@taqwright/lime-cli 0.15.3 → 0.16.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +58 -2
- package/dist/index.js +266 -179
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -272,8 +272,15 @@ lime-cli check # last line: LIME_CHECK_RESULT: PASS | FAIL
|
|
|
272
272
|
`lime-cli check` checks everything that changed since the base branch (committed, uncommitted
|
|
273
273
|
and new files, never `.lime/`) against the current recording: every locator the change adds was
|
|
274
274
|
recorded on the device, every changed spec calls page objects rather than the device, and the
|
|
275
|
-
change holds a test at all
|
|
276
|
-
|
|
275
|
+
change holds a test at all, and every changed spec asserts something (a spec that ends at a tap
|
|
276
|
+
passes with the feature reverted). It names the recording it used (goal and time). A hand-written
|
|
277
|
+
test fails it; `--diff <file>` takes a unified diff for CI.
|
|
278
|
+
|
|
279
|
+
**A deliberately flat suite** declares it with a `lime.json` at the repo root:
|
|
280
|
+
`{ "pageObjects": false }`. Then `check` and `check-pom` skip the page-object rule (and say so),
|
|
281
|
+
`export`/`adapt` brief the agent to write flat specs in the suite's own style, and doctor stops
|
|
282
|
+
flagging direct device use. It is the repo owner's declaration, committed and reviewed; the
|
|
283
|
+
recording rule and the assertion rule still apply.
|
|
277
284
|
|
|
278
285
|
`--lime-dir` writes `.lime/` next to your project: the recorded steps the agent converts
|
|
279
286
|
from, a `run.json` of what was recorded (goal, platform, cloud device), and the briefs.
|
|
@@ -328,6 +335,18 @@ flow on a cloud device, adapts the test into your **test-suite** repo, verifies
|
|
|
328
335
|
and opens a pull request there linking back. LIME holds no GitHub credential — the runner
|
|
329
336
|
already has yours.
|
|
330
337
|
|
|
338
|
+
Where `lime-cli pr` has commented with **suggested tests**, pick one by number instead of typing
|
|
339
|
+
a goal:
|
|
340
|
+
|
|
341
|
+
```
|
|
342
|
+
/automate 2
|
|
343
|
+
```
|
|
344
|
+
|
|
345
|
+
The goal comes from LIME's analysis of that pull request, never from the comment. A suggestion
|
|
346
|
+
behind a feature flag is held back until a test user for it exists. **Every reply on the pull
|
|
347
|
+
request** comes from the **Taqwright GitHub App**: the 👀, "Automating…", and the result. The
|
|
348
|
+
workflow never comments with its own token, so the templates only need `pull-requests: read`.
|
|
349
|
+
|
|
331
350
|
On **GitLab CI/CD**, `docs/gitlab/automate.yml` is the same loop, started from a manual job
|
|
332
351
|
rather than a comment — GitLab CI cannot be triggered by a comment from `.gitlab-ci.yml`
|
|
333
352
|
alone. See [GitLab CI/CD](#gitlab-cicd) below.
|
|
@@ -451,6 +470,43 @@ matrix, other-runner and other-CI variants commented in. For GitLab, `docs/gitla
|
|
|
451
470
|
is the equivalent — mind the `--run-id` note in [GitLab CI/CD](#gitlab-cicd), which is the one
|
|
452
471
|
difference that degrades the board without saying so.
|
|
453
472
|
|
|
473
|
+
### `pr` — what a pull request means for your UI tests
|
|
474
|
+
|
|
475
|
+
Run on the **app** repository's pull requests. LIME traces the change through your app's map
|
|
476
|
+
and answers, as one comment on the pull request:
|
|
477
|
+
|
|
478
|
+
- **Run these**: the tests on screens the change reaches.
|
|
479
|
+
- **Needs update**: tests that assert text this pull request changes (`$` → `SGD`), so an
|
|
480
|
+
intended change is not mistaken for a regression.
|
|
481
|
+
- **Not automated yet**: planned scenarios on the journeys it touches, and journeys with none.
|
|
482
|
+
- **Behind a feature flag**: flags the pull request adds. A test for one needs a test user the
|
|
483
|
+
flag is on for.
|
|
484
|
+
|
|
485
|
+
```yaml
|
|
486
|
+
on: pull_request
|
|
487
|
+
permissions:
|
|
488
|
+
contents: read
|
|
489
|
+
jobs:
|
|
490
|
+
lime:
|
|
491
|
+
runs-on: ubuntu-latest
|
|
492
|
+
steps:
|
|
493
|
+
- uses: actions/checkout@v4
|
|
494
|
+
with: { fetch-depth: 0 }
|
|
495
|
+
- run: npx -y @taqwright/lime-cli pr --project "Shop" --app-dir mobile
|
|
496
|
+
env:
|
|
497
|
+
LIME_CI_TOKEN: ${{ secrets.LIME_CI_TOKEN }}
|
|
498
|
+
```
|
|
499
|
+
|
|
500
|
+
The comment is posted by the **Taqwright GitHub App**, never by your workflow, so every pull
|
|
501
|
+
request sees one author. Install the App on the app repository and connect it in LIME →
|
|
502
|
+
Integrations. Until then the comment goes to the job summary instead.
|
|
503
|
+
|
|
504
|
+
The screens come from the app map you published from VS Code; `pr` keeps the import graph
|
|
505
|
+
current from the pull request's own checkout. What leaves the runner is file names, the part of
|
|
506
|
+
the import graph that can reach them, changed strings and source diffs with credential-looking
|
|
507
|
+
values masked. Credential files are never read. It always exits 0: a fork without secrets or a
|
|
508
|
+
shallow checkout is reported and skipped, never a failed build.
|
|
509
|
+
|
|
454
510
|
## Server API contract
|
|
455
511
|
|
|
456
512
|
`lime-cli` is the only client of the LIME server's `/api/cli/*` endpoints, so
|