@robota-sdk/team

Multi-agent collaboration system for the Robota SDK with advanced workflow management.

🚀 Why Choose @robota-sdk/team?

🎯 Multi-Agent Collaboration

• Task Coordination: Primary agent coordinates with temporary specialist agents for complex tasks
• Specialist Templates: Pre-configured templates for different types of work (research, creative, analysis, etc.)
• Dynamic Agent Creation: Creates temporary expert agents tailored for specific subtasks

💡 Use Cases

• Complex Analysis: Break down multi-faceted analysis tasks across specialist agents
• Content Creation: Coordinate research, writing, and review phases
• Multi-Step Workflows: Handle workflows that benefit from different types of expertise

🏆 Key Features

• Template-Based Specialists: 7 pre-configured specialist templates ready to use
• Cross-Provider Teams: Mix OpenAI, Anthropic, and Google AI providers within teams
• Resource Management: Automatic cleanup and lifecycle management of temporary agents
• Team Analytics: Performance monitoring through ExecutionAnalyticsPlugin

Overview

The @robota-sdk/team package enables multi-agent collaboration with template-based specialist coordination. It provides a system where a primary agent can delegate subtasks to temporary specialist agents, using pre-configured templates for different types of work.

Installation

npm install @robota-sdk/team @robota-sdk/agents

Quick Start

import { createTeam } from '@robota-sdk/team';
import { OpenAIProvider } from '@robota-sdk/openai';
import { AnthropicProvider } from '@robota-sdk/anthropic';

// Create AI providers
const openaiProvider = new OpenAIProvider({ apiKey: process.env.OPENAI_API_KEY });
const anthropicProvider = new AnthropicProvider({ apiKey: process.env.ANTHROPIC_API_KEY });

// Create team with template-based expert selection
const team = createTeam({
  aiProviders: [openaiProvider, anthropicProvider]
});

// Execute complex task - automatic expert delegation
const result = await team.execute(`
Analyze the market for sustainable energy solutions and develop 3 innovative product concepts.
Include market research, competitive analysis, and detailed product specifications.
`);

console.log(result);

Features

🧠 Intelligent Template System

• Automatic Expert Selection: AI analyzes requests and selects optimal specialist templates
• Built-in Expert Templates: 6 specialized templates (task_coordinator, domain_researcher, creative_ideator, summarizer, ethical_reviewer, fast_executor)
• Optimized AI Providers: Each template uses optimal AI model and settings for its specialty
• Natural Language Interface: Simply describe your needs - the system handles expert delegation

🤝 Advanced Collaboration

• Task Decomposition: Automatically breaks complex requests into specialist tasks
• Context Preservation: Ensures delegated tasks maintain proper context and requirements
• Dynamic Agent Creation: Creates temporary expert agents from templates as needed
• Result Synthesis: Combines specialist outputs into comprehensive responses

📊 Workflow Analytics & Monitoring

• Workflow Visualization: Generate flowcharts and relationship diagrams
• Performance Metrics: Track execution time, success rates, and resource usage
• Template Usage Statistics: Monitor which experts are used most frequently
• Cost Optimization: Track costs across different AI providers and models

🔧 Template Management

• Custom Templates: Add your own expert templates with specialized prompts
• Template Registry: Centralized management of all available expert templates
• Provider Configuration: Each template specifies optimal AI provider and settings
• Version Control: Template versioning and update management

📋 Built-in Expert Templates

1. Task Coordinator 🎯

• Role: Analyzes complex requests and delegates to specialists
• Best For: Project management, workflow orchestration
• Model: GPT-4 (high reasoning capability)

2. Domain Researcher 🔍

• Role: Deep research and comprehensive analysis
• Best For: Market research, technical investigation
• Model: Claude 3.5 Sonnet (large context window)

3. Creative Ideator 💡

• Role: Brainstorming and innovative solutions
• Best For: Product concepts, marketing ideas
• Model: GPT-4 (creative capabilities)

4. Fast Executor ⚡

• Role: Quick implementation and coding
• Best For: Code generation, rapid prototyping
• Model: GPT-3.5-turbo (fast and efficient)

