Class OpenAiModel

See also VS Code ILanguageModelChatMetadata

Implements

LanguageModel

Constructors

constructor

new OpenAiModel(
    id: string,
    model: string,
    status: LanguageModelStatus,
    enableStreaming: boolean,
    apiKey: () => undefined | string,
    apiVersion: () => undefined | string,
    supportsStructuredOutput: boolean,
    url: undefined | string,
    deployment: undefined | string,
    openAiModelUtils: OpenAiModelUtils,
    responseApiUtils: OpenAiResponseApiUtils,
    developerMessageSettings?: DeveloperMessageSettings,
    maxRetries?: number,
    useResponseApi?: boolean,
    proxy?: string,
    reasoningSupport?: ReasoningSupport,
    maxInputTokens?: number,
    serverTools?: ServerToolDescriptor[],
    serverSideCompactionSupport?: boolean,
    serverSideCompactionEnabledByDefault?: boolean,
    serverSideCompactionTokenThresholdByDefault?: number,
): OpenAiModel
Parameters
- id: string
  the unique id for this language model. It will be used to identify the model in the UI.
- model: string
  the model id as it is used by the OpenAI API
- status: LanguageModelStatus
- enableStreaming: boolean
  whether the streaming API shall be used
- apiKey: () => undefined | string
  a function that returns the API key to use for this model, called on each request
- apiVersion: () => undefined | string
  a function that returns the OpenAPI version to use for this model, called on each request
- supportsStructuredOutput: boolean
- url: undefined | string
  the OpenAI API compatible endpoint where the model is hosted. If not provided the default OpenAI endpoint will be used.
- deployment: undefined | string
- openAiModelUtils: OpenAiModelUtils
- responseApiUtils: OpenAiResponseApiUtils
- developerMessageSettings: DeveloperMessageSettings = 'developer'
  how to handle system messages
- maxRetries: number = 3
  the maximum number of retry attempts when a request fails
- useResponseApi: boolean = false
  whether to use the newer OpenAI Response API instead of the Chat Completion API
- Optionalproxy: string
- OptionalreasoningSupport: ReasoningSupport
- OptionalmaxInputTokens: number
- OptionalserverTools: ServerToolDescriptor[]
- serverSideCompactionSupport: boolean = false
  whether this model supports server-side compaction (only available via the Response API)
- serverSideCompactionEnabledByDefault: boolean = false
  resolved default enablement of server-side compaction (global preference folded with the per-provider override)
- OptionalserverSideCompactionTokenThresholdByDefault: number
Returns OpenAiModel
- Defined in packages/ai-openai/src/node/openai-language-model.ts:100

Properties

apiKey

apiKey: () => undefined | string

a function that returns the API key to use for this model, called on each request

apiVersion

apiVersion: () => undefined | string

a function that returns the OpenAPI version to use for this model, called on each request

deployment

deployment: undefined | string

developerMessageSettings

developerMessageSettings: DeveloperMessageSettings = 'developer'

how to handle system messages

enableStreaming

enableStreaming: boolean

whether the streaming API shall be used

`Readonly`id

id: string

the unique id for this language model. It will be used to identify the model in the UI.

`Optional`maxInputTokens

maxInputTokens?: number

maxRetries

maxRetries: number = 3

the maximum number of retry attempts when a request fails

model

model: string

the model id as it is used by the OpenAI API

openAiModelUtils

openAiModelUtils: OpenAiModelUtils

`Optional`proxy

proxy?: string

`Optional`reasoningSupport

reasoningSupport?: ReasoningSupport

responseApiUtils

responseApiUtils: OpenAiResponseApiUtils

`Protected`runnerOptions

runnerOptions: RunnerOptions = ...

The options for the OpenAI runner.

serverSideCompactionEnabledByDefault

serverSideCompactionEnabledByDefault: boolean = false

resolved default enablement of server-side compaction (global preference folded with the per-provider override)

serverSideCompactionSupport

serverSideCompactionSupport: boolean = false

whether this model supports server-side compaction (only available via the Response API)

`Optional`serverSideCompactionTokenThresholdByDefault

serverSideCompactionTokenThresholdByDefault?: number

`Optional`serverTools

serverTools?: ServerToolDescriptor[]

Server tools this model offers, declared code-level by the provider package. Note: If you provide these, you must also provide vendor because server tools are vendor-specific.

status

status: LanguageModelStatus

supportsStructuredOutput

supportsStructuredOutput: boolean

url

url: undefined | string

the OpenAI API compatible endpoint where the model is hosted. If not provided the default OpenAI endpoint will be used.

useResponseApi

useResponseApi: boolean = false

whether to use the newer OpenAI Response API instead of the Chat Completion API

`Readonly`vendor

vendor: "openai" = 'openai'

Methods

`Protected`applyResponseApiCompaction

applyResponseApiCompaction(
settings: Record<string, unknown>,
request: LanguageModelRequest,
): Record<string, unknown>
Augments the Response API settings with the server-side compaction directive when compaction is enabled for the given request. When disabled, the settings are returned unchanged so the default path is byte-for-byte identical.
Parameters
- settings: Record<string, unknown>
- request: LanguageModelRequest
Returns Record<string, unknown>
- Defined in packages/ai-openai/src/node/openai-language-model.ts:266

