gridpilot.gg/docs/TESTS.md

# Testing Strategy

## Overview

GridPilot employs a comprehensive BDD (Behavior-Driven Development) testing strategy across three distinct layers: **Unit**, **Integration**, and **End-to-End (E2E)**. Each layer validates different aspects of the system while maintaining a consistent Given/When/Then approach that emphasizes behavior over implementation.

This document provides practical guidance on testing philosophy, test organization, tooling, and execution patterns for GridPilot.

---

## BDD Philosophy

### Why BDD for GridPilot?

GridPilot manages complex business rules around league management, team registration, event scheduling, result processing, and standings calculation. These rules must be:

- **Understandable** by non-technical stakeholders (league admins, race organizers)
- **Verifiable** through automated tests that mirror real-world scenarios
- **Maintainable** as business requirements evolve

BDD provides a shared vocabulary (Given/When/Then) that bridges the gap between domain experts and developers, ensuring tests document expected behavior rather than technical implementation details.

### Given/When/Then Format

All tests—regardless of layer—follow this structure:

```typescript
// Given: Establish initial state/context
// When: Perform the action being tested
// Then: Assert the expected outcome
```

**Example (Unit Test):**
```typescript
describe('League Domain Entity', () => {
  it('should add a team when team limit not reached', () => {
    // Given
    const league = new League('Summer Series', { maxTeams: 10 });
    const team = new Team('Racing Legends');
    
    // When
    const result = league.addTeam(team);
    
    // Then
    expect(result.isSuccess()).toBe(true);
    expect(league.teams).toContain(team);
  });
});
```

This pattern applies equally to integration tests (with real database operations) and E2E tests (with full UI workflows).

---

## Test Types & Organization

### Unit Tests (`/tests/unit`)

**Scope:** Domain entities, value objects, and application use cases with mocked ports (repositories, external services).

**Tooling:** Vitest (fast, TypeScript-native, ESM support)

**Execution:** Parallel, target <1 second total runtime

**Purpose:**
- Validate business logic in isolation
- Ensure domain invariants hold (e.g., team limits, scoring rules)
- Test use case orchestration with mocked dependencies

**Examples from Architecture:**

1. **Domain Entity Test:**
   ```typescript
   // League.addTeam() validation
   Given a League with maxTeams=10 and 9 current teams
   When addTeam() is called with a valid Team
   Then the team is added successfully
   
   Given a League with maxTeams=10 and 10 current teams
   When addTeam() is called
   Then a DomainError is returned with "Team limit reached"
   ```

2. **Use Case Test:**
   ```typescript
   // GenerateStandingsUseCase
   Given a League with 5 teams and completed races
   When execute() is called
   Then LeagueRepository.findById() is invoked
   And ScoringRule.calculatePoints() is called for each team
   And sorted standings are returned
   ```

3. **Scoring Rule Test:**
   ```typescript
   // ScoringRule.calculatePoints()
   Given a F1-style scoring rule (25-18-15-12-10-8-6-4-2-1)
   When calculatePoints(position=1) is called
   Then 25 points are returned
   
   Given the same rule
   When calculatePoints(position=11) is called
   Then 0 points are returned
   ```

**Key Practices:**
- Mock only at architecture boundaries (ports like `ILeagueRepository`)
- Never mock domain entities or value objects
- Keep tests fast (<10ms per test)
- Use in-memory test doubles for simple cases

---

### Integration Tests (`/tests/integration`)

**Scope:** Repository implementations, infrastructure adapters (PostgreSQL, Redis, OAuth clients, result importers).

**Tooling:** Vitest + Testcontainers (spins up real PostgreSQL/Redis in Docker)

**Execution:** Sequential, ~10 seconds per suite

**Purpose:**
- Validate that infrastructure adapters correctly implement port interfaces
- Test database queries, migrations, and transaction handling
- Ensure external API clients handle authentication and error scenarios

**Examples from Architecture:**

1. **Repository Test:**
   ```typescript
   // PostgresLeagueRepository
   Given a PostgreSQL container is running
   When save() is called with a League entity
   Then the league is persisted to the database
   And findById() returns the same league with correct attributes
   ```

2. **OAuth Client Test:**
   ```typescript
   // IRacingOAuthClient
   Given valid iRacing credentials
   When authenticate() is called
   Then an access token is returned
   And the token is cached in Redis for 1 hour
   
   Given expired credentials
   When authenticate() is called
   Then an AuthenticationError is thrown
   ```

3. **Result Importer Test:**
   ```typescript
   // EventResultImporter
   Given an Event exists in the database
   When importResults() is called with iRacing session data
   Then Driver entities are created/updated
   And EventResult entities are persisted with correct positions/times
   And the Event status is updated to 'COMPLETED'
   ```

**Key Practices:**
- Use Testcontainers to spin up real databases (not mocks)
- Clean database state between tests (truncate tables or use transactions)
- Seed minimal test data via SQL fixtures
- Test both success and failure paths (network errors, constraint violations)

---

### End-to-End Tests (`/tests/e2e`)

**Scope:** Full user workflows spanning web-client → web-api → database.

**Tooling:** Playwright + Docker Compose (orchestrates all services)

**Execution:** ~2 minutes per scenario

**Purpose:**
- Validate complete user journeys from UI interactions to database changes
- Ensure services integrate correctly in a production-like environment
- Catch regressions in multi-service workflows

