Test Your LLM Like You Test Your UI

LLMAssert adds five LLM-powered matchers to Playwright to check hallucinations, PII, tone, format, and semantic accuracy.

Matchers call a judge model, return { pass, score, reasoning }, and produce a numeric score from 0.0 to 1.0 or null if inconclusive.

When the judge is unavailable tests are marked inconclusive and treated as passing to avoid blocking CI.

The matchers are asynchronous and require await; setup requires installing @llmassert/playwright and setting OPENAI_API_KEY.