← 插件

Unit Test Pulse

Live unit test progress, failures, stack traces, and logs alongside your Paseo agents.

$ paseo plugin add mbranicky/unit-test-pulse
报告问题
mbranicky侧边栏1次安装fff899b51c50源代码
Unit Test Pulse 截图 1Unit Test Pulse 截图 2Unit Test Pulse 截图 3Unit Test Pulse 截图 4Unit Test Pulse 截图 5Unit Test Pulse 截图 6Unit Test Pulse 截图 7Unit Test Pulse 截图 8Unit Test Pulse 截图 9

该插件的概览尚未翻译为你的语言,因此以英文显示。

Unit Test Pulse follows JavaScript and TypeScript unit tests beside your Paseo agents. The workspace panel shows live results grouped by file, failure messages, stack traces, assertion details, current source previews, and captured logs. A header indicator keeps execution visible when the panel is closed.

Version 1.0.0 integrates with Vitest 4/5, Jest 30, Mocha 11/12, Bun 1.4.3+, Jasmine 6/7, and Node.js 22.18+/24 built-in tests through a local helper. Vitest/Jest/Mocha/Jasmine use the tested project's installed runner; Node tests use the helper's Node executable without installing a framework. Bun uses an existing executable from the project or PATH, or an explicitly configured path. The panel's optional agent skill explains how to launch monitored tests while retaining the project's normal environment and prerequisites. Existing uninstrumented commands cannot be attached after launch.

For runners with native watch support, each watch cycle is a separate result, with an idle state between cycles. Concurrent commands retain independent identities. Filters and search narrow the displayed tests while run counts remain complete for the captured execution. Failure reports and paged output can be copied for debugging.

Node tests retain original assertion stacks and expected/actual values, with source previews for test(), it(), and context subtests. Existing native reporters, loader settings, and test selection are preserved. Node watch cycles contain only files re-executed by the runner.

Mocha runs through its native CLI with the existing reporter, configuration and loaders preserved. Failures, hook errors, retries, skipped/pending tests, and BDD/TDD source previews are captured. Parallel results arrive at file completion; Mocha process output remains at run scope. Dry runs are shown as no executed tests.

Bun 1.4.x uses a private loopback inspector for live results, retries, source previews, concurrent tests, and watch cycles. Its errors and console messages stay in Run output because they lack test identities; some assertions provide no stack frames. Native parallel workers, hot reload, and other inspector sessions are not supported by this adapter. Existing Bun configuration, environment files, preloads, JUnit output, and exit behavior are preserved.

Jasmine retains its native configuration, helpers, reporters, randomization, filters, parallel execution, and exit policy. Spec failures include messages and stacks; suite/global failures stay visible. Pending, excluded, and not-applicable results appear as skipped. Runner-supplied debug logs and pending reasons can be test-scoped; ordinary console output remains in Run output. Jasmine has no native watch mode, so this integration uses one-shot runs.

Vitest console records can carry test identity. Jest console buffers are file-scoped when supplied, and Node.js output records are file-scoped, while process stdout/stderr stays associated with the run. Runner settings can suppress output. Capture limits and unavailable details are shown explicitly.

SDK compatibility is checked against Paseo 0.11.2 and 0.12.0; desktop monitoring is verified on Paseo 0.12.0 on macOS. The README includes real 0.12.0 desktop captures of live results, diagnostics, source previews, logs, setup, recent runs, settings, and Ghostty/Light themes. Runner checks cover Node 22.18.0/24.16.0, Vitest 4.1.11/5.0.4, Jest 30.5.2, Mocha 11.8.0/12.0.3, Bun 1.4.3, and Jasmine 6.3.0/7.0.0. Other platforms, browser-mode tests, and complex multi-project configurations need further verification.

Results and logs are stored on the daemon host under ~/.local/state/unit-test-pulse. The plugin does not upload them. Logs can contain application secrets; copying/exporting shares the selected text, and an agent using the optional skill can read it and send it to its configured provider. History and capture policies are currently fixed; advanced reporting and adapters for other independent runners are planned.