Setup

You need Python 3.11 or newer, uv, and a Gemini API key. The key is free at aistudio.google.com/apikey, no credit card needed.

curl -LsSf https://astral.sh/uv/install.sh | sh     # installs uv
git clone https://github.com/omarcevi/trabzonspor_analyst
cd trabzonspor_analyst
cp .env.example .env                                 # paste your key into .env
uv sync
uv run adk web steps

Open http://localhost:8000, pick a step from the dropdown and ask it something. The events panel shows every tool call and hand-off.

The .env file holds your key, GOOGLE_GENAI_USE_VERTEXAI=FALSE (the free Gemini API, not Vertex AI) and MODEL=gemini-3.8-flash. The model is set in one place, matchday/config.py:

MODEL_NAME = os.getenv("MODEL", "gemini-3.8-flash")

# Free-tier keys hit rate limits when many agents run at once: retry 429s with backoff.
MODEL = Gemini(
    model=MODEL_NAME,
    retry_options=types.HttpRetryOptions(attempts=5, initial_delay=2, max_delay=30),
)

The retries matter more than they look: one question to step 5 can mean a few dozen model calls, and on a free key that hits a 429 now and then. If your laptop won't cooperate, notebook/match_day_agents.ipynb runs the same files in Colab, one cell per step.