How luuisotorres/langgraph-agent-orchestrator works
This repository is a command-line customer-support triage agent. A user enters a query, the LangGraph workflow uses an OpenAI chat model to classify it, conditionally dispatches to billing, technical, or general response logic, and prints the reply. The graph state carries messages and category; an in-memory checkpointer preserves thread context. README-level cyclic conversational behavior is represented by the persistent conversation state rather than an additional routing loop. Source evidence is limited to main.py and the README.
Main entry points: main.py StateGraph workflow
Components 12
CLI interaction
- CLI loop
main.py
Triage workflow
- StateGraph workflow
main.py - Intent classifier
main.py - Conditional router
main.py - OpenAI chat model
- Workflow end
main.py
Department responses
- Billing response
main.py - Technical response
main.py - General response
main.py
Conversation state
- Graph state
main.py - In-memory checkpoint
main.py
Other components
- Support user
Connections 17
- Support user → CLI loop: enters query
main.py - CLI loop → StateGraph workflow: invokes workflow
main.py - StateGraph workflow → Intent classifier: starts at
main.py - Intent classifier → OpenAI chat model: classifies with
main.py - Intent classifier → Conditional router: routes by category
main.py - Conditional router → Billing response: dispatches billing
main.py - Conditional router → Technical response: dispatches technical
main.py - Conditional router → General response: dispatches general
main.py - Billing response → OpenAI chat model: generates response
main.py - Technical response → OpenAI chat model: generates response
main.py - General response → OpenAI chat model: generates response
main.py - Billing response → Workflow end: completes workflow
main.py - Technical response → Workflow end: completes workflow
main.py - General response → Workflow end: completes workflow
main.py - StateGraph workflow → Graph state: passes state
main.py - StateGraph workflow → In-memory checkpoint: persists thread state
main.py - CLI loop → Support user: prints reply
main.py