**Examples from Architecture:**

1. **League Creation Workflow:**
   ```gherkin
   Given an authenticated league admin
   When they navigate to "Create League"
   And fill in league name, scoring system, and team limit
   And submit the form
   Then the league appears in the admin dashboard
   And the database contains the new league record
   And the league is visible to other users
   ```

2. **Team Registration Workflow:**
   ```gherkin
   Given a published league with 5/10 team slots filled
   When a team captain navigates to the league page
   And clicks "Join League"
   And fills in team name and roster
   And submits the form
   Then the team appears in the league's team list
   And the team count updates to 6/10
   And the captain receives a confirmation email
   ```

3. **Automated Result Import:**
   ```gherkin
   Given a League with an upcoming Event
   And iRacing OAuth credentials are configured
   When the scheduled import job runs
   Then the job authenticates with iRacing
   And fetches session results for the Event
   And creates EventResult records in the database
   And updates the Event status to 'COMPLETED'
   And triggers standings recalculation
   ```

4. **Companion App Login Automation:**
   ```gherkin
   Given a League Admin enables companion app login automation
   When the companion app is launched
   Then the app polls for a generated login token from web-api
   And auto-fills iRacing credentials from the admin's profile
   And logs into iRacing automatically
   And confirms successful login to web-api
   ```

**Key Practices:**
- Use Playwright's Page Object pattern for reusable UI interactions
- Test both happy paths and error scenarios (validation errors, network failures)
- Clean database state between scenarios (via API or direct SQL)
- Run E2E tests in CI before merging to main branch

---

## Test Data Strategy

### Fixtures & Seeding

**Unit Tests:**
- Use in-memory domain objects (no database)
- Factory functions for common test entities:
  ```typescript
  function createTestLeague(overrides?: Partial<LeagueProps>): League {
    return new League('Test League', { maxTeams: 10, ...overrides });
  }
  ```

**Integration Tests:**
- Use Testcontainers to spin up fresh PostgreSQL instances
- Seed minimal test data via SQL scripts:
  ```sql
  -- tests/integration/fixtures/leagues.sql
  INSERT INTO leagues (id, name, max_teams) VALUES
    ('league-1', 'Test League', 10);
  ```
- Clean state between tests (truncate tables or rollback transactions)

**E2E Tests:**
- Pre-seed database via migrations before Docker Compose starts
- Use API endpoints to create test data when possible (validates API behavior)
- Database cleanup between scenarios:
  ```typescript
  // tests/e2e/support/database.ts
  export async function cleanDatabase() {
    await sql`TRUNCATE TABLE event_results CASCADE`;
    await sql`TRUNCATE TABLE events CASCADE`;
    await sql`TRUNCATE TABLE teams CASCADE`;
    await sql`TRUNCATE TABLE leagues CASCADE`;
  }
  ```

---

## Docker E2E Setup

### Architecture

E2E tests run against a full stack orchestrated by `docker-compose.test.yml`:

```yaml
services:
  postgres:
    image: postgres:16
    environment:
      POSTGRES_DB: gridpilot_test
      POSTGRES_USER: test
      POSTGRES_PASSWORD: test

  redis:
    image: redis:7-alpine

  web-api:
    build: ./src/apps/web-api
    depends_on:
      - postgres
      - redis
    environment:
      DATABASE_URL: postgres://test:test@postgres:5432/gridpilot_test
      REDIS_URL: redis://redis:6379
    ports:
      - "3000:3000"
```

### Execution Flow

1. **Start Services:** `docker compose -f docker-compose.test.yml up -d`
2. **Run Migrations:** `npm run migrate:test` (seeds database)
3. **Execute Tests:** Playwright targets `http://localhost:3000`
4. **Teardown:** `docker compose -f docker-compose.test.yml down -v`

### Environment Setup

```bash
# tests/e2e/setup.ts
export async function globalSetup() {
  // Wait for web-api to be ready
  await waitForService('http://localhost:3000/health');
  
  // Run database migrations
  await runMigrations();
}

export async function globalTeardown() {
  // Stop Docker Compose services
  await exec('docker compose -f docker-compose.test.yml down -v');
}
```

---

## BDD Scenario Examples

### 1. League Creation (Success + Failure)

```gherkin
Scenario: Admin creates a new league
  Given an authenticated admin user
  When they submit a league form with:
    | name           | Summer Series 2024  |
    | maxTeams       | 12                  |
    | scoringSystem  | F1                  |
  Then the league is created successfully
  And the admin is redirected to the league dashboard
  And the database contains the new league

Scenario: League creation fails with duplicate name
  Given a league named "Summer Series 2024" already exists
  When an admin submits a league form with name "Summer Series 2024"
  Then the form displays error "League name already exists"
  And no new league is created in the database
```

