AIMentor — AI Expertise Learning Platform

Untimed. Answers are graded when you submit, with explanations and spaced-repetition scheduling.

Question 1easy
Why do production agents often need a graph (LangGraph) rather than a linear LCEL chain?
Graphs are always fasterAgents need branching, cycles, shared state, and pause/resume — which a one-way linear chain can't expressChains cannot call an LLMGraphs remove the need for tools
Question 2easy
In a LangGraph StateGraph, what is a 'node'?
A vector in the embedding spaceA function that takes the shared state and returns the fields to updateA database tableA prompt template only
Question 3medium
What does the shared state schema (e.g. a TypedDict) provide across nodes?
A single object that every node can read and write, with merges accumulating updatesA cache of API keysThe list of installed packagesThe embedding model weights
Question 4medium
What does a conditional edge do?
Runs two nodes at onceChooses the next node by running a function on the current state (branching)Deletes the stateStreams tokens to the client
Question 5medium
How is the agent's reason→act→observe loop represented in a LangGraph?
As a single nodeAs a cycle: an 'agent' node and a 'tools' node with a tools→agent edge, plus a stop-condition conditional edgeAs a linear chain with no edgesIt cannot be represented
Question 6medium
What does attaching a checkpointer to a compiled graph do?
Speeds up token generationPersists the graph's state after every node (keyed by thread id), so runs can resume after a crash or pauseTrains the model on the conversationRemoves the recursion limit
Question 7hard
How does human-in-the-loop approval work in LangGraph?
The graph emails an admin automaticallyinterrupt_before pauses execution before a sensitive node; because state is checkpointed, a human approves and the run resumes from the checkpointIt disables all toolsIt requires retraining the model
Question 8hard
To resume a graph that paused at an interrupt, what do you invoke it with?
The original input againNone (with the same thread_id config), so it continues from the checkpointA fresh StateGraphA new checkpointer
Question 9hard
When is reaching for LangGraph justified over a chain or AgentExecutor? (Select all that apply)
You need explicit branching between pathsYou need cycles you must inspect and controlYou need durable state across pauses or human-in-the-loop gatesYou have a single one-shot prompt with no branching
Select all that apply.
Question 10medium
Why is MemorySaver unsuitable as a production checkpointer?
It is too accurateIt stores state in volatile process memory; production needs a DB-backed checkpointer for real durabilityIt cannot store TypedDictsIt disables interrupts
Question 11hard
What keeps a LangGraph cycle from looping forever by default?
Nothing — cycles are always infiniteA recursion limit that bounds the number of steps, plus a stop-condition edgeThe temperature settingThe embedding dimension
Question 12medium
What graph feature (one word) lets a LangGraph pause before a sensitive node for human approval?
Question 13hard
In LangGraph, an agent's tool loop is expressed as what graph structure (one word) that a linear chain can't have?

0 / 13 answered

LangGraph Orchestration — Practice quiz