LlmCallParams

Parameters accepted by `kernel.callLLM()`.

SDK v2026.4.0

Kind: interface

Signature

Example

export interface LlmCallParams

prompt (string): Prompt body sent to the provider.
model (string): Model identifier requested for the call.
systemPrompt? (string): Optional system prompt prepended by the caller.
temperature? (number): Optional sampling temperature.
maxTokens? (number): Optional maximum completion token budget.
reasoningEffort? (LlmReasoningEffort): Optional reasoning effort hint for reasoning-capable models.
stopSequences? (string[]): Optional stop sequences forwarded to the provider.
dlpPolicy? (DlpScanPolicyConfig): Optional per-call DLP override applied by the runtime and proxy.

kernel

Public kernel proxy used inside agent steps for LLM calls, mediated tools, data access, and deterministic runtime utilities.

LlmReasoningEffort

Reasoning effort hint for reasoning-capable LLM models.