5. Ethical Reviewer 🛡️

• Role: Evaluates ethical implications and risks
• Best For: Policy review, compliance checking
• Model: Claude 3 (ethical reasoning)

6. Summarizer 📝

• Role: Condenses and synthesizes information
• Best For: Report generation, executive summaries
• Model: Gemini 1.5 Flash (fast processing)

API Reference

TeamContainer

Main class for managing multi-agent teams.

class TeamContainer {
  constructor(config: TeamConfig)
  
  // Workflow execution
  execute(input: string): Promise<string>
  
  // Team management
  addAgent(name: string, agent: AgentInterface): void
  removeAgent(name: string): void
  getAgent(name: string): AgentInterface | undefined
  
  // Analytics
  getStats(): TeamStats
}

Team Configuration

interface TeamConfig {
  name: string;
  agents: Record<string, AgentInterface>;
  workflow?: {
    description: string;
    steps: WorkflowStep[];
  };
  options?: {
    maxConcurrent?: number;
    timeout?: number;
    retryPolicy?: RetryPolicy;
  };
}

Workflow System

interface WorkflowStep {
  agent: string;
  task: string;
  dependsOn?: string[];
  condition?: (context: WorkflowContext) => boolean;
  timeout?: number;
}

interface WorkflowResult {
  success: boolean;
  results: Record<string, string>;
  metadata: {
    totalTime: number;
    agentsUsed: string[];
    stepResults: StepResult[];
  };
}

Architecture

Module Structure

packages/team/src/
├── team-container.ts       # Main TeamContainer class
├── create-team.ts          # Team creation utilities
├── workflow-formatter.ts   # Workflow result formatting
├── types.ts               # TypeScript definitions
└── index.ts               # Public exports

Integration Points

• Agent System: Works with any agent implementing AgentInterface
• Plugin System: Inherits all agent plugin capabilities
• Analytics: Aggregates metrics from all team members
• Error Handling: Unified error management across team operations

Workflow Patterns

Sequential Workflow

Agents execute tasks in a defined order:

const team = createTeam({
  name: 'Sequential Team',
  workflow: {
    steps: [
      { agent: 'researcher', task: 'Gather information' },
      { agent: 'analyst', task: 'Analyze data' },
      { agent: 'reporter', task: 'Generate report' }
    ]
  }
});

Parallel Workflow

Agents execute tasks concurrently:

const team = createTeam({
  name: 'Parallel Team',
  workflow: {
    steps: [
      { agent: 'agent1', task: 'Task A' },
      { agent: 'agent2', task: 'Task B' },
      { agent: 'agent3', task: 'Task C' }
    ]
  },
  options: { maxConcurrent: 3 }
});

Conditional Workflow

Agents execute based on conditions:

const team = createTeam({
  name: 'Conditional Team',
  workflow: {
    steps: [
      { agent: 'classifier', task: 'Classify input' },
      { 
        agent: 'specialist1', 
        task: 'Handle type A',
        condition: (ctx) => ctx.classification === 'typeA'
      },
      { 
        agent: 'specialist2', 
        task: 'Handle type B',
        condition: (ctx) => ctx.classification === 'typeB'
      }
    ]
  }
});

Team Templates

Pre-configured team setups for common scenarios:

Research Team

• Researcher: Gathers information and sources
• Analyst: Processes and analyzes data
• Writer: Creates final documentation

Development Team

• Architect: Designs system architecture
• Developer: Writes code implementations
• Tester: Creates tests and validates functionality
• Reviewer: Reviews and improves code quality

Content Team

• Content Strategist: Plans content strategy
• Writer: Creates written content
• Editor: Reviews and refines content
• SEO Specialist: Optimizes for search engines

Development

Building

npm run build

Testing

npm run test

Linting

npm run lint
npm run lint:fix

Examples

• Basic Team Setup
• Sequential Workflows
• Parallel Processing
• Conditional Logic
• Team Analytics

Best Practices

1. Agent Specialization: Design agents with specific roles and capabilities
2. Error Handling: Implement robust error recovery in workflows
3. Resource Management: Monitor and optimize agent resource usage
4. Task Granularity: Break down complex tasks into manageable steps
5. Performance Monitoring: Regularly analyze team performance metrics

License

MIT