`Protected`createTools

createTools(
request: LanguageModelRequest,
): undefined | RunnableToolFunctionWithoutParse[]
Parameters
- request: LanguageModelRequest
Returns undefined | RunnableToolFunctionWithoutParse[]
- Defined in packages/ai-openai/src/node/openai-language-model.ts:231

`Protected`getSettings

getSettings(
request: LanguageModelRequest,
forResponseApi?: boolean,
): Record<string, unknown>
Reasoning-level translation lives in openAiReasoningFor.
Parameters
- request: LanguageModelRequest
- forResponseApi: boolean = false
Returns Record<string, unknown>
- Defined in packages/ai-openai/src/node/openai-language-model.ts:125

`Protected`handleChatCompletionsRequest

handleChatCompletionsRequest(
    openai: OpenAI,
    request: UserRequest,
    cancellationToken?: CancellationToken,
): Promise<LanguageModelResponse>
Parameters
- openai: OpenAI
- request: UserRequest
- OptionalcancellationToken: CancellationToken
Returns Promise<LanguageModelResponse>
- Defined in packages/ai-openai/src/node/openai-language-model.ts:140

`Protected`handleNonStreamingRequest

handleNonStreamingRequest(
openai: OpenAI,
request: UserRequest,
): Promise<LanguageModelTextResponse>
Parameters
- openai: OpenAI
- request: UserRequest
Returns Promise<LanguageModelTextResponse>
- Defined in packages/ai-openai/src/node/openai-language-model.ts:184

`Protected`handleResponseApiRequest

handleResponseApiRequest(
    openai: OpenAI,
    request: UserRequest,
    cancellationToken?: CancellationToken,
): Promise<LanguageModelResponse>
Parameters
- openai: OpenAI
- request: UserRequest
- OptionalcancellationToken: CancellationToken
Returns Promise<LanguageModelResponse>
- Defined in packages/ai-openai/src/node/openai-language-model.ts:280

`Protected`handleStructuredOutputRequest

handleStructuredOutputRequest(
openai: OpenAI,
request: UserRequest,
): Promise<LanguageModelParsedResponse>
Parameters
- openai: OpenAI
- request: UserRequest
Returns Promise<LanguageModelParsedResponse>
- Defined in packages/ai-openai/src/node/openai-language-model.ts:207

`Protected`initializeOpenAi

initializeOpenAi(): OpenAI
Returns OpenAI
- Defined in packages/ai-openai/src/node/openai-language-model.ts:243

`Protected`isNonStreamingModel

isNonStreamingModel(_model: string): boolean
Parameters
- _model: string
Returns boolean
- Defined in packages/ai-openai/src/node/openai-language-model.ts:203

`Protected`processMessages

processMessages(messages: LanguageModelMessage[]): ChatCompletionMessageParam[]
Parameters
- messages: LanguageModelMessage[]
Returns ChatCompletionMessageParam[]
- Defined in packages/ai-openai/src/node/openai-language-model.ts:307

request

request(
request: UserRequest,
cancellationToken?: CancellationToken,
): Promise<LanguageModelResponse>
Parameters
- request: UserRequest
- OptionalcancellationToken: CancellationToken
Returns Promise<LanguageModelResponse>
Implementation of LanguageModel.request
- Defined in packages/ai-openai/src/node/openai-language-model.ts:132

Class OpenAiModel

Implements

Index

Constructors

Properties

Methods

Constructors

constructor

Parameters

Returns OpenAiModel

Properties

apiKey

apiVersion

deployment

developerMessageSettings

enableStreaming

Readonlyid

OptionalmaxInputTokens

maxRetries

model

openAiModelUtils

Optionalproxy

OptionalreasoningSupport

responseApiUtils

ProtectedrunnerOptions

serverSideCompactionEnabledByDefault

serverSideCompactionSupport

OptionalserverSideCompactionTokenThresholdByDefault

OptionalserverTools

status

supportsStructuredOutput

url

useResponseApi

Readonlyvendor

Methods

ProtectedapplyResponseApiCompaction

Parameters

Returns Record<string, unknown>

ProtectedcreateTools

Parameters

Returns undefined | RunnableToolFunctionWithoutParse[]

ProtectedgetSettings

Parameters

Returns Record<string, unknown>

ProtectedhandleChatCompletionsRequest

Parameters

Returns Promise<LanguageModelResponse>

ProtectedhandleNonStreamingRequest

Parameters

Returns Promise<LanguageModelTextResponse>

ProtectedhandleResponseApiRequest

Parameters

Returns Promise<LanguageModelResponse>

ProtectedhandleStructuredOutputRequest

Parameters

Returns Promise<LanguageModelParsedResponse>

ProtectedinitializeOpenAi

Returns OpenAI

ProtectedisNonStreamingModel

Parameters

Returns boolean

ProtectedprocessMessages

Parameters

Returns ChatCompletionMessageParam[]

request

Parameters

Returns Promise<LanguageModelResponse>

Settings

On This Page

`Readonly`id

`Optional`maxInputTokens

`Optional`proxy

`Optional`reasoningSupport

`Protected`runnerOptions

`Optional`serverSideCompactionTokenThresholdByDefault

`Optional`serverTools

`Readonly`vendor

`Protected`applyResponseApiCompaction

`Protected`createTools

`Protected`getSettings

`Protected`handleChatCompletionsRequest

`Protected`handleNonStreamingRequest

`Protected`handleResponseApiRequest

`Protected`handleStructuredOutputRequest

`Protected`initializeOpenAi

`Protected`isNonStreamingModel

`Protected`processMessages