LLMOpsThe LLMOps lifecycle · MLflow 3 · FastAPI · Docker · AWS/Azure
Dashboard
0%
1
Curious builder0 XP earned · 300 to level 2
0 daysFinish a lesson to begin
Badge collection0 of 6 unlocked
50 small wins to finish your pathNext lesson →

The workflow that wires the nodes

The workflow is the graph that runs the two nodes in order. A LangGraph StateGraph starts, runs the index-video node, then the auditor node, then ends, passing the VideoAuditState from one to the next. Two nodes, but each does heavy work.

Last updated: 15 Oct, 2026 · LLMOps

The auditor was the second node. This lesson is the orchestration: the one place the order of the steps is defined, so no node calls the next one by hand.

Building the LangGraph workflow · from the End to End Multimodal LLMOps Project video · 2:36:00 to 2:39:00

Nodes, then edges

python
from langgraph.graph import StateGraph, END

# the graph carries the VideoAuditState from node to node
workflow = StateGraph(VideoAuditState)

# the two nodes, registered by name
workflow.add_node("index_video", index_video_node)
workflow.add_node("audit_content", audit_content_node)

# the flow: enter at index, then audit, then end
workflow.set_entry_point("index_video")
workflow.add_edge("index_video", "audit_content")
workflow.add_edge("audit_content", END)

app = workflow.compile()   # the runnable compliance graph

The flow is the whole control: an entry point at index, an edge from index to audit, and an edge from audit to end. Because the graph passes the state, the transcript the index node wrote is there for the auditor automatically; nothing hands it over by hand, which the video names as the point of using a graph.

The compiled LangGraph workflow drawn left to right: a START pill, the index_video node, the audit_content node, and an END pill, joined by three edges. A note says the state flows along the edges, so the auditor reads what the index node wrote.

Why a graph for two nodes

  • The order lives in one place. Changing the flow, adding a node, is an edge change, not a rewrite of the nodes.
  • The state is the hand-off. No node needs to know who runs next or pass data along manually; the graph does it.
  • It is ready to grow. A third node, a retry branch or a human-approval step slots in as another node and edge, the shape from Conditional edges.
Watch out. Two nodes can look like overkill for a graph, but the value is not the node count; it is that the flow is declared, inspectable and extendable. The same two-node graph is what a trace tool draws when you debug a run.
Try it yourself
  • Add a third node, a reporter that emails the report, and name the two edges that change.
  • Explain how the auditor node receives the transcript without the index node passing it directly.
PreviousNode two: the auditor

You understood something today that you didn't yesterday.