Linearbytes releases Bug0 QA Agent for cached Playwright regression testing

Bug0 QA Agent lets developers write declarative test steps in plain English inside standard Playwright specs. On the first run, the agent inspects the accessibility tree and screenshots to understand the interface and carry out each step.

Once a single-action step succeeds, the framework compiles it into concrete browser operations and stores them in an in-memory Redis cache. Later runs replay those cached actions through Playwright. If a replayed action fails because the interface has changed, the library invalidates the cache and reruns the agent loop instead of failing the whole run; multi-action directives always run through the agent.

Assertions are not cached and are checked with vision models on every pass. Claude and Gemini each evaluate an assertion, and a third model breaks ties. The library also includes disposable email through mailsink and community adapters such as Mailinator, so tests can extract one-time passwords and magic links. The code is released under the Functional Source License 1.1, with an ALv2 future license, which is source-available rather than open source and includes non-compete restrictions for the first two years.

Teams testing web interfaces can cut latency and token costs on stable regression suites, because cached single-action steps skip LLM inference on replay.

Assertions still call vision models on every run, so the per-run cost and latency of verification remain and should be budgeted for.

The non-compete terms of the Functional Source License restrict some commercial uses during the first two years, so check them before building on the library.