Implementation:Mlc ai Web llm Chat Completion Tool

Overview

Chat_Completion_Tool implements the Mlc_ai_Web_llm_Tool_Definition principle by providing the TypeScript type definitions for declaring callable functions as tools in the web-llm chat completion API. These types directly mirror the OpenAI function calling specification.

Source Reference

• File: src/openai_api_protocols/chat_completion.ts, Lines 801-835
• Repository: mlc-ai/web-llm

Code Reference

The implementation consists of three type definitions that form a hierarchy:

/**
 * The parameters the functions accepts, described as a JSON Schema object.
 * Omitting `parameters` defines a function with an empty parameter list.
 */
export type FunctionParameters = Record<string, unknown>;

export interface FunctionDefinition {
  /**
   * The name of the function to be called. Must be a-z, A-Z, 0-9, or contain
   * underscores and dashes, with a maximum length of 64.
   */
  name: string;

  /**
   * A description of what the function does, used by the model to choose when and
   * how to call the function.
   */
  description?: string;

  /**
   * The parameters the functions accepts, described as a JSON Schema object.
   * Omitting `parameters` defines a function with an empty parameter list.
   */
  parameters?: FunctionParameters;
}

export interface ChatCompletionTool {
  function: FunctionDefinition;

  /**
   * The type of the tool. Currently, only `function` is supported.
   */
  type: "function";
}

I/O Contract

Type hierarchy:

ChatCompletionTool wraps a FunctionDefinition and tags it with a type discriminator.

Type	Purpose	Fields
FunctionParameters	JSON Schema for function parameters	Record<string, unknown> (any valid JSON Schema object)
FunctionDefinition	Describes a single callable function	name (required), description (optional), parameters (optional)
ChatCompletionTool	Wraps a function definition as a tool	function (required), type: "function" (required, literal)

Integration point:

ChatCompletionTool objects are passed in the tools field of ChatCompletionRequest (line 235):

// From ChatCompletionRequest (line 235)
tools?: Array<ChatCompletionTool>;

Import:

import { ChatCompletionTool } from "@mlc-ai/web-llm";

Usage Examples

Basic tool definition:

import * as webllm from "@mlc-ai/web-llm";

const weatherTool: webllm.ChatCompletionTool = {
  type: "function",
  function: {
    name: "get_current_weather",
    description: "Get the current weather in a given location",
    parameters: {
      type: "object",
      properties: {
        location: {
          type: "string",
          description: "The city and state, e.g. San Francisco, CA",
        },
        unit: {
          type: "string",
          enum: ["celsius", "fahrenheit"],
        },
      },
      required: ["location"],
    },
  },
};

Multiple tools in a request:

const tools: Array<webllm.ChatCompletionTool> = [
  {
    type: "function",
    function: {
      name: "get_stock_fundamentals",
      description: "Get fundamental data for a given stock symbol",
      parameters: {
        type: "object",
        properties: {
          symbol: {
            type: "string",
            description: "The stock ticker symbol, e.g. TSLA",
          },
        },
        required: ["symbol"],
      },
    },
  },
  {
    type: "function",
    function: {
      name: "get_stock_price",
      description: "Get the current price for a stock symbol",
      parameters: {
        type: "object",
        properties: {
          symbol: { type: "string" },
        },
        required: ["symbol"],
      },
    },
  },
];

Parameterless tool:

const tool: webllm.ChatCompletionTool = {
  type: "function",
  function: {
    name: "get_server_time",
    description: "Returns the current server time in UTC",
    // No parameters field -- empty parameter list
  },
};

Related Pages

• Principle:Mlc_ai_Web_llm_Tool_Definition
• Mlc_ai_Web_llm_Function_Calling_Model_Ids -- Model identifiers that support these tool definitions
• Mlc_ai_Web_llm_Tool_Choice_Request -- Controlling which tool is called
• Mlc_ai_Web_llm_Get_Tool_Call_From_Output -- Parsing model output into tool call objects