Running Tests

Amprealize uses pytest with markers to separate test tiers. For CI or infrastructure-dependent tests, use run_tests.sh with BreakerAmp.

Quick Start

# Run all unit tests (no DB or infra needed)
pytest -m unit

# Run a specific test file
pytest tests/test_behavior_service.py

# Run with verbose output
pytest -m unit -v

Test Markers

Run by marker:

pytest -m unit                    # Fast, no infra
pytest -m integration             # Needs DB
pytest -m "not slow"              # Skip slow tests
pytest -m "unit or integration"   # Both tiers

Using run_tests.sh

The test runner script handles infrastructure setup, especially when using BreakerAmp:

# Standard run
./scripts/run_tests.sh

# With BreakerAmp (spins up Postgres in container)
./scripts/run_tests.sh --breakeramp

# Specific markers
./scripts/run_tests.sh --markers unit

# Fresh environment (destroys and recreates)
./scripts/run_tests.sh --breakeramp --fresh

BreakerAmp mode:

1. Runs breakeramp plan --blueprint test to check infrastructure
2. Runs breakeramp apply to provision containers
3. Waits for health checks
4. Runs migrations (alembic upgrade head)
5. Executes pytest
6. Optionally tears down with breakeramp destroy

Test Database

Unit tests: Use mocked services or in-memory stores. No real DB needed.

Integration tests: Require a database. Options:

• BreakerAmp (recommended): Auto-provisions Postgres in Podman container
• Local Postgres: Set DATABASE_URL in .env and run alembic upgrade head
• SQLite: amprealize context add test --storage sqlite --path /tmp/test.db

Writing Tests

Follow the test pyramid: 70% unit, 20% integration, 10% E2E.

import pytest

pytestmark = pytest.mark.unit  # Mark entire module

class TestBehaviorService:
    def test_create_behavior(self, mock_pool):
        """Test creating a new behavior draft."""
        service = BehaviorService(pool=mock_pool)
        result = service.create_behavior(name="test", description="desc")
        assert result.status == "draft"

Key patterns:

• Use conftest.py fixtures for shared setup (mock pools, services)
• Mark tests with appropriate markers
• Keep unit tests under 100ms each
• Test both happy path and error paths

Coverage

pytest --cov=amprealize --cov-report=html -m unit
open htmlcov/index.html

Parity Tests

Cross-surface parity tests verify CLI, API, and MCP produce identical results:

pytest -m parity -v

These tests are in tests/test_*_parity.py files.