### 2. Team Registration (Success + Failure)

```gherkin
Scenario: Team registers for a league
  Given a published league with 5/10 team slots
  When a team captain submits registration with:
    | teamName  | Racing Legends     |
    | drivers   | Alice, Bob, Carol  |
  Then the team is added to the league
  And the team count updates to 6/10
  And the captain receives a confirmation email

Scenario: Registration fails when league is full
  Given a published league with 10/10 team slots
  When a team captain attempts to register
  Then the form displays error "League is full"
  And the team is not added to the league
```

### 3. Automated Result Import (Success + Failure)

```gherkin
Scenario: Import results from iRacing
  Given a League with an Event scheduled for today
  And iRacing OAuth credentials are configured
  When the scheduled import job runs
  Then the job authenticates with iRacing API
  And fetches session results for the Event
  And creates EventResult records for each driver
  And updates the Event status to 'COMPLETED'
  And triggers standings recalculation

Scenario: Import fails with invalid credentials
  Given an Event with expired iRacing credentials
  When the import job runs
  Then an AuthenticationError is logged
  And the Event status remains 'SCHEDULED'
  And an admin notification is sent
```

### 4. Parallel Scoring Calculation

```gherkin
Scenario: Calculate standings for multiple leagues concurrently
  Given 5 active leagues with completed events
  When the standings recalculation job runs
  Then each league's standings are calculated in parallel
  And the process completes in <5 seconds
  And all standings are persisted correctly
  And no race conditions occur (validated via database integrity checks)
```

### 5. Companion App Login Automation

```gherkin
Scenario: Companion app logs into iRacing automatically
  Given a League Admin enables companion app login automation
  And provides their iRacing credentials
  When the companion app is launched
  Then the app polls web-api for a login token
  And retrieves the admin's encrypted credentials
  And auto-fills the iRacing login form
  And submits the login request
  And confirms successful login to web-api
  And caches the session token for 24 hours
```

---

## Coverage Goals

### Target Coverage Levels

- **Domain/Application Layers:** >90% (critical business logic)
- **Infrastructure Layer:** >80% (repository implementations, adapters)
- **Presentation Layer:** Smoke tests (basic rendering, no exhaustive UI coverage)

### Running Coverage Reports

```bash
# Unit + Integration coverage
npm run test:coverage

# View HTML report
open coverage/index.html

# E2E coverage (via Istanbul)
npm run test:e2e:coverage
```

### What to Prioritize

1. **Domain Entities:** Invariants, validation rules, state transitions
2. **Use Cases:** Orchestration logic, error handling, port interactions
3. **Repositories:** CRUD operations, query builders, transaction handling
4. **Adapters:** External API clients, OAuth flows, result importers

**What NOT to prioritize:**
- Trivial getters/setters
- Framework boilerplate (Express route handlers)
- UI styling (covered by visual regression tests if needed)

---

## Continuous Testing

### Watch Mode (Development)

```bash
# Auto-run unit tests on file changes
npm run test:watch

# Auto-run integration tests (slower, but useful for DB work)
npm run test:integration:watch
```

### CI/CD Pipeline

```mermaid
graph LR
  A[Code Push] --> B[Unit Tests]
  B --> C[Integration Tests]
  C --> D[E2E Tests]
  D --> E[Deploy to Staging]
```

**Execution Order:**
1. **Unit Tests** (parallel, <1 second) — fail fast on logic errors
2. **Integration Tests** (sequential, ~10 seconds) — catch infrastructure issues
3. **E2E Tests** (sequential, ~2 minutes) — validate full workflows
4. **Deploy** — only if all tests pass

**Parallelization:**
- Unit tests run in parallel (Vitest default)
- Integration tests run sequentially (avoid database conflicts)
- E2E tests run sequentially (UI interactions are stateful)

---

## Testing Best Practices

### 1. Test Behavior, Not Implementation

**❌ Bad (overfitted to implementation):**
```typescript
it('should call repository.save() once', () => {
  const repo = mock<ILeagueRepository>();
  const useCase = new CreateLeagueUseCase(repo);
  useCase.execute({ name: 'Test' });
  expect(repo.save).toHaveBeenCalledTimes(1);
});
```

**✅ Good (tests observable behavior):**
```typescript
it('should persist the league to the repository', async () => {
  const repo = new InMemoryLeagueRepository();
  const useCase = new CreateLeagueUseCase(repo);
  
  const result = await useCase.execute({ name: 'Test' });
  
  expect(result.isSuccess()).toBe(true);
  const league = await repo.findById(result.value.id);
  expect(league?.name).toBe('Test');
});
```

### 2. Mock Only at Architecture Boundaries

**Ports (interfaces)** should be mocked in use case tests:
```typescript
const mockRepo = mock<ILeagueRepository>({
  save: jest.fn().mockResolvedValue(undefined),
});
```

**Domain entities** should NEVER be mocked:
```typescript
// ❌ Don't do this
const mockLeague = mock<League>();

// ✅ Do this
const league = new League('Test League', { maxTeams: 10 });
```

### 3. Keep Tests Readable and Maintainable

**Arrange-Act-Assert Pattern:**
```typescript
it('should calculate standings correctly', () => {
  // Arrange: Set up test data
  const league = createTestLeague();
  const teams = [createTestTeam('Team A'), createTestTeam('Team B')];
  const results = [createTestResult(teams[0], position: 1)];
  
  // Act: Perform the action
  const standings = league.calculateStandings(results);
  
  // Assert: Verify the outcome
  expect(standings[0].team).toBe(teams[0]);
  expect(standings[0].points).toBe(25);
});
```

