Skip to main content

Exercise 1 - Your First Gemini Call

Goal: Get the Gemini API working and compare its output to llama3.2 on the same prompt.

Before you start​

You need an API key from Google AI Studio. It is free. Store it in your environment - never in code:

export GEMINI_API_KEY="your-key-here"

Install the SDK:

pip install google-generativeai

Assignment​

Open 01_gemini_call.py.

  1. Write a generate_gemini(prompt) function using google.generativeai with model gemini-2.0-flash.
  2. Write a generate_ollama(prompt) function using the Ollama HTTP API with llama3.2.
  3. Run both on the same three prompts:
    • A factual question with a clear answer
    • A request to summarise a short paragraph
    • An out-of-context question with no good answer
  4. Print both responses side by side for each prompt.
  5. Time each call. Print the latency.

Thinking questions​

  • Did the two models give the same answer? Different answers? Which felt more accurate?
  • Gemini was faster or slower than your local llama3.2? What does that depend on?
  • Your API key is in an environment variable. What would happen if you accidentally committed a file that contained the key as a string?

Next: Exercise 2 - Swap the Generator →