AI SDK Kotlin/ai.torad.aisdk/TextGenerator

TextGenerator

class TextGenerator @JvmOverloads constructor(model: LanguageModel, config: CallConfig = CallConfig())(source)

High-level text generation facade for a single LanguageModel.

TextGenerator turns GenerationInput plus optional Output decoding into cold Flow-based calls. Use generate for one-shot responses and collect the returned flow with .first(). Use stream or streamResult when the caller wants incremental StreamEvent values or stream metadata.

Streaming has no default deadline. Set CallConfig.timeout or configure the underlying HTTP/on-device engine socket timeout so a silent SSE or local engine stream cannot wait forever.

Since

0.3.0-beta01

Constructors

TextGenerator

@JvmOverloads

constructor(model: LanguageModel, config: CallConfig = CallConfig())

Functions

generate

fun generate(input: GenerationInput): Flow<GenerateTextResult<String>>

Cold flow: no model call starts until collected. Each collection reruns the generation from the beginning, so provider calls may incur cost or billing again. One-shot callers should collect exactly one value, usually with .first().

fun <T> generate(input: GenerationInput, output: Output<T>): Flow<GenerateTextResult<T>>

Cold flow: no model call starts until collected. Each collection reruns the generation from the beginning, including structured-output decoding, so provider calls may incur cost or billing again. One-shot callers should collect exactly one value, usually with .first().

stream

fun stream(input: GenerationInput): Flow<StreamEvent>

Starts a text stream when collected.

streamResult

fun streamResult(input: GenerationInput): StreamTextResult

Creates a replayable stream result with request/response metadata.