Prompt file imported from charris22/SocratesLearnAgent (.github/prompts/write-tests.prompt.md). Copyright stays with the author.

Write Tests

Owen wants to add tests. Guide him through the testing patterns used in this project.

Context

• Tests live in tests/
• We use pytest with pytest-asyncio for async test support
• Config in pyproject.toml: asyncio_mode = "auto", testpaths = ["tests"]
• Run with: pytest -v

Testing layers (explain the tradeoffs of each):

1. Unit tests (no LLM, no network)

These test pure logic — models, data transformations, grading.

# Example: tests/test_models.py
from app.models import TopicScore

def test_accuracy():
    score = TopicScore(subject="Math", topic="Fractions", attempts=10, correct=7)
    assert score.accuracy == 0.7

When to write these: For any logic that doesn't call Azure OpenAI.

2. Service tests (mock the LLM)

These test service logic with a fake LLM response.

# Example pattern:
from unittest.mock import AsyncMock, patch

@patch("app.services.chat_service.get_openai_client")
async def test_chat_returns_response(mock_client):
    mock_client.return_value.chat.completions.create = AsyncMock(
        return_value=MockResponse("Hello!")
    )
    result = await chat(ChatRequest(message="hi"))
    assert result.reply == "Hello!"

Explain: What is mocking? Why mock the LLM? (cost, speed, determinism)

3. API tests (FastAPI TestClient)

These test the HTTP layer.

from fastapi.testclient import TestClient
from app.main import app

client = TestClient(app)

def test_health():
    response = client.get("/health")
    assert response.status_code == 200

Steps:

1. Identify what to test — ask Owen what behavior they want to verify
2. Choose the right testing layer
3. Write the test, explaining each assertion
4. Run pytest -v and interpret the output
5. Discuss code coverage and what's worth testing vs over-testing