0xA0Lesson 11 of 11

Capstone: build a RAG pipeline

Chunk, index, retrieve and answer with citations - and abstain when the sources don’t cover the question.

35 min 5-question quiz 2 code exercises

By the end of this lesson you can

Chunk documents into citable sentences
Retrieve the best chunk for a question
Answer with a citation, or abstain below a threshold

Time to build a whole (tiny) RAG pipeline. To keep it runnable anywhere, the “generator” is extractive: it answers with the best-matching sentence itself, plus its citation. Swapping in a language model later changes only the last step - the indexing, retrieval, citation and abstention logic stays the same.

Two exercises:

Index: split each document into sentences with stable IDs like returns#2.
Answer: score sentences against the question, cite the best one, and say “I don’t know” when the best score is too low.

stem.py

def stem(word):
    return word[:-1] if word.endswith("s") and len(word) > 3 else word
print([stem(word) for word in ["costs", "electronics", "is", "days"]])

Output

['cost', 'electronic', 'is', 'day']

Key takeaways

A RAG pipeline is chunk → index → retrieve → generate with citations.
Stable chunk IDs make citations possible.
A score threshold lets the system abstain instead of guessing.

Lesson quiz

5 questions · pass with 4 correct · up to 50 XP

Passing this quiz completes the lesson and keeps your streak going. Questions you miss come back in review sessions later.

Practice: write Python

Write Python in the editor and run it against sample inputs. Python runs locally in your browser using a WebAssembly runtime.

Exercise 1

Step 1: index sentences

+25 XP

Each line is doc|text. Split the text into sentences with re.split(r"(?<=[.!?])\s+", text) and print each as doc#N: sentence, numbering from 1 within each document.

Two documents

main.py

Loading editor…

Python runs in a sandboxed browser worker with a 60 second time limit. Its runtime loads from the Pyodide CDN; your code stays in this browser.

Exercise 2

Step 2: answer with a citation

+25 XP

Lines before --- are documents (doc|text); lines after are questions. Index sentences as in step 1. For each question:

take its words (lowercase [a-z0-9]+), stem them with stem, and drop STOP words;
score each sentence by how many of those question words appear among its stemmed words;
pick the highest score (the first sentence wins ties).

Print question -> sentence [doc#N], or question -> I don't know - no source covers that. if the best score is below 2.

Answer, answer, abstain

main.py

Loading editor…

Python runs in a sandboxed browser worker with a 60 second time limit. Its runtime loads from the Pyodide CDN; your code stays in this browser.

Questions about this lesson

Stuck? Ask. Figured something out? Share it. Explaining is one of the best ways to learn.

Loading posts…