Checkpoint: CLI Q&A Over Notes and Images

What this is

No new concepts in this lesson. This is a checkpoint: a small, real script built entirely out of ideas from Lessons 1 through 8, combined into one thing. If you can read lesson.py and understand why every piece is there, you've mastered the Beginner tier. If any piece feels unfamiliar, that's a sign to revisit the lesson it came from before continuing to Intermediate.

What it does

Builds the mixed store (Lesson 6: five notes, four images captioned and embedded), then answers four questions, one answerable only from a note's text, one answerable only from an image, and two spanning different notes and images, printing the whole run so you can see retrieval pick the right modality for each question without being told which modality to use.

Where each piece came from

def build_mixed_store() -> list[dict]:
    ...

Lesson 6, unchanged: text records from fixtures/notes/, image records from fixtures/images/ (captioned via Lesson 4's caption_image, embedded via Lesson 5's embed_texts), concatenated into one list.

def retrieve(query: str, store: list[dict], k: int) -> list[dict]:
    ...

Lesson 7, unchanged: the same retrieve() naive_rag has used since its own Lesson 6, run here against a store that happens to contain captions as well as document chunks.

def generate_answer(query: str, retrieved: list[dict]) -> str:
    ...

Lesson 8, unchanged: build contents record by record, re-attaching a retrieved image's original file, or a retrieved chunk's text, then generate.

def ask(query: str, store: list[dict], k: int = 2) -> str:
    retrieved = retrieve(query, store, k)
    return generate_answer(query, retrieved)

The whole pipeline as one function call, the same shape naive_rag Lesson 8 introduced, now running over text and images without any branch in ask() itself, the branching happens once, inside generate_answer.