diff --git a/.changeset/acp-executor.md b/.changeset/acp-executor.md deleted file mode 100644 index 271c22bc82..0000000000 --- a/.changeset/acp-executor.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"@e2e-dev/acp": minor ---- - -**Release:** New package: run `agent.act` and `agent.assert` on your own coding agent over the Agent Client Protocol, signed in with the agent's own login. `claudeCode()` and `codex()` start those agents with e2e's step tools as their only tools; `acp({ command, args })` starts any other ACP agent. Every action goes through the runner, so steps replay from the cache without the agent, and a tool of the agent's own that runs without asking fails the step with `POLICY_DENIED`. diff --git a/.changeset/acp-fresh-agent-after-tail-tool.md b/.changeset/acp-fresh-agent-after-tail-tool.md deleted file mode 100644 index 5a6093f055..0000000000 --- a/.changeset/acp-fresh-agent-after-tail-tool.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"@e2e-dev/acp": patch ---- - -A step that starts right after a step that timed out now gets a fresh agent when the old agent runs a tool of its own as it stops. Before, the step failed with `MODEL_PROVIDER_FAILED` ("ACP connection closed"), because it reused the session that tool had closed. diff --git a/.changeset/agent-drag-at.md b/.changeset/agent-drag-at.md deleted file mode 100644 index c39a9d01e8..0000000000 --- a/.changeset/agent-drag-at.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -Add `drag_at` to the agent's pixel tools, in `agent.act` and `e2e mcp`, and `dragAt(from, to)` to `ExecutorActions`: a drag from one screenshot point to another, for a shape on a canvas or a pin on a map. Like `tap_at`, the bare points go to the engine's `dragTo` pointer kind, where the model aimed; an engine without it drags by id when the points land on two different listed controls, like `drag`. The replay cache records both points and replays each the way it replays a `tap_at` point. diff --git a/.changeset/alert-iso-timestamp.md b/.changeset/alert-iso-timestamp.md deleted file mode 100644 index 7d607390bf..0000000000 --- a/.changeset/alert-iso-timestamp.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -A replay no longer fails its end check on an alert that shows an ISO 8601 timestamp, such as `Saved at 2026-10-24T15:41:27.955Z`, when the time differs from the recording. diff --git a/.changeset/app-recording.md b/.changeset/app-recording.md deleted file mode 100644 index 094aaf78f3..0000000000 --- a/.changeset/app-recording.md +++ /dev/null @@ -1,16 +0,0 @@ ---- -"e2e": patch ---- - -`app.startRecording()` and `app.stopRecording()` record one span of the screen, such as a single transition, on the web and on devices. `stopRecording` returns the videos: a file `path`, or a `url` when a hosted browser or device keeps the recording. A start fails while the test's `video` records, so set `{ video: 'off' }` on the test. - -```ts -test('the dialog opens', { video: 'off' }, async ({ app, screen }) => { - await app.open(); - await app.startRecording({ name: 'dialog' }); - await screen.getByRole('button', 'Open').tap(); - const [video] = await app.stopRecording(); -}); -``` - -An engine's `startVideo` and `stopVideo` may now be called more than once per attempt, never overlapping. diff --git a/.changeset/before-action-observer.md b/.changeset/before-action-observer.md deleted file mode 100644 index f1bad84df6..0000000000 --- a/.changeset/before-action-observer.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -'e2e': patch ---- - -Add `onBeforeAction` to `agent.act`, a hook that runs before each action, so a test can capture the screen an action started on. diff --git a/.changeset/booted-device-skips-held.md b/.changeset/booted-device-skips-held.md deleted file mode 100644 index f85268dbd7..0000000000 --- a/.changeset/booted-device-skips-held.md +++ /dev/null @@ -1,6 +0,0 @@ ---- -"@e2e-dev/mobile": patch -"e2e": patch ---- - -A mobile target without `device` skips a booted simulator or emulator that another agent-device session holds, such as one another tool is driving, and drives the free ones. When every booted device is held, the run fails at `prepare` naming the holding sessions instead of failing on the first one. diff --git a/.changeset/browserstack-provider.md b/.changeset/browserstack-provider.md deleted file mode 100644 index b87de197e8..0000000000 --- a/.changeset/browserstack-provider.