What this is

No new concepts in this lesson. This is a checkpoint: a small, real script built entirely out of ideas from Lessons 10 through 17, combined into one thing. If you can read lesson.py and understand why every piece is there, you've mastered the Intermediate tier. If any piece feels unfamiliar, that's a sign to revisit the lesson it came from before continuing to Advanced.

What it does

Loads all five fixture notes, then answers questions through ask(): retrieve, three-way confidence grade (Lesson 12), refine at the strip level (Lessons 10-11), rewrite and re-retrieve if nothing survives (Lessons 6-7), generate with source citations, and print whether correction actually changed anything, the same visible indicator idea from Lesson 9's Beginner checkpoint, now backed by the more precise Intermediate-tier machinery instead of binary grading.

Where each piece came from

graded = [{**c, "confidence": grade_confidence(query, c["text"])} for c in retrieved]
kept = [c for c in graded if c["confidence"] in ("correct", "ambiguous")]

Lesson 12's three-way grade and action mapping: both "correct" and "ambiguous" chunks are kept (and refined below), only "incorrect" is dropped outright.

context = "\n\n".join(f"[{c['source']}]\n{refine(effective_query, c['text'])}" for c in kept)

Lesson 11's refine(), applied to every kept chunk before it reaches the prompt, and Lesson 15's citation format ([source.md]), combined into one context string.

used_sources = {c["source"] for c in kept}
corrected = naive_top1_source is not None and naive_top1_source not in used_sources

Lesson 9's correction indicator, unchanged in spirit: compare what naive top-1 alone would have used against what the full pipeline actually used.

Try this yourself

Without looking anything up:

  • Ask Lesson 16's exact question ("How often does Project Aurora's wind speed sensor need re-oiling?") through ask(). Does the citation correctly point at weather-station.md? Would this pipeline have any way to notice if weather-station.md's content had been subtly wrong, the way Lesson 16's fabricated passage was?
  • Trace through what happens if every retrieved chunk grades "incorrect": which branch of ask() runs, and what does the final indicator print?
  • Compare this lesson's citation style to naive_rag Lesson 25's capstone. Both use [source.md], on purpose, matching conventions across this series makes a future course able to reuse either one's prompt template without translation.

If you can make these changes confidently, you're ready for the Advanced tier, starting at Lesson 19.