### 4. Test Error Scenarios

Don't just test the happy path:
```typescript
describe('League.addTeam()', () => {
  it('should add team successfully', () => { /* ... */ });
  
  it('should fail when team limit reached', () => {
    const league = createTestLeague({ maxTeams: 1 });
    league.addTeam(createTestTeam('Team A'));
    
    const result = league.addTeam(createTestTeam('Team B'));
    
    expect(result.isFailure()).toBe(true);
    expect(result.error.message).toBe('Team limit reached');
  });
  
  it('should fail when adding duplicate team', () => { /* ... */ });
});
```

---

## Common Patterns

### Setting Up Test Fixtures

**Factory Functions:**
```typescript
// tests/support/factories.ts
export function createTestLeague(overrides?: Partial<LeagueProps>): League {
  return new League('Test League', {
    maxTeams: 10,
    scoringSystem: 'F1',
    ...overrides,
  });
}

export function createTestTeam(name: string): Team {
  return new Team(name, { drivers: ['Driver 1', 'Driver 2'] });
}
```

### Mocking Ports in Use Case Tests

```typescript
// tests/unit/application/CreateLeagueUseCase.test.ts
describe('CreateLeagueUseCase', () => {
  let mockRepo: jest.Mocked<ILeagueRepository>;
  let useCase: CreateLeagueUseCase;
  
  beforeEach(() => {
    mockRepo = {
      save: jest.fn().mockResolvedValue(undefined),
      findById: jest.fn().mockResolvedValue(null),
      findByName: jest.fn().mockResolvedValue(null),
    };
    useCase = new CreateLeagueUseCase(mockRepo);
  });
  
  it('should create a league when name is unique', async () => {
    const result = await useCase.execute({ name: 'New League' });
    
    expect(result.isSuccess()).toBe(true);
    expect(mockRepo.save).toHaveBeenCalledWith(
      expect.objectContaining({ name: 'New League' })
    );
  });
});
```

### Database Cleanup Strategies

**Integration Tests:**
```typescript
// tests/integration/setup.ts
import { sql } from './database';

export async function cleanDatabase() {
  await sql`TRUNCATE TABLE event_results CASCADE`;
  await sql`TRUNCATE TABLE events CASCADE`;
  await sql`TRUNCATE TABLE teams CASCADE`;
  await sql`TRUNCATE TABLE leagues CASCADE`;
}

beforeEach(async () => {
  await cleanDatabase();
});
```

**E2E Tests:**
```typescript
// tests/e2e/support/hooks.ts
import { test as base } from '@playwright/test';

export const test = base.extend({
  page: async ({ page }, use) => {
    // Clean database before each test
    await fetch('http://localhost:3000/test/cleanup', { method: 'POST' });
    await use(page);
  },
});
```

### Playwright Page Object Pattern

```typescript
// tests/e2e/pages/LeaguePage.ts
export class LeaguePage {
  constructor(private page: Page) {}
  
  async navigateToCreateLeague() {
    await this.page.goto('/leagues/create');
  }
  
  async fillLeagueForm(data: { name: string; maxTeams: number }) {
    await this.page.fill('[name="name"]', data.name);
    await this.page.fill('[name="maxTeams"]', data.maxTeams.toString());
  }
  
  async submitForm() {
    await this.page.click('button[type="submit"]');
  }
  
  async getSuccessMessage() {
    return this.page.textContent('.success-message');
  }
}

// Usage in test
test('should create league', async ({ page }) => {
  const leaguePage = new LeaguePage(page);
  await leaguePage.navigateToCreateLeague();
  await leaguePage.fillLeagueForm({ name: 'Test', maxTeams: 10 });
  await leaguePage.submitForm();
  
  expect(await leaguePage.getSuccessMessage()).toBe('League created');
});
```

---

## Real E2E Testing Strategy (No Mocks)

GridPilot requires two distinct E2E testing strategies due to the nature of its automation adapters:

1. **Strategy A (Docker)**: Test `BrowserDevToolsAdapter` with Puppeteer against a fixture server
2. **Strategy B (Native macOS)**: Test `NutJsAutomationAdapter` on real hardware with display access

### Constraint: iRacing Terms of Service

- **Production**: nut.js OS-level automation only (no Puppeteer/CDP for actual iRacing automation)
- **Testing**: Puppeteer CAN be used to test `BrowserDevToolsAdapter` against static HTML fixtures

### Test Architecture Overview

```mermaid
graph TB
    subgraph Docker E2E - CI
        FX[Static HTML Fixtures] --> FS[Fixture Server Container]
        FS --> HC[Headless Chrome Container]
        HC --> BDA[BrowserDevToolsAdapter Tests]
    end
    
    subgraph Native E2E - macOS Runner
        SCR[Screen Capture] --> TM[Template Matching Tests]
        WF[Window Focus Tests] --> NJA[NutJsAutomationAdapter Tests]
        KB[Keyboard/Mouse Tests] --> NJA
    end
```

