AIMentor — AI Expertise Learning Platform

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

Question 1easy
Fundamentally, what does a RAG framework like LangChain or LlamaIndex provide over a hand-rolled pipeline?
A capability impossible to build yourselfThe same embed→chunk→index→retrieve→generate stages packaged as reusable, connector-rich componentsA different mathematical approach to similarityA replacement for the LLM
Question 2easy
LlamaIndex is best characterized as:
Composition-first, built around chaining arbitrary stepsIndex-first, built around retrieval with a fast path to a RAG baseline and many data connectorsA vector databaseAn embedding model
Question 3easy
In LangChain's LCEL, what does the `|` operator mean in `retriever | prompt | llm`?
Bitwise OR of the componentsFeed each component's output into the next — function composition, like a pipeRun the components in parallelA logical OR fallback between components
Question 4medium
What is LangChain's primary strength relative to LlamaIndex?
It requires no dependenciesExplicit composition of swappable stages that extends into multi-step chains and agentsIt has the only PDF loaderIt eliminates the need for a prompt
Question 5medium
In the abstraction mapping, your hand-rolled `recursive_split` corresponds to which framework concept?
A retrieverA text splitter / node parser (e.g. RecursiveCharacterTextSplitter, SentenceSplitter)A vector storeAn output parser
Question 6medium
When is hand-rolling a RAG pipeline a reasonable choice? (Select all that apply)
The pipeline is simple and stableYou want zero dependency/version churnYou must understand and audit every line for debugging or complianceYou need 100+ data connectors and complex agentic branching
Select all that apply.
Question 7medium
What does LlamaIndex's `VectorStoreIndex.from_documents(docs)` do in one call?
Only loads files from diskChunks, embeds, and indexes the documentsGenerates the final answerTrains an embedding model
Question 8hard
Why wrap a framework behind your own thin `retrieve()`/`answer()` interface?
To make the code run fasterSo a framework migration (or a drop back to hand-rolled) touches one adapter, not the whole codebaseBecause frameworks forbid direct callsTo avoid writing any prompts
Question 9hard
You migrate a working pipeline to a framework. What makes that swap SAFE?
Trusting the framework's defaultsRunning the golden-set evaluation before and after and confirming the scorecard didn't regressDeleting the old code immediatelyIncreasing the temperature
Question 10medium
Can LangChain and LlamaIndex be used together?
No, they are mutually exclusiveYes — e.g. LlamaIndex for ingestion/retrieval inside a LangChain agentOnly if you fork both librariesOnly for non-RAG tasks
Question 11medium
An ingestion-heavy app must pull from Notion, Slack, Google Drive, and SQL into one index. Which factor most favors adopting LlamaIndex?
Its temperature controlIts large library of data connectors/loadersIt is the only option with vectorsIt removes the need for embeddings
Question 12hard
Which caution is MOST specific to using these frameworks?
Embeddings must be normalizedTheir class names and import paths change fast between releases, so months-old snippets often breakThey cannot do rerankingThey require a GPU
Question 13medium
What is the name (acronym) of LangChain's expression language that composes components with the `|` operator?
Question 14hard
Complete the rule of thumb: start hand-rolled for learning and simple cases; adopt a framework when the glue code you're writing is exactly what the framework already ____.

0 / 14 answered

RAG Frameworks — LangChain & LlamaIndex — Practice quiz