LLM.Stream.ToolCall (LLM v0.1.2)

A tool call chunk from a streaming response.

Emitted when the model decides to call a tool. During streaming, tool call arguments arrive incrementally. The library accumulates them internally and only emits a chunk when complete: true.

Fields

:id — unique identifier for this tool call (used in tool result messages)
:name — the tool name to call
:arguments — decoded arguments map (complete when complete: true)
:index — position index for multiple simultaneous tool calls
:complete — whether this chunk contains the full tool call (always true when emitted)

Example

%LLM.Stream.ToolCall{
  id: "call_abc123",
  name: "get_weather",
  arguments: %{"location" => "San Francisco"},
  index: 0,
  complete: true
}

Summary

Types

t()

Types

t()

@type t() :: %LLM.Stream.ToolCall{
  arguments: map(),
  complete: boolean(),
  id: String.t(),
  index: non_neg_integer(),
  name: String.t()
}