---

### Strategy A: Docker-Based E2E Tests

#### Purpose
Test the complete 18-step workflow using `BrowserDevToolsAdapter` against real HTML fixtures without mocks.

#### Architecture

```yaml
# docker/docker-compose.e2e.yml
services:
  # Headless Chrome with remote debugging enabled
  chrome:
    image: browserless/chrome:latest
    ports:
      - "9222:3000"
    environment:
      - CONNECTION_TIMEOUT=600000
      - MAX_CONCURRENT_SESSIONS=1
      - PREBOOT_CHROME=true
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:3000/json/version"]
      interval: 5s
      timeout: 10s
      retries: 3

  # Static server for iRacing HTML fixtures
  fixture-server:
    build:
      context: ./fixture-server
      dockerfile: Dockerfile
    ports:
      - "3456:80"
    volumes:
      - ../resources/iracing-hosted-sessions:/usr/share/nginx/html:ro
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:80/01-hosted-racing.html"]
      interval: 5s
      timeout: 10s
      retries: 3
```

#### Fixture Server Configuration

```dockerfile
# docker/fixture-server/Dockerfile
FROM nginx:alpine

# Configure nginx for static HTML serving
COPY nginx.conf /etc/nginx/conf.d/default.conf

EXPOSE 80
```

```nginx
# docker/fixture-server/nginx.conf
server {
    listen 80;
    server_name localhost;
    root /usr/share/nginx/html;
    
    location / {
        try_files $uri $uri/ =404;
        add_header Access-Control-Allow-Origin *;
    }
}
```

#### BDD Scenarios for Docker E2E

```gherkin
Feature: BrowserDevToolsAdapter Workflow Automation
  As the automation engine
  I want to execute the 18-step hosted session workflow
  So that I can verify browser automation against real HTML fixtures

  Background:
    Given the Docker E2E environment is running
    And the fixture server is serving iRacing HTML pages
    And the headless Chrome container is connected

  Scenario: Complete workflow navigation through all 18 steps
    Given the BrowserDevToolsAdapter is connected to Chrome
    When I execute step 2 HOSTED_RACING
    Then the adapter should navigate to the hosted racing page
    And the page should contain the create race button
    
    When I execute step 3 CREATE_RACE
    Then the wizard modal should open
    
    When I execute step 4 RACE_INFORMATION
    And I fill the session name field with "Test Race"
    Then the form field should contain "Test Race"
    
    # ... steps 5-17 follow same pattern
    
    When I execute step 18 TRACK_CONDITIONS
    Then the automation should stop at the safety checkpoint
    And the checkout button should NOT be clicked

  Scenario: Modal step handling - Add Car modal
    Given the automation is at step 8 SET_CARS
    When I click the "Add Car" button
    Then the ADD_CAR modal should open
    When I search for "Dallara F3"
    And I select the first result
    Then the modal should close
    And the car should be added to the selection

  Scenario: Form field validation with real selectors
    Given I am on the RACE_INFORMATION page
    Then the selector "input[name='sessionName']" should exist
    And the selector ".form-group:has label:has-text Session Name input" should exist
    
  Scenario: Error handling when element not found
    Given I am on a blank page
    When I try to click selector "#nonexistent-element"
    Then the result should indicate failure
    And the error message should contain "not found"
```

#### Test Implementation Structure

```typescript
// tests/e2e/docker/browserDevToolsAdapter.e2e.test.ts
import { describe, it, expect, beforeAll, afterAll } from 'vitest';
import { BrowserDevToolsAdapter } from '@infrastructure/adapters/automation/BrowserDevToolsAdapter';
import { StepId } from '@domain/value-objects/StepId';

describe('E2E: BrowserDevToolsAdapter - Docker Environment', () => {
  let adapter: BrowserDevToolsAdapter;
  const CHROME_WS_ENDPOINT = process.env.CHROME_WS_ENDPOINT || 'ws://localhost:9222';
  const FIXTURE_BASE_URL = process.env.FIXTURE_BASE_URL || 'http://localhost:3456';

  beforeAll(async () => {
    adapter = new BrowserDevToolsAdapter({
      browserWSEndpoint: CHROME_WS_ENDPOINT,
      defaultTimeout: 30000,
    });
    await adapter.connect();
  });

  afterAll(async () => {
    await adapter.disconnect();
  });

  describe('Step Workflow Execution', () => {
    it('should navigate to hosted racing page - step 2', async () => {
      const result = await adapter.navigateToPage(`${FIXTURE_BASE_URL}/01-hosted-racing.html`);
      expect(result.success).toBe(true);
    });

    it('should fill race information form - step 4', async () => {
      await adapter.navigateToPage(`${FIXTURE_BASE_URL}/03-race-information.html`);
      const stepId = StepId.create(4);
      const result = await adapter.executeStep(stepId, {
        sessionName: 'E2E Test Session',
        password: 'testpass123',
        description: 'Automated E2E test session',
      });
      expect(result.success).toBe(true);
    });

    // ... additional step tests
  });

  describe('Modal Operations', () => {
    it('should handle ADD_CAR modal - step 9', async () => {
      await adapter.navigateToPage(`${FIXTURE_BASE_URL}/09-add-a-car.html`);
      const stepId = StepId.create(9);
      const result = await adapter.handleModal(stepId, 'open');
      expect(result.success).toBe(true);
    });
  });

  describe('Safety Checkpoint', () => {
    it('should stop at step 18 without clicking checkout', async () => {
      await adapter.navigateToPage(`${FIXTURE_BASE_URL}/18-track-conditions.html`);
      const stepId = StepId.create(18);
      const result = await adapter.executeStep(stepId, {});
      expect(result.success).toBe(true);
      expect(result.metadata?.safetyStop).toBe(true);
    });
  });
});
```

---

### Strategy B: Native macOS E2E Tests

#### Purpose
Test OS-level screen automation using nut.js on real hardware. These tests CANNOT run in Docker because nut.js requires actual display access.

