AI SDK Kotlin/ai.torad.aisdk/StepResult

StepResult

data class StepResult(val stepNumber: Int, val text: String, val reasoning: String, val toolCalls: List<ContentPart.ToolCall>, val toolResults: List<ContentPart.ToolResult>, val toolApprovalRequests: List<ContentPart.ToolApprovalRequest>, val finishReason: FinishReason, val usage: Usage, val warnings: List<CallWarning> = emptyList(), val request: LanguageModelRequestMetadata = LanguageModelRequestMetadata(), val response: LanguageModelResponseMetadata = LanguageModelResponseMetadata(), val providerMetadata: Map<String, JsonElement> = emptyMap(), val rawFinishReason: String? = null, val model: String? = null, val experimentalContext: Any? = null)

Snapshot of one completed loop step — surfaced to OnStepFinishEvent and accumulated in the loop state for PrepareStepScope.steps.

Constructors

constructor(stepNumber: Int, text: String, reasoning: String, toolCalls: List<ContentPart.ToolCall>, toolResults: List<ContentPart.ToolResult>, toolApprovalRequests: List<ContentPart.ToolApprovalRequest>, finishReason: FinishReason, usage: Usage, warnings: List<CallWarning> = emptyList(), request: LanguageModelRequestMetadata = LanguageModelRequestMetadata(), response: LanguageModelResponseMetadata = LanguageModelResponseMetadata(), providerMetadata: Map<String, JsonElement> = emptyMap(), rawFinishReason: String? = null, model: String? = null, experimentalContext: Any? = null)

Properties

dynamicToolCalls

val dynamicToolCalls: List<ContentPart.ToolCall>

dynamicToolResults

val dynamicToolResults: List<ContentPart.ToolResult>

experimentalContext

val experimentalContext: Any?

The evolving agent context live for this step (upstream's experimental_context).

val finishReason: FinishReason

val model: String?

The model id that produced this step (upstream's model.modelId).

providerMetadata

val providerMetadata: Map<String, JsonElement>

rawFinishReason

val rawFinishReason: String?

val reasoning: String

val reasoningText: String?

The non-empty reasoning text for this step, or null — mirrors upstream's reasoningText.

val request: LanguageModelRequestMetadata

val response: LanguageModelResponseMetadata

staticToolCalls

val staticToolCalls: List<ContentPart.ToolCall>

Tool calls/results split by static (tool(...)) vs dynamic (dynamicTool(...)), per upstream.

staticToolResults

val staticToolResults: List<ContentPart.ToolResult>

val stepNumber: Int

val text: String

toolApprovalRequests

val toolApprovalRequests: List<ContentPart.ToolApprovalRequest>

val toolCalls: List<ContentPart.ToolCall>

val toolResults: List<ContentPart.ToolResult>

val usage: Usage

val warnings: List<CallWarning>