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -'@e2e-dev/browserstack': minor ---- - -**Release:** New package: `@e2e-dev/browserstack`. `browserstack()` leases BrowserStack App Automate real Android and iOS devices for the mobile engine, one isolated agent-device daemon per worker slot. An `https` `app` is uploaded once per run, an attempt's video links BrowserStack's recording of the session, and a release fails the run when agent-device can't confirm it ended the session. diff --git a/.changeset/bump-agent-device-0-21-33.md b/.changeset/bump-agent-device-0-21-33.md deleted file mode 100644 index c2f5ef2ae1..0000000000 --- a/.changeset/bump-agent-device-0-21-33.md +++ /dev/null @@ -1,7 +0,0 @@ ---- -"@e2e-dev/mobile": patch -"@e2e-dev/limrun": patch -"@e2e-dev/testmu": patch ---- - -Bump agent-device to 0.21.33. diff --git a/.changeset/bump-agent-device-0-21-35.md b/.changeset/bump-agent-device-0-21-35.md deleted file mode 100644 index f17c5af687..0000000000 --- a/.changeset/bump-agent-device-0-21-35.md +++ /dev/null @@ -1,7 +0,0 @@ ---- -'@e2e-dev/mobile': patch -'@e2e-dev/limrun': patch -'@e2e-dev/testmu': patch ---- - -Update agent-device and @agent-device/testmu to 0.21.35. An unscoped boot picks the newest of several shut-down devices, and on iOS, pressing return with nothing focused keeps the runner session. diff --git a/.changeset/bump-deps-2026-10-25.md b/.changeset/bump-deps-2026-10-25.md deleted file mode 100644 index 31ddffff88..0000000000 --- a/.changeset/bump-deps-2026-10-25.md +++ /dev/null @@ -1,9 +0,0 @@ ---- -"e2e": patch -"@e2e-dev/decision": patch -"@e2e-dev/mobile": patch -"@e2e-dev/limrun": patch -"@e2e-dev/testmu": patch ---- - -Bump agent-device to 0.21.31, zod to 4.6.7, @modelcontextprotocol/server to 2.4.0, and oxc to 0.155.0. diff --git a/.changeset/cleanup-failure-exit-code.md b/.changeset/cleanup-failure-exit-code.md deleted file mode 100644 index eec9c8e1eb..0000000000 --- a/.changeset/cleanup-failure-exit-code.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -A run now exits with the cleanup error's code, usually 3, when a test's cleanup failed. The list reporter shows the cleanup failure in the summary and under the test for passed and flaky tests, so a CI job that goes red names the cleanup failure instead of reporting a green summary. diff --git a/.changeset/cli-empty-reporter.md b/.changeset/cli-empty-reporter.md deleted file mode 100644 index 326f448a2d..0000000000 --- a/.changeset/cli-empty-reporter.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -`e2e run` and `e2e explore` refuse an empty `--reporter` (`--reporter ''`, `--reporter ','`, or `--reporter "$REPORTERS"` with the variable unset) as a usage error, exit 2, instead of running with no reporters and printing nothing. diff --git a/.changeset/cli-exit-when-done.md b/.changeset/cli-exit-when-done.md deleted file mode 100644 index 49f9306342..0000000000 --- a/.changeset/cli-exit-when-done.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -An `e2e` command whose config or test file leaves a timer or socket open exits two seconds after it finishes, instead of staying open. diff --git a/.changeset/device-agent-device.md b/.changeset/device-agent-device.md deleted file mode 100644 index c24ff569ea..0000000000 --- a/.changeset/device-agent-device.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"@e2e-dev/mobile": patch ---- - -Add the unstable `device.unstable_raw`, which may change or go away in any release: the test's agent-device session name, the device's id and name, and an agent-device `client` bound to that session, for what the `device` methods don't cover. Client calls aren't recorded as steps, replayed from the agent cache, or redacted. diff --git a/.changeset/drop-pixel-taint.md b/.changeset/drop-pixel-taint.md deleted file mode 100644 index 703233b52b..0000000000 --- a/.changeset/drop-pixel-taint.md +++ /dev/null @@ -1,6 +0,0 @@ ---- -'e2e': minor -'@e2e-dev/decision': patch ---- - -**Breaking:** a secret fill no longer turns off screenshots for the model. `agent.assert(..., { vision: 'only' })`, `vision: true`, the `screenshot` and point tools, `e2e mcp`, and `e2e explore` findings keep working after a sign-in, in serial groups, and in tests that restore a saved session. Screenshots still mask secure fields, and secret values are still redacted from all text. A generic secret filled into a plain input, or a value the app shows outside a secure field, such as behind a show-password toggle, is in the screenshot and reaches your model provider, so use test credentials. `StepExecutorContext.pixelsTainted` is removed, and a new step's `visionDegraded` is never `PIXEL_TAINTED`. diff --git a/.changeset/expect-count-missing-frame.md b/.changeset/expect-count-missing-frame.md deleted file mode 100644 index 37d1f84dbe..0000000000 --- a/.changeset/expect-count-missing-frame.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -`toHaveCount` and the list forms of `toHaveText` and `toContainText` read a locator under a missing frame as zero matches, so `toHaveCount(0)` passes after an iframe closes instead of failing with `LOCATOR_NOT_FOUND`. diff --git a/.changeset/explore-runs-init.md b/.changeset/explore-runs-init.md deleted file mode 100644 index cb2d08933a..0000000000 --- a/.changeset/explore-runs-init.