#### Requirements
- macOS CI runner with display access
- Screen recording permissions granted
- Accessibility permissions enabled
- Real Chrome/browser window visible

#### BDD Scenarios for Native E2E

```gherkin
Feature: NutJsAutomationAdapter OS-Level Automation
  As the automation engine
  I want to perform OS-level screen automation
  So that I can interact with iRacing without browser DevTools

  Background:
    Given I am running on macOS with display access
    And accessibility permissions are granted
    And screen recording permissions are granted

  Scenario: Screen capture functionality
    When I capture the full screen
    Then a valid image buffer should be returned
    And the image dimensions should match screen resolution

  Scenario: Window focus management
    Given a Chrome window titled "iRacing" is open
    When I focus the browser window
    Then the Chrome window should become the active window

  Scenario: Template matching detection
    Given I have a template image for the "Create Race" button
    And the iRacing hosted racing page is visible
    When I search for the template on screen
    Then the template should be found
    And the location should have confidence > 0.8

  Scenario: Mouse click at detected location
    Given I have detected a button at coordinates 500,300
    When I click at that location
    Then the mouse should move to 500,300
    And a left click should be performed

  Scenario: Keyboard input simulation
    Given a text field is focused
    When I type "Test Session Name"
    Then the text should be entered character by character
    With appropriate delays between keystrokes

  Scenario: Login state detection
    Given the iRacing login page is displayed
    When I detect the login state
    Then the result should indicate logged out
    And the login form indicator should be detected

  Scenario: Safe automation - no checkout
    Given I am on the Track Conditions step
    When I execute step 18
    Then no click should be performed on the checkout button
    And the automation should report safety stop
```

#### Test Implementation Structure

```typescript
// tests/e2e/native/nutJsAdapter.e2e.test.ts
import { describe, it, expect, beforeAll, afterAll } from 'vitest';
import { NutJsAutomationAdapter } from '@infrastructure/adapters/automation/NutJsAutomationAdapter';

describe('E2E: NutJsAutomationAdapter - Native macOS', () => {
  let adapter: NutJsAutomationAdapter;

  beforeAll(async () => {
    // Skip if not on macOS with display
    if (process.platform !== 'darwin' || !process.env.DISPLAY_AVAILABLE) {
      return;
    }
    
    adapter = new NutJsAutomationAdapter({
      mouseSpeed: 500,
      keyboardDelay: 25,
      defaultTimeout: 10000,
    });
    await adapter.connect();
  });

  afterAll(async () => {
    if (adapter?.isConnected()) {
      await adapter.disconnect();
    }
  });

  describe('Screen Capture', () => {
    it('should capture full screen', async () => {
      const result = await adapter.captureScreen();
      expect(result.success).toBe(true);
      expect(result.imageData).toBeDefined();
      expect(result.dimensions.width).toBeGreaterThan(0);
    });

    it('should capture specific region', async () => {
      const region = { x: 100, y: 100, width: 200, height: 200 };
      const result = await adapter.captureScreen(region);
      expect(result.success).toBe(true);
    });
  });

  describe('Window Focus', () => {
    it('should focus Chrome window', async () => {
      const result = await adapter.focusBrowserWindow('Chrome');
      // May fail if Chrome not open, which is acceptable
      expect(result).toBeDefined();
    });
  });

  describe('Template Matching', () => {
    it('should find element by template', async () => {
      const template = {
        id: 'test-button',
        imagePath: './resources/templates/test-button.png',
        confidence: 0.8,
      };
      const location = await adapter.findElement(template);
      // Template may not be on screen - test structure only
      expect(location === null || location.confidence > 0).toBe(true);
    });
  });
});
```

---

### Test File Structure

```
tests/
├── e2e/
│   ├── docker/                          # Docker-based E2E tests
│   │   ├── browserDevToolsAdapter.e2e.test.ts
│   │   ├── workflowSteps.e2e.test.ts
│   │   ├── modalHandling.e2e.test.ts
│   │   └── selectorValidation.e2e.test.ts
│   ├── native/                          # Native OS automation tests
│   │   ├── nutJsAdapter.e2e.test.ts
│   │   ├── screenCapture.e2e.test.ts
│   │   ├── templateMatching.e2e.test.ts
│   │   └── windowFocus.e2e.test.ts
│   ├── automation.e2e.test.ts           # Existing selector validation
│   └── features/                         # Gherkin feature files
│       └── hosted-session-automation.feature
├── integration/
│   └── infrastructure/
│       └── BrowserDevToolsAdapter.test.ts
└── unit/
    └── ...

docker/
├── docker-compose.e2e.yml              # E2E test environment
└── fixture-server/
    ├── Dockerfile
    └── nginx.conf

.github/
└── workflows/
    ├── e2e-docker.yml                  # Docker E2E workflow
    └── e2e-macos.yml                   # macOS native E2E workflow
```

