Skip to content

PlayMonkey v0.4.0

Choose a tag to compare

View all tags
@github-actions github-actions released this 13 Oct 08:19
· 30 commits to main since this release
v0.4.0
f288827
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

The first tagged release: milestones M1 to M4 of the roadmap.

Monkey and findings (M1)

  • monkey run: seeded random monkey on Playwright, with stable element refs, stayOnDomain and limits per run.
  • JSON config with schema validation, ${VAR} secrets from the environment, masked in all output.
  • Form login and saved sessions (storageState), several roles, one browser context each.
  • Oracles: uncaught exceptions, console errors, HTTP 4xx and 5xx, crashes, hangs, error pages, horizontal
    overflow; findings deduplicated by key, with the last actions, a screenshot and ignore rules.
  • Action log (actions.json) and monkey replay, which repeats a run without randomness or LLM.
  • report.json and a self-contained report.html; trace and video per run; exit codes 0, 1, 2.

LLM for stuck situations (M2, M3)

  • Backends anthropic-foundry (Claude on Microsoft Foundry) and openai (Responses API or Chat Completions on any
    OpenAI-compatible endpoint), behind one provider interface.
  • Stuck detection, handoff, fallback to the monkey, circuit breaker, budgets (llm.maxCalls, llm.maxTokens);
    secrets reach the model only as placeholders.
  • monkey preflight and a preflight before every run; token and estimated cost accounting in the report.
  • Spectator mode: --watch, --slow-mo.

Real-world pages (M3)

  • Cookie banners and overlays, native dialogs, same-origin iframes, popups and tabs (monkey.maxTabs), file
    uploads, downloads with the download oracle.
  • More actions: double-click, right-click, hover, long press, drag; action log version 2 (version 1 still read).

Comfort (M4)

  • Parallel runs: seeds (one run per seed and role, in <role>/<seed>/) and parallel; one combined report with
    findings aggregated across runs; shared LLM budget and circuit breaker; a failed run does not stop the others.
  • CI example workflow (.github/workflows/monkey.yml), how-to guides for CI, parallel runs and reproducing
    findings, CONTRIBUTING.md.
  • Install from git as a dev dependency: npm install --save-dev github:besessener/PlayMonkey#v0.4.0.
Assets 2
Loading