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -`e2e explore` in a project with no config runs the `e2e init` wizard first instead of failing with `CONFIG_NOT_FOUND`. The exploration starts right after when nothing is left to do by hand: dependencies installed, a model with its key set or its sign-in done, and a web target, whose URL init names. Otherwise init lists the steps left, ending with the `e2e explore` command as typed. Without a terminal, as in CI or under a coding agent, nothing changes: no files are written and the error stays. diff --git a/.changeset/filter-has-not.md b/.changeset/filter-has-not.md deleted file mode 100644 index 6e60509277..0000000000 --- a/.changeset/filter-has-not.md +++ /dev/null @@ -1,7 +0,0 @@ ---- -"e2e": patch -"@e2e-dev/web": patch -"@e2e-dev/mobile": patch ---- - -`locator.filter()` takes `hasNot`, which drops matches whose subtree resolves a locator, as in Playwright. Third-party engines must handle the new `hasNot` field of a `filter` expression. diff --git a/.changeset/fresh-mobile-inputs.md b/.changeset/fresh-mobile-inputs.md deleted file mode 100644 index dafb68f8de..0000000000 --- a/.changeset/fresh-mobile-inputs.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -'@e2e-dev/mobile': patch ---- - -Re-read mobile fill and clear targets before typing, including cached replay. Missing inputs, changed security state, or incomplete captures retry after a fresh observation. diff --git a/.changeset/github-explore-comment.md b/.changeset/github-explore-comment.md deleted file mode 100644 index c6521d6785..0000000000 --- a/.changeset/github-explore-comment.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"@e2e-dev/github": patch ---- - -An `e2e explore` run posts its own pull request comment. Before, a job that ran `e2e run` and then `e2e explore` replaced the test results comment with the exploration's. Test run comments keep their marker, so the comments already on open pull requests are still updated. An exploration comment an earlier version posted keeps the test run's marker: the next `e2e explore` posts a new comment beside it, and a test run in the same job takes it back. A `--last-failed` rerun no longer folds in an exploration's report that shares its output directory. diff --git a/.changeset/grammar-tools-export.md b/.changeset/grammar-tools-export.md deleted file mode 100644 index 3f00773d5a..0000000000 --- a/.changeset/grammar-tools-export.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -Export the built-in agent's action tools, verdict tool, and rules from `e2e/agent`: `createGrammarTools`, `createVerdictTool`, `BASE_RULES`, and `VERDICT_RULES`, plus `isRuntimeHardStop` and the `ScreenOutput` type, for a custom tool loop or executor that keeps the built-in actions and verdicts. diff --git a/.changeset/hosted-fill-agent-device.md b/.changeset/hosted-fill-agent-device.md deleted file mode 100644 index 28170789c5..0000000000 --- a/.changeset/hosted-fill-agent-device.md +++ /dev/null @@ -1,7 +0,0 @@ ---- -'@e2e-dev/mobile': patch -'@e2e-dev/limrun': patch -'@e2e-dev/testmu': patch ---- - -Update agent-device to 0.21.34. Hosted WebDriver fills replace the confirmed focused field instead of appending, and an unsupported active-element lookup refuses before tapping. The pre-tap focus probe no longer reads the previous field's geometry. diff --git a/.changeset/init-example-angular.md b/.changeset/init-example-angular.md deleted file mode 100644 index e84e7809ff..0000000000 --- a/.changeset/init-example-angular.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -`e2e init --example with-angular` starts a new project from the new Angular example: a standalone component greeting app with three locator tests and two agent tests, run against the Angular dev server. diff --git a/.changeset/init-help-existing-files.md b/.changeset/init-help-existing-files.md deleted file mode 100644 index b4c40e13f7..0000000000 --- a/.changeset/init-help-existing-files.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -`e2e init --help` no longer says it leaves existing files untouched. It says an existing config and test files stay as they are, while `package.json` and `.gitignore` get only what they are missing. diff --git a/.changeset/init-multiple-engines.md b/.changeset/init-multiple-engines.md deleted file mode 100644 index 502939b959..0000000000 --- a/.changeset/init-multiple-engines.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -`e2e init` lets you select both the web and mobile engines. The engine question is now a multiselect, and `--engine` takes a comma-separated list (`--engine web,mobile`). With both, the config gets a target for each, and the examples go in `tests/web.e2e.ts` and `tests/mobile.e2e.ts`, each test with `requires` so the other target skips it. Selecting no engine replaces the **None** option. diff --git a/.changeset/ios-compose-text-fields.md b/.changeset/ios-compose-text-fields.md deleted file mode 100644 index f8ceb6454a..0000000000 --- a/.changeset/ios-compose-text-fields.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -'@e2e-dev/mobile': patch ---- - -Read the placeholder an empty Compose Multiplatform field folds into its iOS label as the field's placeholder, so an empty field no longer reads as holding the example text. `fill` and `clear` type into an unclassified iOS element that holds a value or only text, the way a custom-drawn Compose Multiplatform or Flutter field can read. diff --git a/.changeset/judgment-params.md b/.changeset/judgment-params.md deleted file mode 100644 index e32c84923d..0000000000 --- a/.changeset/judgment-params.md +++ /dev/null @@ -1,11 +0,0 @@ ---- -"e2e": patch ---- - -`agent.assert` and `agent.waitFor` take `params`, like `agent.act`. The judge reads `{name}` with its value, a `unique()` value included. A `Secret` reaches the judge by name only, so check a secret's value on screen with a locator assertion. - -```ts -const name = unique(`acme-${Date.now()}`); -await agent.act('create a company named {name}', { params: { name } }); -await agent.assert('the table shows company {name}', { params: { name } }); -``` diff --git a/.changeset/mcp-run-code.md b/.changeset/mcp-run-code.md deleted file mode 100644 index 5540da7e66..0000000000 --- a/.changeset/mcp-run-code.