mojentic-core/com.mojentic.llm/LlmBroker

LlmBroker

class LlmBroker(gateway: LlmGateway, tracer: Tracer = NullTracer, toolRunner: ToolRunner = SerialToolRunner(), json: Json = DEFAULT_JSON)

Coordinator above an LlmGateway that handles tool execution and provides a Kotlin-idiomatic chat surface.

The broker is the primary integration point for application code. It:

Forwards single-shot, streaming, and structured-output requests to the gateway.
Dispatches tool calls through a ToolRunner (serial by default) and recursively re-prompts the LLM with the tool results.
Honours CompletionConfig.maxToolIterations as a hard recursion ceiling.
Records observability hooks to a Tracer (no-op by default).

The broker is stateless from the caller's point of view: each call carries its own message list. Use ChatSession (Phase 2) when you want managed conversation state.

Constructors

LlmBroker

constructor(gateway: LlmGateway, tracer: Tracer = NullTracer, toolRunner: ToolRunner = SerialToolRunner(), json: Json = DEFAULT_JSON)

Types

Companion

object Companion

Functions

complete

suspend fun complete(model: String, messages: List<LlmMessage>, tools: List<LlmTool> = emptyList(), config: CompletionConfig = CompletionConfig(), correlationId: String? = null): LlmGatewayResponse

completeJson

inline suspend fun <T> completeJson(model: String, messages: List<LlmMessage>, config: CompletionConfig = CompletionConfig(), correlationId: String? = null): T

generateResponse

suspend fun generateResponse(model: String, messages: List<LlmMessage>, tools: List<LlmTool> = emptyList(), config: CompletionConfig = CompletionConfig(), correlationId: String? = null): LlmGatewayResponse

Return one native response without dispatching tools or changing context.

generateStreamEvents

fun generateStreamEvents(model: String, messages: List<LlmMessage>, config: CompletionConfig = CompletionConfig(), correlationId: String? = null): Flow<CompletionStreamEvent>

Stream one turn as CompletionStreamEvents with terminal completion evidence.

stream

fun stream(model: String, messages: List<LlmMessage>, tools: List<LlmTool> = emptyList(), config: CompletionConfig = CompletionConfig(), correlationId: String? = null): Flow<StreamEvent>