OfficeLLM

Main API for creating and managing multi-agent systems

import { OfficeLLM } from 'officellm';

Constructor

new OfficeLLM(config: OfficeLLMConfig)

Creates a new officeLLM instance with manager and worker agents.

config · required

Configuration object containing manager and workers

Example

const office = new OfficeLLM({
  manager: managerConfig,
  workers: [workerConfig1, workerConfig2],
});

Methods

executeTask

executeTask(task: Task): Promise<TaskResult>

Execute a task through the manager agent.

task · required

Task object with title, description, and priority

Returns: Promise resolving to task result

callWorker

callWorker(workerName: string, params: Record<string, any>): Promise<TaskResult>

Call a specific worker agent directly.

workerName · required

Name of the worker agent to call

params · required

Parameters to pass to the worker

Returns: Promise resolving to task result

getWorkers

getWorkers(): string[]

Get list of available worker agent names.

Returns: Array of worker names

getManager

getManager(): { name: string; description?: string }

Get manager agent information.

Returns: Object with manager name and description

Types

OfficeLLMConfig

interface OfficeLLMConfig {
  manager: ManagerConfig;
  workers: WorkerConfig[];
}

ManagerConfig

interface ManagerConfig {
  name: string;
  description?: string;
  provider: ProviderConfig;
  systemPrompt: string;
  tools: ToolDefinition[];
}

WorkerConfig

interface WorkerConfig {
  name: string;
  description?: string;
  provider: ProviderConfig;
  systemPrompt: string;
  tools: ToolDefinition[];
}

Task

interface Task {
  title: string;
  description: string;
  priority?: 'low' | 'medium' | 'high';
  [key: string]: any; // Additional parameters
}

TaskResult

interface TaskResult {
  success: boolean;
  content: string;
  toolCalls?: ToolCall[];
  usage?: {
    promptTokens: number;
    completionTokens: number;
    totalTokens: number;
  };
  error?: string;
}

Usage Examples

Basic Task Execution

const office = new OfficeLLM({
  manager: {
    name: 'Project Manager',
    provider: { type: 'openai', apiKey: '...', model: 'gpt-4' },
    systemPrompt: 'You coordinate AI agents...',
    tools: [
      {
        name: 'math_solver',
        description: 'Solve math problems',
        parameters: z.object({
          task: z.string(),
        }),
      },
    ],
  },
  workers: [
    {
      name: 'Math Solver',
      provider: { type: 'anthropic', apiKey: '...', model: 'claude-3-sonnet' },
      systemPrompt: 'You are a math expert...',
      tools: [
        {
          name: 'calculate',
          description: 'Calculate expressions',
          parameters: z.object({
            expression: z.string(),
          }),
        },
      ],
    },
  ],
});

// Execute task
const result = await office.executeTask({
  title: 'Calculate area',
  description: 'What is the area of a circle with radius 5?',
  priority: 'medium',
});

console.log(result.content); // "The area is approximately 78.54 square units..."

Direct Worker Call

// Call worker directly
const workerResult = await office.callWorker('Math Solver', {
  task: 'Solve for x: 2x + 3 = 7',
});

console.log(workerResult.content);

Getting Available Workers

const workers = office.getWorkers();
console.log(workers); // ['Math Solver', 'Research Assistant', ...]

const manager = office.getManager();
console.log(manager.name); // 'Project Manager'