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -`e2e mcp` sessions have a `run_code` tool. Pass it test statements as a test body writes them, such as `await screen.getByLabel('Email').fill('ada@example.com')`, and the session runs them on its `screen` and `app` with a test's waits and matchers. It stops at the first failure and returns each statement's result, then what changed on screen. The statements that pass, except reads, join the `test_lines` draft. A project tool named `run_code` now fails the config load with `INVALID_CONFIG`. The bundled skill tells the coding agent to run the lines no action printed, such as assertions, with `run_code` before it writes them into the test. diff --git a/.changeset/mcp-run-test.md b/.changeset/mcp-run-test.md deleted file mode 100644 index ccc696c873..0000000000 --- a/.changeset/mcp-run-test.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -`e2e mcp`'s `open_session` takes a `test`, named like an `e2e run` positional (`tests/checkout.e2e.ts:12`), and runs that one test in the session before handing it over: its hooks, `test.extend` fixtures, and body, as `e2e run` runs them. The session then stays on the screen where the test passed or failed, and the opening text leads with the outcome: the error code and message, the failing line with the code around it, the URL, the nodes nearest a locator that matched nothing, and the steps that ran. `afterEach` hooks and fixture teardown run at `close_session`. The test's own steps print no test line; the session's lines start with the agent's first action, and `test_lines` returns them as the lines that continue the test. Agent steps replay and record in the replay cache as on a run's first attempt. A name that selects several tests fails with `TEST_AMBIGUOUS`, and a serial group member, a test that starts from a setup test's session, or a test that runs as several agents fails with `SESSION_TEST_UNSUPPORTED`. The bundled skill tells the coding agent to fix a failing test this way instead of clicking through the flow again. diff --git a/.changeset/mobile-keyboard-transitions.md b/.changeset/mobile-keyboard-transitions.md deleted file mode 100644 index e4a89e21bc..0000000000 --- a/.changeset/mobile-keyboard-transitions.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"@e2e-dev/mobile": patch ---- - -A test step on a control the keyboard moves now waits for it to land instead of tapping where it was. Moving-target refusals are retried when agent-device sent no input. diff --git a/.changeset/mobile-lease-helper.md b/.changeset/mobile-lease-helper.md deleted file mode 100644 index 74f7422660..0000000000 --- a/.changeset/mobile-lease-helper.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -'@e2e-dev/mobile': patch ---- - -Add `@e2e-dev/mobile/lease` for device-provider packages: `agentDeviceLeases()` leases devices from a cloud agent-device runs sessions on, one isolated daemon per worker slot, and fails a release that agent-device can't confirm ended the session. diff --git a/.changeset/mobile-ref-generation.md b/.changeset/mobile-ref-generation.md deleted file mode 100644 index a361bee6ed..0000000000 --- a/.changeset/mobile-ref-generation.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"@e2e-dev/mobile": patch ---- - -A mobile action on a node that a newer snapshot replaced finds the node again instead of acting on a different element. diff --git a/.changeset/mobile-stale-transition-ref.md b/.changeset/mobile-stale-transition-ref.md deleted file mode 100644 index c9f9ac3a03..0000000000 --- a/.changeset/mobile-stale-transition-ref.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"@e2e-dev/mobile": patch ---- - -Re-resolve a locator when its target changes during the transition wait, instead of tapping a stale native reference that may now identify another control. diff --git a/.changeset/multi-file-downloads.md b/.changeset/multi-file-downloads.md deleted file mode 100644 index 725c7b21f8..0000000000 --- a/.changeset/multi-file-downloads.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"@e2e-dev/web": patch ---- - -Add `browser.waitForDownloads(trigger, { count })` to collect several files from one trigger. `waitForDownload` now also bounds completion and saving by its timeout. diff --git a/.changeset/navigate-app-relative.md b/.changeset/navigate-app-relative.md deleted file mode 100644 index 6fcc70eae8..0000000000 --- a/.changeset/navigate-app-relative.