OpenAIModelOptions interface

  • Reference
Package:
@microsoft/teams-ai

Options for configuring an OpenAIModel to call an OpenAI hosted model.

Extends
BaseOpenAIModelOptions

Properties

apiKey

API key to use when calling the OpenAI API.
defaultModel

Default model to use for completions.
endpoint

Optional. Endpoint to use when calling the OpenAI API.
organization

Optional. Organization to use when calling the OpenAI API.
project

Optional. Project to use when calling the OpenAI API.

Inherited Properties

clientOptions

Optional. Custom client options to use when calling the OpenAI API.
logRequests

Optional. Whether to log requests to the console.
maxRetries

Optional. Maximum number of retries to use when calling the OpenAI API.
requestConfig
responseFormat

Optional. Forces the model return a specific response format.
retryPolicy
seed

Optional. A static seed to use when making model calls.
stream

Optional. Whether the models responses should be streamed back using Server Sent Events (SSE.)
useSystemMessages

Optional. Whether to use system messages when calling the OpenAI API.

Property Details

apiKey

API key to use when calling the OpenAI API.

apiKey: string

Property Value

string

Remarks

A new API key can be created at https://platform.openai.com/account/api-keys.

defaultModel

Default model to use for completions.

defaultModel: string

Property Value

string

endpoint

Optional. Endpoint to use when calling the OpenAI API.

endpoint?: string

Property Value

string

Remarks

For Azure OpenAI this is the deployment endpoint.

organization

Optional. Organization to use when calling the OpenAI API.

organization?: string

Property Value

string

project

Optional. Project to use when calling the OpenAI API.

project?: string

Property Value

string

Inherited Property Details

clientOptions

Optional. Custom client options to use when calling the OpenAI API.

clientOptions?: ClientOptions

Property Value

ClientOptions

Inherited From BaseOpenAIModelOptions.clientOptions

logRequests

Optional. Whether to log requests to the console.

logRequests?: boolean

Property Value

boolean

Remarks

This is useful for debugging prompts and defaults to false.

Inherited From BaseOpenAIModelOptions.logRequests

maxRetries

Optional. Maximum number of retries to use when calling the OpenAI API.

maxRetries?: number

Property Value

number

Remarks

The default is to retry twice.

Inherited From BaseOpenAIModelOptions.maxRetries

requestConfig

Warning

This API is now deprecated.

Optional. Request options to use when calling the OpenAI API.

requestConfig?: AxiosRequestConfig<any>

Property Value

AxiosRequestConfig<any>

Inherited From BaseOpenAIModelOptions.requestConfig

responseFormat

Optional. Forces the model return a specific response format.

responseFormat?: { type: "json_object" }

Property Value

{ type: "json_object" }

Remarks

This can be used to force the model to always return a valid JSON object.

Inherited From BaseOpenAIModelOptions.responseFormat

retryPolicy

Warning

This API is now deprecated.

Optional. Retry policy to use when calling the OpenAI API.

retryPolicy?: number[]

Property Value

number[]

Remarks

Use maxRetries instead.

Inherited From BaseOpenAIModelOptions.retryPolicy

seed

Optional. A static seed to use when making model calls.

seed?: number

Property Value

number

Remarks

The default is to use a random seed. Specifying a seed will make the model deterministic.

Inherited From BaseOpenAIModelOptions.seed

stream

Optional. Whether the models responses should be streamed back using Server Sent Events (SSE.)

stream?: boolean

Property Value

boolean

Remarks

Defaults to false.

Inherited From BaseOpenAIModelOptions.stream

useSystemMessages

Optional. Whether to use system messages when calling the OpenAI API.

useSystemMessages?: boolean

Property Value

boolean

Remarks

The current generation of models tend to follow instructions from user messages better then system messages so the default is false, which causes any system message in the prompt to be sent as user messages instead.

Inherited From BaseOpenAIModelOptions.useSystemMessages