Tutorial Natural Language Tests

Tutorial: Natural Language Tests

One English sentence → Playwright spec. Requires an LLM (no rule fallback).

Deep: Tutorial 03.

1. Configure LLM

# .env
LLM_PROVIDER=anthropic          # or openai | azure | google | ollama
ANTHROPIC_API_KEY=sk-ant-...
LLM_MODEL=claude-sonnet-4       # provider-specific
# Local free option:
# LLM_PROVIDER=ollama
# OLLAMA_BASE_URL=http://localhost:11434

2. Create

zyvor-qa create "Verify the homepage shows the Schedule Demo button and all product names"

Pipeline: NL → structured requirements (tests/fixtures/requirements.json) → generator → tests/generated/nl-….spec.ts.

Tip: name the route and exact visible text in the sentence (/vm, “KubeVirt”) so the model emits goto + getByText.

3. Create and run

zyvor-qa create "Check that /guestkit loads and mentions Offline VM intelligence" --execute

4. Keep permanently

• Promote file into tests/manual/, or
• Rewrite as a versioned markdown spec (Spec to Test)

Mission Control also has an AI test action card for the same flow.

Next

Spec to Test · CLI Commands · Configuration