AI Experts Assignment (Python)

How to run tests

Locally

1. Create and activate a virtual environment:

   python3 -m venv .venv
   source .venv/bin/activate

2. Install dependencies:

   pip install -r requirements.txt

3. Run tests:

   pytest -v

With Docker

1. Build the image:

   docker build -t ai-experts-assignment .

2. Run the tests:

   docker run --rm ai-experts-assignment

• set up a small Python project to run reliably (locally + in Docker),
• pin dependencies for reproducible installs,
• write focused tests to reproduce a bug,
• implement a minimal, reviewable fix.

What you will do

1) Dockerfile (required)

Create a Dockerfile so the project can run the test suite in a non-interactive, CI-style environment.

Requirements:

• requirements.txt exists and is used during build (pip install -r requirements.txt)
• pytest must be included/pinned in requirements.txt
• The image must run tests by default (use: CMD ["python", "-m", "pytest", "-v"]).
• The build must install dependencies from requirements.txt.

2) requirements.txt (required)

Create a requirements.txt with pinned versions, using this format:

• package==x.y.z

3) README updates (required)

Update this README to include:

• how to run the tests locally,
• how to build and run tests with Docker.

4) Find + fix a bug (required)

There is a bug somewhere in this repository.

Your tasks:

• Identify the bug through reading code and/or running tests.
• Write tests that reproduce the bug (tests should fail on the current code).
• Apply the smallest possible fix to make the tests pass.
• Keep the change minimal and reviewable (no refactors).

Constraints

• Keep changes minimal and reviewable.
• Do not refactor unrelated code.
• Do not introduce extra tooling unless required.
• You may add tests and the smallest code change needed to fix the bug.

5) EXPLANATION.md (required)

Create EXPLANATION.md (max 250 words) containing:

• What was the bug?
• Why did it happen?
• Why does your fix solve it?
• One realistic case / edge case your tests still don’t cover

Submission

• Submit a public GitHub repository URL containing your solution to the Google form link provided.