Testing Strategy (Vitest & Playwright)

Relevant source files

The following files were used as context for generating this wiki page:

• package.json
• CLAUDE.md
• tests/e2e/api.test.ts
• tests/e2e/chat.test.ts
• lib/ai/prompts.test.ts
• lib/db/schema.test.ts
• lib/db/lv-queries.test.ts
• app/(chat)/lv/import-actions.test.ts/lv/import-actions.test.ts)
• app/(chat)/lv/preisspiegel-actions.test.ts/lv/preisspiegel-actions.test.ts)
• lib/compliance/daech.test.ts

The LV.AI testing strategy ensures the reliability of AI-driven construction management tools. It combines Vitest for fast unit and integration testing with Playwright for comprehensive end-to-end (E2E) validation of the user interface and API workflows.

This strategy enforces GAEB-conformity, D/A/CH region compliance (VOB/ÖNORM/SIA), and data integrity across chat-based project interactions.

Sources: CLAUDE.md:19-27, package.json:20-22

Testing Architecture

The project employs a tiered testing approach:

1. Unit Tests (Vitest): Validate individual functions, AI prompts, and database schemas.
2. Integration Tests (Vitest): Verify Server Actions and database queries using mocks for external dependencies.
3. End-to-End Tests (Playwright): Simulate real user sessions, verifying the chat interface, file uploads, and API responses.

Test Execution Commands

Command	Purpose
pnpm test	Runs all Vitest unit and integration tests.
pnpm test:watch	Starts Vitest in interactive watch mode.
pnpm test:e2e	Executes Playwright E2E tests in a simulated browser environment.

Sources: package.json:20-22

Unit and Integration Testing (Vitest)

Vitest handles logic-heavy components like prompt construction, database schema integrity, and regional compliance modules.

Prompt and AI Logic

Tests verify that AI prompts correctly include identity markers, GAEB formats (X81-X86), and D/A/CH legal frameworks.

• regularPrompt must contain "LV.AI", "AVA", and regional references like "VOB/A" or "SIA 118".
• systemPrompt tests ensure dynamic injection of location hints and tool support.

Sources: lib/ai/prompts.test.ts:10-50, lib/ai/prompts.test.ts:77-100

Database and Query Testing

The system validates the Drizzle ORM schema and complex SQL queries.

• Schema Validation: Ensures all Enums (e.g., landEnum, gaebFormatEnum) and tables (e.g., lvDocument, offer) are correctly defined.
• Query Mocking: Integration tests for lv-queries.ts use vi.hoisted to mock database responses, allowing verification of CRUD operations without a live database connection.

Sources: lib/db/schema.test.ts:32-60, lib/db/lv-queries.test.ts:9-40

Server Actions

Integration tests for actions like importLvFileAction and generatePreisspiegelAction mock authentication and external API calls.

• Import Logic: Validates normalization of GAEB Position numbers (OZ) and handling of session-less requests.
• Preisspiegel Logic: Verifies that at least two valid offers are required to generate a comparison.

Sources: app/(chat)/lv/import-actions.test.ts:60-80, app/(chat)/lv/preisspiegel-actions.test.ts:80-100

The diagram shows the organization of Vitest tests and their reliance on mocking for external dependencies.

End-to-End Testing (Playwright)

Playwright tests the application from the user's perspective, focusing on the Chat interface and API stability.

Chat Interface Workflow

Tests in tests/e2e/chat.test.ts verify:

• Initial visibility of the multimodal-input and send-button.
• Support for multiline text input.
• Functionality of the "Stop" button during AI generation.
• Behavior of suggested actions on empty chat states.

Sources: tests/e2e/chat.test.ts:3-40

API and Navigation

Tests in tests/e2e/api.test.ts simulate API interactions:

1. Message Flow: Sending a message, verifying the "assistant" role appears, and checking for content.
2. Dynamic Routing: Confirming that the URL redirects to a specific /chat/:id after the first message.
3. Error Handling: Intercepting API calls to return a 500 status and verifying the UI displays an error toast.

Sources: tests/e2e/api.test.ts:6-60

The sequence diagram illustrates the standard Playwright test flow for a successful chat interaction.

Specialized AVA Validation

The project includes specific testing procedures for construction-specific logic (AVA). These are outlined as "LV.AI-Prüfverfahren" (Validation Procedures).

• GAEB Conformity: Every change to import/export logic requires X81–X86 schema validation.
• Regional Compliance: Logic in daech.test.ts ensures that requesting "DE" returns "VOB", "AT" returns "ÖNORM", and "CH" returns "SIA".
• Data Integrity: Changes to offer comparison queries require testing against realistic multi-bidder datasets to prevent calculation errors in the Preisspiegel.

Sources: CLAUDE.md:83-95, lib/compliance/daech.test.ts:13-35

Compliance Mappings

Region	Standard (Vergabe)	Vertrag
DE	VOB/A	VOB/B
AT	ÖNORM B2110	ÖNORM B2111
CH	SIA 118	SIA 102

Sources: CLAUDE.md:73-80, lib/compliance/daech.test.ts:38-55

Summary

The LV.AI testing strategy provides a robust safety net for complex construction procurement workflows. Vitest ensures that the underlying legal and AI prompt logic remains correct across regional variations, while Playwright guarantees a seamless user experience for construction professionals. The enforcement of mandatory GAEB and D/A/CH compliance checks during the testing phase maintains the high technical standards required for construction industry software.