Nodes, edges, conditional edges, and routing

Overview

• A node is a function performing one graph step.
• An edge describes the next step, while a conditional edge selects from allowed destinations based on state.
• Routing can use model output, deterministic checks, or both.

Why this matters

Explicit transitions make failure paths and allowed actions reviewable.

An everyday example

A hospital sends a case to a specific department depending on validated symptoms and rules.

A technical example

intent === "claim_status" selects claim lookup; unsupported intent selects clarification.

How it works

1. Run node
2. Validate/update state
3. Router selects named destination
4. Execute next node
5. Terminate at a defined endpoint.

Architecture and flow

Node → bounded route → allowed next node.

The insurance assistant example

A model can select “create draft”; only a deterministic approval check can select “submit approved draft.”

When to use it

Branching tasks and explicit retry/escalation paths.

When to avoid it

Do not let model text become a dynamically imported node name or arbitrary endpoint URL.

Interview answer

“Nodes perform work and edges control execution. I validate model-derived routing and use business rules to control consequential transitions.”

Benefits, trade-offs, and production considerations

• Each node should have a clear input, output, deadline, and failure contract.
• Keep business services independently testable and avoid a giant node containing the whole agent.