { aisdk } from '@openai/agents-extensions';
import { openai } from '@ai-sdk/openai';

const model = aisdk(openai('gpt-4o'));

const agent = new Agent({
  name: 'My Agent',
  model
});

Param

The Vercel AI SDK model to wrap.

Implements

• Model

Constructors

Constructor

new AiSdkModel(model): AiSdkModel;

Parameters

Parameter	Type
model	LanguageModelV1

Returns

AiSdkModel

Methods

getResponse()

getResponse(request): Promise<{
  output: AgentOutputItem[];
  providerData: {
     files?: object[];
     finishReason: LanguageModelV1FinishReason;
     logprobs?: LanguageModelV1LogProbs;
     providerMetadata?: LanguageModelV1ProviderMetadata;
     rawCall: {
        rawPrompt: unknown;
        rawSettings: Record<string, unknown>;
     };
     rawResponse?: {
        body?: unknown;
        headers?: Record<string, string>;
     };
     reasoning?:   | string
        | (
        | {
        signature?: string;
        text: string;
        type: "text";
      }
        | {
        data: string;
        type: "redacted";
      })[];
     request?: {
        body?: string;
     };
     response?: {
        id?: string;
        modelId?: string;
        timestamp?: Date;
     };
     sources?: LanguageModelV1Source[];
     text?: string;
     toolCalls?: LanguageModelV1FunctionToolCall[];
     usage: {
        completionTokens: number;
        promptTokens: number;
     };
     warnings?: LanguageModelV1CallWarning[];
  };
  responseId: string;
  usage: Usage;
}>;

Get a response from the model.

Parameters

Parameter	Type	Description
request	ModelRequest	The request to get a response for.

Returns

Promise<{ output: AgentOutputItem[]; providerData: { files?: object[]; finishReason: LanguageModelV1FinishReason; logprobs?: LanguageModelV1LogProbs; providerMetadata?: LanguageModelV1ProviderMetadata; rawCall: { rawPrompt: unknown; rawSettings: Record<string, unknown>; }; rawResponse?: { body?: unknown; headers?: Record<string, string>; }; reasoning?: | string | ( | { signature?: string; text: string; type: "text"; } | { data: string; type: "redacted"; })[]; request?: { body?: string; }; response?: { id?: string; modelId?: string; timestamp?: Date; }; sources?: LanguageModelV1Source[]; text?: string; toolCalls?: LanguageModelV1FunctionToolCall[]; usage: { completionTokens: number; promptTokens: number; }; warnings?: LanguageModelV1CallWarning[]; }; responseId: string; usage: Usage; }>

Implementation of

Model.getResponse

---

getStreamedResponse()

getStreamedResponse(request): AsyncIterable<StreamEvent>;

Get a streamed response from the model.

Parameters

Parameter	Type
request	ModelRequest

Returns

AsyncIterable<StreamEvent>

Implementation of

Model.getStreamedResponse