Why do LangGraph conditional edges exist when a plain DAG already has edges?

Flashcard·Easy·4.0 · 0·~30s·Asked atMicrosoft Cursor Perplexity

Attempt it

A conditional edge attaches a routing function to a node. After that node runs, the function reads the current state and returns the name of the next node (or END). This is what lets the graph express cycles (the same node can route back to itself), branch on result (a tool result decides whether to call another tool or stop), and retry loops. A pure DAG framework can only express forward, fixed edges, so agent loops and conditional branches cannot exist in it. Conditional edges are the primitive that turns a DAG framework into an agent framework.

Your answer

Quick reference

TL;DR

A conditional edge runs a routing function over state and returns the next node name. That is the primitive that lets a graph express cycles, branches, and retries.

Memory aid

Easy to grasp

Picture a board game. A normal arrow on the board says 'after square 3, always go to square 4'. Every game plays out the same way. A conditional arrow looks at the dice you just rolled and says 'if you rolled a 6, go back to start; if you rolled even, advance two; otherwise stay.' That single change turns a fixed path into a game with strategy. A flowing pipeline only has fixed arrows. An agent that decides what to do next based on what it just saw needs the dice-reading arrow. LangGraph's conditional edges are those dice-reading arrows.

Key concepts

LLM Frameworks Agents

Detailed answer & concept explanation

~4 min read

Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example. Click to expand.

Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example.

Quick scan · 2–4 min

Everything important, quickly.

7 min: routing function shape, three behaviours unlocked, contrast with DAG frameworks, interaction with checkpointer, common production patterns.

Code example

python

from langgraph.graph import StateGraph, END

def call_model(state):
    response = llm.invoke(state["messages"])
    return {"messages": state["messages"] + [response]}

def call_tool(state):
    out = run_tool(state["messages"][-1].tool_calls[0])
    return {"messages": state["messages"] + [out]}

def route(state) -> str:
    last = state["messages"][-1]
    return "tool" if getattr(last, "tool_calls", None) else END

graph = StateGraph(dict)
graph.add_node("model", call_model)
graph.add_node("tool", call_tool)
graph.add_edge("tool", "model")               # fixed edge: tool result re-enters model
graph.add_conditional_edges("model", route)   # routing function decides next step
graph.set_entry_point("model")
app = graph.compile()

Real-world examples

Real products, models, and research that use this idea.

LangGraph's prebuilt `create_react_agent` uses a conditional edge between the LLM node and the tool node to loop until the model emits no more tool calls
Anthropic's research-agent reference implementation routes via conditional edges among planner, retriever, and synthesizer until coverage is sufficient
Mastra's TypeScript workflow runtime adopts the same routing function as edge pattern to express agent loops in TS

Interviewer follow-ups

What an interviewer would ask next. Try answering before peeking at the approach.

QHow does the checkpointer interact with conditional edges during time-travel?

QWhy are route functions required to be pure reads of state?

QWhen would you use Send (graph parallel fan-out) instead of a conditional edge?

Don't say this

Red flags and common mistakes that signal junior thinking. Click to expand.

Most common mistake

Treating a conditional edge as 'just an if statement'. It is the routing primitive that makes cycles legal in the graph, which is exactly what a DAG cannot express.

60-second cheat sheet

The night-before-the-interview bullets. Scan these on the way to the call.

References

Primary sources. Skim if you want the original framing.