Class AiSdkLlm

AI SDK integration that accepts a pre-configured LanguageModel. Enables ADK to work with any provider supported by Vercel's AI SDK.

Hierarchy (view full)

BaseLlm
- AiSdkLlm

Index

Constructors

Properties

Methods

supportedModels generateContentAsyncImpl generateContentAsync maybeAppendUserContent connect

Constructors

constructor

new AiSdkLlm(modelInstance): AiSdkLlm
Constructor accepts a pre-configured LanguageModel instance
Parameters
- modelInstance: LanguageModel
Returns AiSdkLlm
Overrides BaseLlm.constructor
- Defined in packages/adk/src/models/ai-sdk.ts:28

Properties

`Protected`logger

logger: Logger = ...

model

model: string

The name of the LLM, e.g. gemini-2.5-flash or gemini-2.5-flash-001.

Methods

`Static`supportedModels

supportedModels(): string[]
Returns empty array - following Python ADK pattern

Returns string[]
Overrides BaseLlm.supportedModels
- Defined in packages/adk/src/models/ai-sdk.ts:40

`Protected`generateContentAsyncImpl

generateContentAsyncImpl(request, stream?): AsyncGenerator<LlmResponse, void, unknown>
Implementation method to be overridden by subclasses. This replaces the abstract generateContentAsync method.
Parameters
- request: LlmRequest
- stream: boolean = false
Returns AsyncGenerator<LlmResponse, void, unknown>
Overrides BaseLlm.generateContentAsyncImpl
- Defined in packages/adk/src/models/ai-sdk.ts:44

generateContentAsync

generateContentAsync(llmRequest, stream?): AsyncGenerator<LlmResponse, void, unknown>
Generates one content from the given contents and tools.
Parameters
- llmRequest: LlmRequest
  LlmRequest, the request to send to the LLM.
- Optionalstream: boolean
  bool = false, whether to do streaming call.
Returns AsyncGenerator<LlmResponse, void, unknown>
a generator of LlmResponse.

For non-streaming call, it will only yield one LlmResponse.

For streaming call, it may yield more than one response, but all yielded responses should be treated as one response by merging the parts list.
Inherited from BaseLlm.generateContentAsync
- Defined in packages/adk/src/models/base-llm.ts:46

`Protected`maybeAppendUserContent

maybeAppendUserContent(llmRequest): void
Appends a user content, so that model can continue to output.
Parameters
- llmRequest: LlmRequest
  LlmRequest, the request to send to the LLM.
Returns void
Inherited from BaseLlm.maybeAppendUserContent
- Defined in packages/adk/src/models/base-llm.ts:216

connect

connect(_llmRequest): BaseLLMConnection
Creates a live connection to the LLM.
Parameters
- _llmRequest: LlmRequest
Returns BaseLLMConnection
BaseLLMConnection, the connection to the LLM.
Inherited from BaseLlm.connect
- Defined in packages/adk/src/models/base-llm.ts:251