Skip to content

LangGraph - Agents

This page documents the seven supervisor destinations and the nested Knowledge Base specialist. Each specialist is defined in its owning agent module.

Available Agents

The system uses several specialized agents:

Career Agent

  • File: src/chatbot/agentic/agents/career_chat_agent.py
  • Role: Career management and objectives
  • Description: Career chat agent responsible for managing career-related conversations, including objectives and job mapping

Knowledge Base Agent

  • File: src/chatbot/agentic/agents/knowledge_base_agent.py
  • Role: Grounded company-knowledge research
  • Description: A Deep Agent reached by a Career Chat handoff; retrieves RAG and configured public pages concurrently, compares evidence and synthesizes once.

Opportunity Agents

  • File: src/chatbot/agentic/agents/opportunity_agents.py
  • Role: Opportunity search and suggestions
  • Description: Searches for professional opportunities with advanced filters and personalized suggestions

Task Helper Agent

  • File: src/chatbot/agentic/agents/task_helper_agent.py
  • Role: AI assistant for professional tasks
  • Description: AI-powered contextual responses for professional questions

Person Disambiguation Agent

  • File: src/chatbot/agentic/agents/person_disambiguation_agent.py
  • Role: Ambiguous person resolution
  • Description: Resolves ambiguous person references when multiple matches are found in user queries

Talent Finder Agent

  • File: src/chatbot/agentic/agents/talent_finder_agents.py
  • Role: Internal talent discovery
  • Description: Searches and discovers internal talent based on skills, experience, and other criteria

Training Finder Agent

  • File: src/chatbot/agentic/agents/training_finder_agents.py
  • Role: Training search and personalized recommendations
  • Description: Searches for trainings through vector and full-text retrieval, preserves weighted RRF order, and generates collection titles, subtitles and optional plan phases

Profile Analysis Agent

  • File: src/chatbot/agentic/agents/profile_analysis_agent.py
  • Role: Visual profile assessment
  • Description: Presents strengths, development priorities and profile action plans

See the agent catalogue for the full node inventory.

Architecture

Conversational agents use LangChain's create_agent; Knowledge Base uses create_deep_agent. Both run on LangGraph with: - A tenant-configured mini or large LLM profile - Specific tools - A system or dynamic prompt

Usage

Agents are orchestrated by the supervisor in src/chatbot/agentic/graphs.py. The pre-supervisor gate evaluates safety only. Career Chat can transfer to its Knowledge Base Deep Agent; supported answers are delivered verbatim without another agent-model call, while misses return to Career Chat for a contextual follow-up. The guardrail separates a mini Decisions classification from conditional mini boundary-response generation; the supervisor uses a separate typed choice to route to an enabled specialist. Both classifications use Decisions when the configured OpenAI mini supports it, and ordinary structured generation with the configured provider otherwise.