wehub-resource-sync
71 lines
2.5 KiB
TypeScript
71 lines
2.5 KiB
TypeScript
/**
|
|
* Example 08: Multi-Agent System
|
|
*
|
|
* This example demonstrates a sophisticated multi-agent orchestration system.
|
|
* It shows how to:
|
|
* - Implement multiple specialized agents (HR, IT, Sales)
|
|
* - Use an orchestrator to coordinate agent interactions
|
|
* - Handle agent handoffs and context switching
|
|
* - Maintain conversation flow across different agent specializations
|
|
* - Build complex agent hierarchies and routing
|
|
*
|
|
* Key concepts:
|
|
* - Multi-agent orchestration patterns
|
|
* - Agent specialization and routing
|
|
* - Context spreading with spread operator
|
|
* - Handoff detection and management
|
|
* - Conversational state management across agents
|
|
*/
|
|
|
|
import { Client } from '@botpress/client'
|
|
import { execute } from 'llmz'
|
|
|
|
import { CLIChat } from '../utils/cli-chat'
|
|
|
|
// Import specialized agent definitions
|
|
import { HRAgent } from './agent_hr' // Human Resources specialist
|
|
import { ITAgent } from './agent_it' // Information Technology specialist
|
|
import { MainAgent } from './agent_main' // Entry point/coordinator agent
|
|
import { SalesAgent } from './agent_sales' // Sales and marketing specialist
|
|
|
|
// Import the orchestration system
|
|
import { MultiAgentOrchestrator } from './orchestrator'
|
|
|
|
// Initialize Botpress client for LLM communication
|
|
const client = new Client({
|
|
botId: process.env.BOTPRESS_BOT_ID!,
|
|
token: process.env.BOTPRESS_TOKEN!,
|
|
})
|
|
|
|
// Set up chat interface and orchestrator
|
|
const chat = new CLIChat()
|
|
|
|
// Create orchestrator with all agents and set MainAgent as entry point
|
|
// The orchestrator manages agent selection, context, and handoffs
|
|
const orchestrator = new MultiAgentOrchestrator(
|
|
[MainAgent, HRAgent, ITAgent, SalesAgent], // Available agents
|
|
MainAgent.name // Initial/default agent
|
|
)
|
|
|
|
// Main execution loop with agent orchestration
|
|
while (true) {
|
|
// Execute with the current agent's context
|
|
// The orchestrator provides the appropriate instructions, tools, and configuration
|
|
const result = await execute({
|
|
// Spread the orchestrator's context (instructions, tools, exits, etc.)
|
|
// This dynamically configures the execution based on the current agent
|
|
...orchestrator.context,
|
|
client,
|
|
chat,
|
|
})
|
|
|
|
// Check if the current agent has handed off control to another agent
|
|
if (!orchestrator.hasHandedOff(result)) {
|
|
// No handoff occurred - continue with user input
|
|
// Wait for the next user message to continue the conversation
|
|
await chat.prompt()
|
|
}
|
|
// If handoff occurred, the orchestrator has already switched contexts
|
|
// Continue the loop with the new agent's configuration
|
|
}
|