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -Agent `navigate` to a full URL on the app now records it without the origin, so cached steps replay when the app runs on another port or host with the same `app.identity`. Re-record older entries to pick this up. diff --git a/.changeset/negative-text-locator-filter.md b/.changeset/negative-text-locator-filter.md deleted file mode 100644 index ad8541afbb..0000000000 --- a/.changeset/negative-text-locator-filter.md +++ /dev/null @@ -1,7 +0,0 @@ ---- -"e2e": patch -"@e2e-dev/web": patch -"@e2e-dev/mobile": patch ---- - -`locator.filter()` takes `hasNotText`, which drops matches whose text contains a string or matches a RegExp, as in Playwright. Third-party engines must handle the new `hasNotText` field of a `filter` expression. diff --git a/.changeset/no-output-error-invalid.md b/.changeset/no-output-error-invalid.md deleted file mode 100644 index 62b5f315f9..0000000000 --- a/.changeset/no-output-error-invalid.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -A structured model call that returns no parseable output (for example an empty `length` finish) now fails as `MODEL_OUTPUT_INVALID` instead of `MODEL_PROVIDER_FAILED`. diff --git a/.changeset/oauth-windows-browser-open.md b/.changeset/oauth-windows-browser-open.md deleted file mode 100644 index af5d1211a8..0000000000 --- a/.changeset/oauth-windows-browser-open.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -`e2e login` on Windows no longer opens the sign-in URL through a command interpreter. The URL comes from the provider's own response in the device flow, and `cmd /c start` read a `|`, `&`, or `^` in it as command syntax, so a URL that carried one ran a command of its own. The URL is now handed to the default handler for its scheme as a single argument, and only an `http:` or `https:` URL is opened; any other scheme is printed for you to open yourself. diff --git a/.changeset/open-app-target.md b/.changeset/open-app-target.md deleted file mode 100644 index ef8bdd5463..0000000000 --- a/.changeset/open-app-target.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -'@e2e-dev/mobile': patch ---- - -The `open_app` agent tool returns to the target's app when called without an app, so the agent can recover after a system sheet or another app takes the foreground. Device errors no longer end with agent-device's CLI advice, such as "Run apps to discover...". diff --git a/.changeset/point-tool-rules.md b/.changeset/point-tool-rules.md deleted file mode 100644 index f605f3d125..0000000000 --- a/.changeset/point-tool-rules.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -The built-in agent's instructions now say that `tap_at`, `hover_at`, and `drag_at` act on the bare points when the engine supports it, and fall back to the listed controls by id only when it does not. They previously said every point tool acts on a listed control by its id. diff --git a/.changeset/press-enter-focus.md b/.changeset/press-enter-focus.md deleted file mode 100644 index f41b00e0be..0000000000 --- a/.changeset/press-enter-focus.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -'@e2e-dev/mobile': patch ---- - -`press('Enter')` on a field taps it first when it is not focused, as a character key already did. Before, Enter went to whatever held focus, and on iOS failed when the keyboard was down. diff --git a/.changeset/replay-end-path-wait.md b/.changeset/replay-end-path-wait.md deleted file mode 100644 index ff8c9912c4..0000000000 --- a/.changeset/replay-end-path-wait.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -A replayed step whose last action starts a navigation now waits up to 15 seconds for the recorded destination, instead of handing off to the agent after about 5 seconds. diff --git a/.changeset/retain-serial-verdict-duration.md b/.changeset/retain-serial-verdict-duration.md deleted file mode 100644 index 2f2af9b071..0000000000 --- a/.changeset/retain-serial-verdict-duration.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -Report a test's duration from the attempt that decided it when a retry is interrupted or cannot start. diff --git a/.changeset/route-fetch-upstream.md b/.changeset/route-fetch-upstream.md deleted file mode 100644 index 48714aa11b..0000000000 --- a/.changeset/route-fetch-upstream.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"@e2e-dev/web": patch ---- - -Add `route.fetch()` to read the upstream response inside a route handler before deciding it. diff --git a/.changeset/runner-explore.md b/.changeset/runner-explore.md deleted file mode 100644 index 22d59b5f26..0000000000 --- a/.changeset/runner-explore.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -`e2e/runner` exports `explore()`, the counterpart of `e2e explore` that `run()` is to `e2e run`. It takes the project (`cwd`, and `config` as a file path or an `E2EConfig` value), the goal and its budgets (`goal`, `target`, `agent`, `session`, `maxSteps`, `timeoutMs`), `interruptSignal`, and `onEvent`, and resolves with the run's result plus the exploration record (`explore.ended`, `explore.steps`, `explore.findings`) without exiting the caller. Cancellation awaits teardown. The notices `e2e explore` prints before the run starts, such as the target it chose, arrive through `onEvent` as `run` notices. It shares `run()`'s reservation, so overlapping calls reject with `RUNNER_BUSY`, and each exploration that returns a report sends one `e2e_run_completed` event with `command: "runner.explore"`. diff --git a/.changeset/screenshot-caret-stylesheet.md b/.changeset/screenshot-caret-stylesheet.md deleted file mode 100644 index 16367a2397..0000000000 --- a/.changeset/screenshot-caret-stylesheet.