LangGraph - Agents¶

This section documents the nine specialized nodes in the LangGraph chatbot system.

Overview¶

The system uses specialized agents, each responsible for handling specific types of user requests. Agents are orchestrated by a supervisor that analyzes user intent and routes to the appropriate agent.

Available Agents¶

• Career Chat Agent - Career management and objectives
• Opportunity Agents - Opportunity search and suggestions
• Task Helper Agent - AI assistant for professional tasks
• Person Disambiguation Agent - Ambiguous person resolution
• Talent Finder Agent - Internal talent discovery
• Training Finder Agent - Training search with personalized recommendations
• Profile Analysis Agent - Visual profile assessment and action plans

Architecture¶

Each agent is created with LangChain's create_agent on a LangGraph runtime and has:

• LLM profile - Tenant-configured mini or large, with a stable task label
• Specific Tools - Tailored to the agent's domain
• System Prompt - Defines behavior and capabilities
• Middleware - For chat history management and dynamic prompts

Agent Workflow¶

User Message
    ↓
Supervisor Agent
    ↓ (analyzes intent)
Specialized Agent
    ↓ (uses tools)
Response to User

Usage¶

Agents are orchestrated by the supervisor in src/chatbot/agentic/graphs.py. The supervisor maintains conversation state and routes through shared handoff tools.