Claude Fable 5.1 & GPT-6 Astra packages are live

Testing

13 packages, each available for 11 model families.

accessibility

Testing for accessibility — what automation catches, what it cannot, and the keyboard and screen-reader checks that find real barriers.

11 models

e2e

End-to-end tests that are worth their cost — user-visible selectors, deterministic waiting, and keeping the suite small enough to trust.

11 models

go-testing

Testing in Go with the standard toolchain — table-driven subtests, t.Helper, httptest and fstest, golden files, benchmarks, the race detector,…

11 models

integration

Testing across real boundaries — real databases in containers, isolation between tests, and faking third parties without faking their behaviour.

11 models

load

Load testing that predicts production behaviour — modelling real traffic, measuring percentiles not averages, and finding the knee before users do.

11 models

performance

Testing performance as a regression gate — budgets in CI, lab versus field data, and measuring the metrics users actually feel.

11 models

pytest

A pytest suite that stays fast and trustworthy — layout, fixtures with the right scope, parametrize instead of loops, markers that mean something,…

11 models

regression

Preventing bugs from returning — the failing test that must come first, choosing what to keep, and stopping a suite from becoming archaeology.

11 models

security

Testing for security defects — what SAST, DAST and dependency scanning each catch, and the abuse-case tests only a human writes.

11 models

stress

Pushing a system past its limit deliberately — finding the breaking point, verifying it degrades gracefully, and confirming it recovers.

11 models

test-strategy

Deciding what to test and at which level — the shape of a suite, what to do with legacy code, and the signals that a strategy is failing.

11 models

unit

Writing unit tests that catch regressions without cementing implementation — what to assert, what to fake, and why most flaky tests are design…

11 models

visual

Visual regression testing that is not permanently red — deterministic rendering, masking dynamic content, and reviewing diffs instead of regenerating…

11 models

Other categories