Skip to main content

Lab 14: LangGraph Agent Workflow

Difficulty: Advanced · Estimated time: ~4–6 hours

Objective​

Implement a stateful agent workflow (plan → act → synthesize) as a graph:

  • explicit state object
  • deterministic transitions
  • repair loop on failures

Requirements​

  • At least 4 nodes: router, tool, verifier, synthesizer
  • JSONL logging per node
  • A replay mode that re-runs from saved state without calling tools

Deliverables​

  • graph.py
  • state.py
  • logs/ (sample run)
  • README.md explaining the graph and transitions