DeepInfraProvider

Namespace: Vercel.AI.DeepInfra

public sealed class DeepInfraProvider : OpenAICompatibleProvider

DeepInfra provider. OpenAI Chat Completions compatible at https://api.deepinfra.com/v1/openai.

Inheritance: object → ProviderBase → OpenAICompatibleProvider

View source

Constructors

DeepInfraProvider(HttpClient, OpenAICompatibleOptions?)

public DeepInfraProvider(HttpClient httpClient, OpenAICompatibleOptions? options = null)

Creates a provider.

Parameters

Name	Type	Description
httpClient	HttpClient
options	OpenAICompatibleOptions

View source

Fields

ApiKeyVariable

public const string ApiKeyVariable = "DEEPINFRA_API_KEY"

Environment variable for the API key.

Value: string

View source

DefaultBaseUrl

public const string DefaultBaseUrl = "https://api.deepinfra.com/v1/openai"

Default API origin.

Value: string

View source

ProviderId

public const string ProviderId = "deepinfra"

Provider id.

Value: string

View source

Methods

Create(OpenAICompatibleOptions?, HttpMessageHandler?)

public static DeepInfraProvider Create(OpenAICompatibleOptions? options = null, HttpMessageHandler? handler = null)

Creates a provider. Pass a handler from tests.

Parameters

Name	Type	Description
options	OpenAICompatibleOptions
handler	HttpMessageHandler

Returns: DeepInfraProvider

View source

FixReasoningUsage(JsonElement)

public static JsonElement FixReasoningUsage(JsonElement usage)

DeepInfra leaves reasoning tokens out of completion_tokens for some models (for example Gemini and Gemma). When reasoning exceeds completion tokens, the reasoning tokens are added to the completion and total counts.

Parameters

Name	Type	Description
usage	JsonElement

Returns: JsonElement

View source

ImageModel(string)

public override IImageModel ImageModel(string modelId)

Returns an image model.

Parameters

Name	Type	Description
modelId	string

Returns: IImageModel

View source