ECE study buddy

Part A

Get the code and a key

Do Part A one time. After it, the lecture code and your key are ready.

1. Download the code

   Download lec103-code.zip 23 KB

   Right-click the zip file. Click Extract All. Put the folder lec103-code inside your ai-ece folder.

   Double-click the zip file. Move the folder lec103-code into your ai-ece folder.

   Unzip it in your ai-ece folder: unzip lec103-code.zip.

   Sometimes you get a folder inside a folder with the same name. Use the folder that has README.md in it. The folder exercise-103.1 is already in it.

2. Go into the folder

   In VS Code, open the ai-ece folder and a new terminal. Turn the venv on. Then go into the code folder:

   cd lec103-code

   Copy

   The prompt must start with (.venv). If not, see setup step 6.

3. Make a free OpenRouter account

   OpenRouter gives free AI models. You do not need a card. Go to openrouter.ai and sign up.

4. Make a key

   Open Keys. Click Create key. Give it a name, for example ece-lab. Copy the key. OpenRouter shows it only one time.

   Your key is you. Keep it secret. Do not share it in WhatsApp or email. Do not put it in a .py file. Do not upload .env to GitHub. If someone sees your key, delete it on the Keys page and make a new one.

5. Put the key in .env

   In the lec103-code folder, copy the example file to a new file with the name .env:

   Copy-Item .env.example .env

   Copy

   cp .env.example .env

   Copy

   Open .env in VS Code. Paste your key after API_KEY=. Do not add spaces or quotes. Save the file (Ctrl + S, or Cmd + S on macOS).

   API_KEY=paste-your-key-here
   URL=https://openrouter.ai/api/v1/chat/completions
   MODEL=google/gemma-4-31b-it:free

   Do not change URL. MODEL is a free model: its name ends in :free.

6. Check: run your first AI call

   cd 0-basics
   python 01_first_call.py
   cd ..

   Copy

   Type a question and press Enter. The answer from the model comes back in a few seconds. You see an answer? Part A is done. You see an error? Read If something goes wrong.

Part B

Build the study buddy

Go to the exercise folder and run the starter file. You see the full JSON reply.

cd exercise-103.1
python starter.py

starter.py

You already have this file. It is in lec103-code/exercise-103.1/ from the zip in Part A. It works as it is. Each TODO is one step below. Change this file.

# Exercise 103.1 - ECE study buddy (starter file)
# This file works as it is. Run it first:  python starter.py
# Then do steps 1 to 6 from README.md. Each step has a TODO below.
import requests
from config import API_KEY, URL, MODEL

# TODO 1: Change this question to an ECE question.
question = "What is AI?"

# TODO 3: Make a "messages" list. Put a system message first, then the user message.
# TODO 4: Put the code from here to the end in a "while True:" loop.
#         Read the question with input(). Stop the loop when the user types "quit".
# TODO 5: Make the "messages" list one time, before the loop.
#         Add each question and each answer to it.

response = requests.post(
    URL,
    headers={"Authorization": "Bearer " + API_KEY},
    json={"model": MODEL, "messages": [{"role": "user", "content": question}]},
)
data = response.json()

# TODO 1: This prints the full JSON reply. Print only the answer text.
print(data)

# TODO 2: Print the tokens used. Use data.get("usage", {}).
# TODO 6 (bonus): Add up the tokens of all the turns. Print the total after each answer.

Do the steps in order. Run the file after each step.

1. Ask an ECE question

   Goal

   Ask an ECE question. Print only the answer text.

   Hint

   Change the question line. The answer text is at data["choices"][0]["message"]["content"].

   Check

   You see only the answer. You do not see {'id': ...

2. Print the tokens

   Goal

   Print the number of tokens for this call.

   Hint

   Write usage = data.get("usage", {}). Then print usage.get("prompt_tokens"), usage.get("completion_tokens") and usage.get("total_tokens").

   Check

   You see three numbers. Input + output = total.

3. Add a system prompt

   Goal

   Give the model a job.

   Hint

   Make a messages list. Put this message first:

   {"role": "system", "content": "You are a friendly tutor for ECE students. Answer in 3 short lines."}

   Put the user message second. Send messages in the json= part.

   Check

   The answer has 3 short lines. Put # before the system message and run again. Compare the two answers.

4. Ask many questions

   Goal

   Ask questions again and again. Type quit to stop.

   Hint

   Put the code in a while True: loop. Use question = input("You: "). Use break when the text is quit.

   Check

   Type my name is Ravi. Then type what is my name?. The model does not know. It has no memory.

5. Give it a memory

   Goal

   The model remembers the chat.

   Hint

   Make the messages list one time, before the loop. In the loop, append the user message. After the answer, append {"role": "assistant", "content": answer}. Send the whole list each time.

   Check

   Type my name is Ravi. Then type what is my name?. The model says “Ravi”.

6. Bonus: watch the tokens grow

   Goal

   Print the total tokens after each turn.

   Hint

   Before the loop, write tokens_so_far = 0. After each answer, add usage.get("total_tokens", 0) to it. Print it.

   Check

   The input tokens grow each turn. Why? You send the whole history each time. A long chat costs more.

Solution

Try each step first. Then compare your code with the solution.

Test it: type my name is Ravi, then what is my name?, then quit.