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"@e2e-dev/web": patch ---- - -Screenshots no longer change the inline style of editable elements to hide the text caret, so a server-rendered app hydrating while a capture runs no longer reports a React attribute mismatch on its inputs and textareas. A capture runs with the caret hidden in the frames it can reach, and the stylesheet is taken back from every frame afterwards. diff --git a/.changeset/selective-cookie-clearing.md b/.changeset/selective-cookie-clearing.md deleted file mode 100644 index 46f22e1109..0000000000 --- a/.changeset/selective-cookie-clearing.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"@e2e-dev/web": patch ---- - -Add selective cookie clearing with validated name, domain and path filters, preserving other browser storage. diff --git a/.changeset/skill-write-as-you-walk.md b/.changeset/skill-write-as-you-walk.md deleted file mode 100644 index a1d1f27a3f..0000000000 --- a/.changeset/skill-write-as-you-walk.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -The bundled agent skill now has the coding agent write a test while it walks the flow over `e2e mcp`, instead of exploring first and writing the test from memory. It creates the test file before it opens the session and adds each step as it takes it. A step is an `agent.act` goal, written in the words on screen before the agent walks its path, or, for an exact value, the `Test line:` the action printed. When the outcome shows, the agent locates it and adds the `expect`. The `e2e mcp` server's instructions, `e2e mcp --help`, and the MCP server and coding agents docs say the same. diff --git a/.changeset/testmu-strict-release.md b/.changeset/testmu-strict-release.md deleted file mode 100644 index c01a28ae82..0000000000 --- a/.changeset/testmu-strict-release.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -'@e2e-dev/testmu': patch ---- - -A device release now fails the run when agent-device can't confirm it ended the TestMu AI session: a forced daemon stop, a lost shutdown report, or a daemon that exited during the run. Needs `@e2e-dev/mobile` 0.12.5 or later. diff --git a/.changeset/tricky-weeks-grin.md b/.changeset/tricky-weeks-grin.md deleted file mode 100644 index 2aab08c63e..0000000000 --- a/.changeset/tricky-weeks-grin.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -Report rejected recording writes as not-written. diff --git a/.changeset/unique-multi-line-replay.md b/.changeset/unique-multi-line-replay.md deleted file mode 100644 index 270992b10f..0000000000 --- a/.changeset/unique-multi-line-replay.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -A multi-line `unique()` value now replays from the cache. The screen shows it on one line, so the end check now looks for it on one line too. Before, the replay missed: `REPLAY_STALE (end-mismatch)` with `--strict-cache`, a hand-off to the agent without it. An older runner reads an entry recorded with this fix as `invalid-entry`. diff --git a/.changeset/upgrade-agent-skill.md b/.changeset/upgrade-agent-skill.md deleted file mode 100644 index 0d8c508a3d..0000000000 --- a/.changeset/upgrade-agent-skill.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"e2e": patch ---- - -`e2e upgrade` updates the project's copies of the agent skill in `.agents/skills/e2e` and `.claude/skills/e2e` to the newly installed version, instead of telling you to run `e2e init`. It also updates a stale copy when the packages are already up to date, and `--dry-run` names the copies it would update. diff --git a/.changeset/upload-filechooser-button.md b/.changeset/upload-filechooser-button.md deleted file mode 100644 index cbb34d48a9..0000000000 --- a/.changeset/upload-filechooser-button.md +++ /dev/null @@ -1,6 +0,0 @@ ---- -"e2e": patch -"@e2e-dev/web": patch ---- - -`upload` and `setInputFiles` accept a button that opens a file chooser, such as one over a hidden file input. The agent's `upload` tool says so, so the model can pick the button it sees. diff --git a/.changeset/web-app-permissions.md b/.changeset/web-app-permissions.md deleted file mode 100644 index 15c82e135c..0000000000 --- a/.changeset/web-app-permissions.md +++ /dev/null @@ -1,6 +0,0 @@ ---- -"@e2e-dev/web": patch -"e2e": patch ---- - -A `web()` target takes `app.permissions` with `'grant'` states, granted to every page of the attempt: `permissions: { 'clipboard-read': 'grant', 'clipboard-write': 'grant' }` lets the app use the clipboard in headless Chromium. Names are checked against the browser's at config load; `'deny'` and `'reset'` are refused on web. diff --git a/.changeset/web-replaced-weakmap.md b/.changeset/web-replaced-weakmap.md deleted file mode 100644 index 99c5ff3f43..0000000000 --- a/.changeset/web-replaced-weakmap.md +++ /dev/null @@ -1,5 +0,0 @@ ---- -"@e2e-dev/web": patch ---- - -Screenshots no longer fail with `ENGINE_FAILURE` ("closed-shadow root tracking is unavailable") on a page that replaces the global `WeakMap`, as some polyfills do. On such pages, locators find elements inside closed shadow roots again, and CDP reconnects no longer fail. diff --git a/packages/acp/CHANGELOG.md b/packages/acp/CHANGELOG.md new file mode 100644 index 0000000000..42dcd6110c --- /dev/null +++ b/packages/acp/CHANGELOG.md @@ -0,0 +1,11 @@ +# @e2e-dev/acp + +## 0.1.0 + +### Minor Changes + +- [#960](https://github.com/tester-army/e2e/pull/960) [`ffed34e`](https://github.com/tester-army/e2e/commit/ffed34e94b9b75867d4526f2b1b92e7a66764c3f) Thanks [@enekesabel](https://github.com/enekesabel)! - **Release:** New package: run `agent.act` and `agent.assert` on your own coding agent over the Agent Client Protocol, signed in with the agent's own login. `claudeCode()` and `codex()` start those agents with e2e's step tools as their only tools; `acp({ command, args })` starts any other ACP agent. Every action goes through the runner, so steps replay from the cache without the agent, and a tool of the agent's own that runs without asking fails the step with `POLICY_DENIED`. + +### Patch Changes + +- [#1457](https://github.com/tester-army/e2e/pull/1457) [`e3195e3`](https://github.com/tester-army/e2e/commit/e3195e3f79cf7dd845a8ea05bbafa9b30d883c7a) Thanks [@samanyudas](https://github.com/samanyudas)! - A step that starts right after a step that timed out now gets a fresh agent when the old agent runs a tool of its own as it stops. Before, the step failed with `MODEL_PROVIDER_FAILED` ("ACP connection closed"), because it reused the session that tool had closed. diff --git a/packages/acp/package.json b/packages/acp/package.json index d2061ccc3e..0c4e5108a5 100644 --- a/packages/acp/package.json +++ b/packages/acp/package.json @@ -1,6 +1,6 @@ { "name": "@e2e-dev/acp", - "version": "0.0.0", + "version": "0.1.0", "description": "Run e2e agent steps on a coding agent you already use (Claude Code, Codex, Cursor, ...) over the Agent Client Protocol", "keywords": [ "e2e", diff --git a/packages/browserstack/CHANGELOG.md b/packages/browserstack/CHANGELOG.md new file mode 100644 index 0000000000..5d0ea49adf --- /dev/null +++ b/packages/browserstack/CHANGELOG.md @@ -0,0 +1,7 @@ +# @e2e-dev/browserstack + +## 0.1.0 + +### Minor Changes + +- [#1307](https://github.com/tester-army/e2e/pull/1307) [`64432f1`](https://github.com/tester-army/e2e/commit/64432f10891d9772494137b0ba328d8ef440deae) Thanks [@malopezr7](https://github.com/malopezr7)! - **Release:** New package: `@e2e-dev/browserstack`. `browserstack()` leases BrowserStack App Automate real Android and iOS devices for the mobile engine, one isolated agent-device daemon per worker slot. An `https` `app` is uploaded once per run, an attempt's video links BrowserStack's recording of the session, and a release fails the run when agent-device can't confirm it ended the session. diff --git a/packages/browserstack/package.json b/packages/browserstack/package.json index bf68124f03..bd859c3ff1 100644 --- a/packages/browserstack/package.json +++ b/packages/browserstack/package.json @@ -1,6 +1,6 @@ { "name": "@e2e-dev/browserstack", - "version": "0.0.0", + "version": "0.1.0", "description": "BrowserStack App Automate real Android and iOS devices for e2e mobile tests", "keywords": [ "e2e", diff --git a/packages/decision/CHANGELOG.md b/packages/decision/CHANGELOG.md index 7598d6bb00..e70861351e 100644 --- a/packages/decision/CHANGELOG.md +++ b/packages/decision/CHANGELOG.md @@ -1,5 +1,13 @@ # @e2e-dev/decision +## 0.3.1 + +### Patch Changes + +- [#1382](https://github.com/tester-army/e2e/pull/1382) [`3a8d6f1`](https://github.com/tester-army/e2e/commit/3a8d6f1b6cdaa4c7975c55bf25f67d2e76143b21) Thanks [@okwasniewski](https://github.com/okwasniewski)! - Bump agent-device to 0.21.31, zod to 4.6.7, @modelcontextprotocol/server to 2.4.0, and oxc to 0.155.0. + +- [#1437](https://github.com/tester-army/e2e/pull/1437) [`94e6e76`](https://github.com/tester-army/e2e/commit/94e6e7654856b6e3946424534386cb6323a47eb6) Thanks [@okwasniewski](https://github.com/okwasniewski)! - **Breaking:** a secret fill no longer turns off screenshots for the model. `agent.assert(..., { vision: 'only' })`, `vision: true`, the `screenshot` and point tools, `e2e mcp`, and `e2e explore` findings keep working after a sign-in, in serial groups, and in tests that restore a saved session. Screenshots still mask secure fields, and secret values are still redacted from all text. A generic secret filled into a plain input, or a value the app shows outside a secure field, such as behind a show-password toggle, is in the screenshot and reaches your model provider, so use test credentials. `StepExecutorContext.pixelsTainted` is removed, and a new step's `visionDegraded` is never `PIXEL_TAINTED`. + ## 0.3.0 ### Minor Changes diff --git a/packages/decision/package.json b/packages/decision/package.json index 755091a3c4..040cabe7f9 100644 --- a/packages/decision/package.json +++ b/packages/decision/package.json @@ -1,6 +1,6 @@ { "name": "@e2e-dev/decision", - "version": "0.3.0", + "version": "0.3.1", "description": "Decision-model