Skip to content

M4: Ship your first AI app

This is the one you came for. Today you connect your Python to a real AI model and build a chatbot with its own personality, running in front of you, talking back. And because you built the last three modules, this won't be a magic black box: you'll understand every line, from the secret key to the reply on screen.

Today's win: a working AI chatbot you built and can read line by line, you give it a personality, and it remembers the conversation.

Today you will

  • Get an API key, store it safely in .env, and confirm it with one tiny call
  • Understand an LLM from a builder's view: you send messages, it sends a reply
  • Build a minimal chatbot with a personality you choose

Run of show (~70 min)

Time What we do
0:00 Hook + the win we're chasing
0:05 The one idea: request → response, and what a key is (full read in notes.md)
0:10 Lab Part A: key → .env → confirm with a tiny call
0:35 Lab Part B: build and personalize the chatbot
1:05 Show: post a line of your bot's reply to the wins board
1:10 Wrap + take-home

If you get stuck

  • A red AuthenticationError almost always means a key typo or a misnamed .env, re-read the API-key guide's troubleshooting box. Nothing here can harm your computer.
  • Your key is a paid secret. Keep it in .env, never in your code, never in chat. If you ever paste it somewhere public, delete it in the Console and make a new one.
  • Re-read the You should now see line under each step. Compare with your breakout partner.

Instructor note: budget time for billing setup: students must add a little credit before any call works. Suggest everyone set a spend limit (a few dollars covers the whole course). The cheaper claude-haiku-4-5 model is a one-string swap for cost-sensitive practice (see notes).

Optional challenge

Give your bot a job: a patient Python tutor, a travel guide for one city, a pun machine. Change only the SYSTEM text and watch its whole personality change, same code, different soul.