hashbrown

experimental_webllm_tools

Creates application-selected structured tool inference from a loaded text engine. Buffers and validates one JSON argument object before exposing a canonical call; finalization streams the original response schema. The application owns model loading and selects each phase synchronously. This is a structured JSON protocol, independent of native WebLLM tool calling. Runs sharing an engine are exclusive.

API

export declare function experimental_webllm_tools(
  options:  ExperimentalWebLlmToolsTransportOptions ,
):  TransportFactory;
experimental_webllm_tools
@paramoptions:
@returns