wallaby-cli
Wallaby keeps JavaScript, TypeScript, and Python tests live and queryable throughout a coding task. It runs affected tests as files change and keeps current results, coverage, and execution data available, so an agent can inspect what the code did instead of reconstructing it from terminal output.
Wallaby does not replace the project's existing test setup, whether it uses Vitest, Jest, Jasmine, Mocha, Angular's ng test, pytest, Python's unittest, or another supported framework. It runs the existing tests through their framework and augments them with live affected-test execution, retained queryable state, coverage artifacts, execution traces, and runtime inspection for coding agents. For test execution and analysis during a coding task, use Wallaby instead of invoking those frameworks directly. Run a conventional test command such as npm test, npx vitest, or ng test only when the user explicitly asks to run it directly.
Use that live test state to:
- Establish a baseline before editing by checking current failures, coverage, and the tests that exercise the code you plan to change.
- Keep feedback focused while editing. Start with relevant tests, add related tests as the change surface grows, and use project-wide verification at the end when the task requires it.
- Locate coverage gaps across a batch of files in one consolidated
coverage-gaps.mdreport. It lists uncovered lines and partially covered expression ranges per file and links to full per-file coverage and related-test artifacts when deeper inspection is needed. - Read coverage directly beside the complete source in a
.wcovartifact. Every coverable line is markedfull,partial, ornone, and partially covered lines identify the exact uncovered column ranges and expressions. Filter the same view to one test to see only what that test executed and missed. - Analyze a batch of files, one source or test file, or an exact source location from a compact summary, then follow its separate coverage and related-test artifacts only when needed. Use coverage, change risk, test statuses, and timings to decide what to change and how broadly to verify it.
- Analyze one executed test as a unified execution record. Follow recorded source lines in execution order across every file involved. The trace includes imported modules and setup, marks the start of the selected test, and continues through each source line the test reaches. Combine it with per-file test-scoped
.wcovartifacts to see both the route taken and the exact lines and expressions executed or missed. The same report provides the test's status, timing, errors, logs, and covered files as supporting diagnostics. - Inspect multiple variables or expressions across different source locations in one request. Each value is captured in every test context that reaches its location and tied to the test that produced it. Filter the combined results to one test when narrowing the investigation. Use this runtime evidence before changing code or adding temporary logs.
- Use project-wide coverage and test and file metrics such as timing, test count, complexity, and change risk to identify meaningful test gaps, slow or tightly coupled tests, and changes that need wider verification.
When a command produces a report, start with its concise Markdown output, then open linked reports and artifacts only when the current question needs more context. Wallaby saves every generated report and artifact to a timestamped directory on disk, so earlier results remain available as historical snapshots when comparison is useful.