---

### CI/CD Integration

#### Docker E2E Workflow

```yaml
# .github/workflows/e2e-docker.yml
name: E2E Tests - Docker

on:
  push:
    branches: [main, develop]
  pull_request:
    branches: [main]

jobs:
  e2e-docker:
    runs-on: ubuntu-latest
    
    steps:
      - uses: actions/checkout@v4
      
      - name: Setup Node.js
        uses: actions/setup-node@v4
        with:
          node-version: '20'
          cache: 'npm'
      
      - name: Install dependencies
        run: npm ci
      
      - name: Start Docker E2E environment
        run: |
          docker compose -f docker/docker-compose.e2e.yml up -d
          docker compose -f docker/docker-compose.e2e.yml ps
      
      - name: Wait for services to be healthy
        run: |
          timeout 60 bash -c 'until curl -s http://localhost:9222/json/version; do sleep 2; done'
          timeout 60 bash -c 'until curl -s http://localhost:3456/01-hosted-racing.html; do sleep 2; done'
      
      - name: Run Docker E2E tests
        run: npm run test:e2e:docker
        env:
          CHROME_WS_ENDPOINT: ws://localhost:9222
          FIXTURE_BASE_URL: http://localhost:3456
      
      - name: Stop Docker environment
        if: always()
        run: docker compose -f docker/docker-compose.e2e.yml down -v
```

#### macOS Native E2E Workflow

```yaml
# .github/workflows/e2e-macos.yml
name: E2E Tests - macOS Native

on:
  push:
    branches: [main, develop]
  pull_request:
    branches: [main]

jobs:
  e2e-macos:
    runs-on: macos-latest
    
    steps:
      - uses: actions/checkout@v4
      
      - name: Setup Node.js
        uses: actions/setup-node@v4
        with:
          node-version: '20'
          cache: 'npm'
      
      - name: Install dependencies
        run: npm ci
      
      - name: Grant screen recording permissions
        run: |
          # Note: GitHub Actions macOS runners have limited permission support
          # Some tests may be skipped if permissions cannot be granted
          sudo sqlite3 /Library/Application\ Support/com.apple.TCC/TCC.db \
            "INSERT OR REPLACE INTO access VALUES('kTCCServiceScreenCapture','com.apple.Terminal',0,2,0,1,NULL,NULL,0,'UNUSED',NULL,0,$(date +%s));" 2>/dev/null || true
      
      - name: Run native E2E tests
        run: npm run test:e2e:native
        env:
          DISPLAY_AVAILABLE: "true"
      
      - name: Upload screenshots on failure
        if: failure()
        uses: actions/upload-artifact@v4
        with:
          name: e2e-screenshots
          path: tests/e2e/native/screenshots/
```

---

### NPM Scripts

```json
{
  "scripts": {
    "test:e2e": "vitest run --config vitest.e2e.config.ts",
    "test:e2e:docker": "vitest run --config vitest.e2e.config.ts tests/e2e/docker/",
    "test:e2e:native": "vitest run --config vitest.e2e.config.ts tests/e2e/native/",
    "docker:e2e:up": "docker compose -f docker/docker-compose.e2e.yml up -d",
    "docker:e2e:down": "docker compose -f docker/docker-compose.e2e.yml down -v",
    "docker:e2e:logs": "docker compose -f docker/docker-compose.e2e.yml logs -f"
  }
}
```

---

### Environment Configuration

```bash
# .env.test.example
# Docker E2E Configuration
CHROME_WS_ENDPOINT=ws://localhost:9222
FIXTURE_BASE_URL=http://localhost:3456
E2E_TIMEOUT=120000

# Native E2E Configuration
DISPLAY_AVAILABLE=true
NUT_JS_MOUSE_SPEED=500
NUT_JS_KEYBOARD_DELAY=25
```

---

## Cross-References

- **[`ARCHITECTURE.md`](./ARCHITECTURE.md)** — Layer boundaries, port definitions, and dependency rules that guide test structure
- **[`TECH.md`](./TECH.md)** — Detailed tooling specifications (Vitest, Playwright, Testcontainers configuration)
- **[`package.json`](../package.json)** — Test scripts and commands (`test:unit`, `test:integration`, `test:e2e`, `test:coverage`)

---

## Summary

GridPilot's testing strategy ensures:
- **Business logic is correct** (unit tests for domain/application layers)
- **Infrastructure works reliably** (integration tests for repositories/adapters)
- **User workflows function end-to-end** (E2E tests for full stack)
- **Browser automation works correctly** (Docker E2E tests with real fixtures)
- **OS-level automation works correctly** (Native macOS E2E tests with display access)

### Hosted Session Automation Test Pyramid

For the iRacing hosted-session automation, confidence is provided by these concrete suites:

