AIMentor — AI Expertise Learning Platform

13 questions · pass mark 80% · 20 minute limit. The timer auto-submits at zero.

Time remaining: 20:00

Question 1easy
What is the single unifying abstraction that most LangChain components implement?
The EmbeddingThe Runnable (with invoke / stream / batch)The VectorStoreThe Tokenizer
Question 2easy
In LCEL, what does `prompt | model | parser` build?
Three separate programsA single composed Runnable where each stage's output feeds the nextA bitwise OR of the componentsA vector index
Question 3medium
Why do .stream and .batch work on a composed LCEL chain without extra code?
Because the chain is compiled to CBecause the composition is itself a Runnable, inheriting the whole Runnable interfaceBecause streaming is disabled by defaultBecause batch only works on single models
Question 4medium
What does a plain dict like {"context": retriever, "question": RunnablePassthrough()} become in LCEL?
A RunnableParallel that runs each branch on the same input and collects a dict of resultsA Python setA sequential chainAn error
Question 5medium
What is RunnablePassthrough used for?
To call the LLM twiceTo forward the input unchanged (optionally adding fields), e.g. keep the original question alongside computed contextTo parse JSON outputTo embed the input
Question 6medium
What does RunnableWithMessageHistory add to a chain?
Faster inferencePer-session conversation memory: it loads and appends message history keyed by a session idAutomatic fine-tuningA vector database
Question 7hard
Why is InMemoryChatMessageHistory unsuitable for production?
It is too slowIt lives in process memory — lost on restart and not shared across servers; history also grows unboundedIt cannot store stringsIt requires a GPU
Question 8medium
In a LangChain @tool function, what becomes the tool's description that the model uses to route?
The function name onlyThe function's docstringA separate YAML fileThe return type annotation
Question 9hard
What does an AgentExecutor handle for you that you wrote by hand in earlier topics? (Select all that apply)
The reason→act→observe loopThe bounded iteration count (max_iterations)tool_call_id plumbing and error handlingChoosing which embedding model to train
Select all that apply.
Question 10hard
What is the most durable thing to learn about LangChain, given its fast-moving API?
The exact import paths of the current releaseThe mental model — Runnables, LCEL composition, and the agent loop — which stays stable while imports churnThe list of every deprecated classThe internal source code line numbers
Question 11medium
For a single one-shot prompt (no runtime decisions), what should you use in LangChain?
A full AgentExecutor with toolsA simple LCEL chain like prompt | model | parserA multi-agent supervisorFine-tuning
Question 12medium
What is the name of the interface (one word) that prompts, models, parsers, chains, and agents all implement in LangChain?
Question 13hard
What acronym names LangChain's expression language for composing Runnables with the pipe operator?

0 / 13 answered

LangChain Essentials — Final Exam