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

Embeds all five fixture notes once, then answers four questions in a row, one needing retrieval, one needing the time tool, one needing neither, and one more needing retrieval again, printing which tool (if any) the model reached for isn't shown explicitly here the way Lesson 8 showed it, deliberately, this checkpoint reads like something you'd actually hand a user: just questions and answers, the tool routing invisible because it's supposed to be.

Where each piece came from

store = build_vector_store()

Lessons 2, 4: embed every fixture note once, up front, the expensive step that should only happen once per run.

TOOLS = types.Tool(function_declarations=[SEARCH_NOTES_DECLARATION, GET_CURRENT_DATETIME_DECLARATION])
CONFIG = types.GenerateContentConfig(tools=[TOOLS])

Lesson 7: both tools declared together, so the model can pick between them (or neither) per question.

def run_tool(call: types.FunctionCall, store: list[dict]) -> str:
    if call.name == "search_notes": ...
    elif call.name == "get_current_datetime": ...

Lesson 7's dispatch, unchanged: match the model's requested tool name to the real function that implements it.

def ask(query: str, store: list[dict]) -> str:
    ...

Lesson 5's loop shape, Lesson 6's "sometimes there's no call at all" branch, Lesson 7's multi-tool dispatch, all in one function: model call, tool call if requested, tool result handed back, final answer.

Try this yourself

Without looking anything up:

  • Add a third tool of your own, something as trivial as get_current_datetime() was (a word-counter, a Celsius-to-Fahrenheit converter, anything needing no new dependency). Declare it, add a branch to run_tool(), and ask a question only it can answer. Does the model reach for it correctly on the first try?
  • Ask a question that's genuinely ambiguous between the notes tool and general knowledge, for example "What's a good daily routine for learning a skill?" (it could answer generically, or pull from language-journal.md). Which way does it go, and does the tool's description explain why?
  • Remove get_current_datetime from TOOLS entirely (leave the function and its branch in run_tool() in place) and re-ask the New York time question. What does the model do when a tool it would have used is no longer declared? This previews Lesson 16's failure-modes lesson: a model can't call a tool it doesn't know exists, and its fallback behavior when that happens is worth seeing on purpose.

If you can make these changes confidently, you're ready for the Intermediate tier, starting at Lesson 10.