- **Domain/Application unit tests**
  - Entities and value objects such as [`AutomationSession`](packages/domain/entities/AutomationSession.ts:1), [`SessionState`](packages/domain/value-objects/SessionState.ts:1), [`CheckoutState`](packages/domain/value-objects/CheckoutState.ts:1), [`CheckoutConfirmation`](packages/domain/value-objects/CheckoutConfirmation.ts:1), and [`RaceCreationResult`](packages/domain/value-objects/RaceCreationResult.ts:1).
  - Use cases such as [`StartAutomationSessionUseCase`](packages/application/use-cases/StartAutomationSessionUseCase.ts:1), [`VerifyAuthenticatedPageUseCase`](packages/application/use-cases/VerifyAuthenticatedPageUseCase.ts:1), [`CompleteRaceCreationUseCase`](packages/application/use-cases/CompleteRaceCreationUseCase.ts:1), and [`ConfirmCheckoutUseCase`](packages/application/use-cases/ConfirmCheckoutUseCase.ts:1).

- **Infrastructure / automation integration tests**
  - Adapter wiring and fixture serving via [`FixtureServer.integration.test.ts`](tests/integration/infrastructure/FixtureServer.integration.test.ts:1).
  - Playwright lifecycle and overlay wiring such as [`BrowserModeIntegration.test.ts`](tests/integration/infrastructure/BrowserModeIntegration.test.ts:1) and automation-flow tests like [`CarsFlow.integration.test.ts`](tests/integration/infrastructure/automation/CarsFlow.integration.test.ts:1).
  - UI/overlay integration via [`renderer-overlay.integration.test.ts`](tests/integration/interface/renderer/renderer-overlay.integration.test.ts:1).

- **Fixture-based step E2E tests (per-step behavior)**
  - One test per wizard step under [`tests/e2e/steps`](tests/e2e/steps:1), all using the shared [`StepHarness`](tests/e2e/support/StepHarness.ts:1) and [`FixtureServer`](packages/infrastructure/adapters/automation/engine/FixtureServer.ts:1).
  - These validate DOM-level selectors / flows for each step (1–18) against fixture-backed HTML, and are considered canonical for step behavior.

- **Fixture-based workflow E2E tests (button → auth entry → automation → confirmation boundary)**
  - Workflow-focused tests under [`tests/e2e/workflows`](tests/e2e/workflows:1) that drive the `PlaywrightAutomationAdapter` + `WizardStepOrchestrator` across multiple steps using the fixture server.
  - Example: [`steps-07-09-cars-flow.e2e.test.ts`](tests/e2e/workflows/steps-07-09-cars-flow.e2e.test.ts:1) exercises cross-step cars flow, while [`full-hosted-session.workflow.e2e.test.ts`](tests/e2e/workflows/full-hosted-session.workflow.e2e.test.ts:1) runs a full 1–18 workflow via [`MockAutomationEngineAdapter`](packages/infrastructure/adapters/automation/engine/MockAutomationEngineAdapter.ts:1) and [`StartAutomationSessionUseCase`](packages/application/use-cases/StartAutomationSessionUseCase.ts:1), asserting final `SessionState` and step position.
  - Additional workflow scenarios cover mid-flow failure using [`MockBrowserAutomationAdapter`](packages/infrastructure/adapters/automation/engine/MockBrowserAutomationAdapter.ts:1), ensuring failure states and diagnostics are surfaced without emitting false confirmations.

- **Opt-in real-world automation smoke tests (true iRacing / NutJs)**
  - The legacy real iRacing automation suite [`automation.e2e.test.ts`](tests/e2e/automation.e2e.test.ts:1) is treated as a smoke-only, opt-in layer.
  - It is gated by `RUN_REAL_AUTOMATION_SMOKE=1` and should not run in normal CI; it exists to validate that NutJs / template-based automation still matches the real UI for a small number of manual smoke runs on a prepared macOS environment.

#### Confidence expectations

- For **normal changes** to hosted-session automation (selectors, step logic, overlay behavior, authentication, or confirmation flows), the following suites must pass to claim "high confidence":
  - All relevant **unit tests** in `tests/unit` that touch the changed domain/use-case code.
  - All relevant **integration tests** in `tests/integration` for the affected adapters.
  - All **step E2E tests** under [`tests/e2e/steps`](tests/e2e/steps:1).
  - All **workflow E2E tests** under [`tests/e2e/workflows`](tests/e2e/workflows:1).

- The **real-world smoke suite** in [`tests/e2e/automation.e2e.test.ts`](tests/e2e/automation.e2e.test.ts:1) is an additional, manual confidence layer and should be run only on configured machines when validating large changes to NutJs automation, template packs, or iRacing UI assumptions.

- When adding new behavior:
  - Prefer **unit tests** for domain/application changes.
  - Add or extend **integration tests** when introducing new adapters or external integration.
  - Add **step E2E tests** when changing DOM/step behavior for a specific wizard step.
  - Add or extend **workflow E2E tests** when behavior spans multiple steps, touches authentication/session lifecycle, or affects confirmation/checkout behavior end-to-end.

By following BDD principles and maintaining clear test organization, the team can confidently evolve GridPilot while preserving correctness and stability, with a dedicated, layered confidence story for hosted-session automation.