Make LangGraph Agents Durable with Dapr Workflow - Schedule Planner

LangGraph gives an LLM the structure to act as a state machine: nodes that read and write shared state, edges that route between them, and a conditional edge that keeps the model and its tools looping until the model is done. What it doesn't give you is durability. Every node result and every message lives in a Python variable, so killing the process mid-loop loses the conversation and every LLM call you already paid for. In this self-paced track you'll see how Dapr Workflow turns the same compiled graph into a durable, crash-proof application.

What you'll run

You'll work with Schedule Planner, a LangGraph agent that looks up available slots for a venue. You'll read through the graph, then run it on Dapr Workflow and find its checkpointed state in Redis. Finally you'll crash a running graph halfway through and restart it: it resumes from the last checkpoint instead of redoing completed work.

Ensure you have your own OpenAI API key before you start this track.