executors and test filters for e2e", "license": "Apache-2.0", "repository": { diff --git a/packages/e2e/CHANGELOG.md b/packages/e2e/CHANGELOG.md index 163b1bd417..ed9a466f45 100644 --- a/packages/e2e/CHANGELOG.md +++ b/packages/e2e/CHANGELOG.md @@ -1,5 +1,96 @@ # e2e +## 0.22.0 + +### Minor Changes + +- [#1437](https://github.com/tester-army/e2e/pull/1437) [`94e6e76`](https://github.com/tester-army/e2e/commit/94e6e7654856b6e3946424534386cb6323a47eb6) Thanks [@okwasniewski](https://github.com/okwasniewski)! - **Breaking:** a secret fill no longer turns off screenshots for the model. `agent.assert(..., { vision: 'only' })`, `vision: true`, the `screenshot` and point tools, `e2e mcp`, and `e2e explore` findings keep working after a sign-in, in serial groups, and in tests that restore a saved session. Screenshots still mask secure fields, and secret values are still redacted from all text. A generic secret filled into a plain input, or a value the app shows outside a secure field, such as behind a show-password toggle, is in the screenshot and reaches your model provider, so use test credentials. `StepExecutorContext.pixelsTainted` is removed, and a new step's `visionDegraded` is never `PIXEL_TAINTED`. + +### Patch Changes + +- [#1282](https://github.com/tester-army/e2e/pull/1282) [`7518ae5`](https://github.com/tester-army/e2e/commit/7518ae5b2d1879169185455777022e8ac28f77f8) Thanks [@MarcosCamara01](https://github.com/MarcosCamara01)! - Add `drag_at` to the agent's pixel tools, in `agent.act` and `e2e mcp`, and `dragAt(from, to)` to `ExecutorActions`: a drag from one screenshot point to another, for a shape on a canvas or a pin on a map. Like `tap_at`, the bare points go to the engine's `dragTo` pointer kind, where the model aimed; an engine without it drags by id when the points land on two different listed controls, like `drag`. The replay cache records both points and replays each the way it replays a `tap_at` point. + +- [#1447](https://github.com/tester-army/e2e/pull/1447) [`82d5095`](https://github.com/tester-army/e2e/commit/82d5095a1f29ee2660b41db5235703d81f329436) Thanks [@21prnv](https://github.com/21prnv)! - A replay no longer fails its end check on an alert that shows an ISO 8601 timestamp, such as `Saved at 2026-10-24T15:41:27.955Z`, when the time differs from the recording. + +- [#1381](https://github.com/tester-army/e2e/pull/1381) [`d2ffb31`](https://github.com/tester-army/e2e/commit/d2ffb319cfd81e267a5661c9d9229f874c38bb0d) Thanks [@okwasniewski](https://github.com/okwasniewski)! - `app.startRecording()` and `app.stopRecording()` record one span of the screen, such as a single transition, on the web and on devices. `stopRecording` returns the videos: a file `path`, or a `url` when a hosted browser or device keeps the recording. A start fails while the test's `video` records, so set `{ video: 'off' }` on the test. + + ```ts + test('the dialog opens', { video: 'off' }, async ({ app, screen }) => { + await app.open(); + await app.startRecording({ name: 'dialog' }); + await screen.getByRole('button', 'Open').tap(); + const [video] = await app.stopRecording(); + }); + ``` + + An engine's `startVideo` and `stopVideo` may now be called more than once per attempt, never overlapping. + +- [#1387](https://github.com/tester-army/e2e/pull/1387) [`4461e9d`](https://github.com/tester-army/e2e/commit/4461e9d10458919205198c31307d9b06a59f1722) Thanks [@bheemreddy-samsara](https://github.com/bheemreddy-samsara)! - Add `onBeforeAction` to `agent.act`, a hook that runs before each action, so a test can capture the screen an action started on. + +- [#1379](https://github.com/tester-army/e2e/pull/1379) [`9e53c40`](https://github.com/tester-army/e2e/commit/9e53c40a94fed6585d4e57cf14b8e91c405cb5d8) Thanks [@okwasniewski](https://github.com/okwasniewski)! - A mobile target without `device` skips a booted simulator or emulator that another agent-device session holds, such as one another tool is driving, and drives the free ones. When every booted device is held, the run fails at `prepare` naming the holding sessions instead of failing on the first one. + +- [#1382](https://github.com/tester-army/e2e/pull/1382) [`3a8d6f1`](https://github.com/tester-army/e2e/commit/3a8d6f1b6cdaa4c7975c55bf25f67d2e76143b21) Thanks [@okwasniewski](https://github.com/okwasniewski)! - Bump agent-device to 0.21.31, zod to 4.6.7, @modelcontextprotocol/server to 2.4.0, and oxc to 0.155.0. + +- [#1233](https://github.com/tester-army/e2e/pull/1233) [`34287cf`](https://github.com/tester-army/e2e/commit/34287cfe78224d1d6b7050ee218a25f9606e4236) Thanks [@Varshith-Kali](https://github.com/Varshith-Kali)! - A run now exits with the cleanup error's code, usually 3, when a test's cleanup failed. The list reporter shows the cleanup failure in the summary and under the test for passed and flaky tests, so a CI job that goes red names the cleanup failure instead of reporting a green summary. + +- [#1351](https://github.com/tester-army/e2e/pull/1351) [`bfae325`](https://github.com/tester-army/e2e/commit/bfae3257074b19a4deaa04b13c17a65d5cd95c5a) Thanks [@samanyudas](https://github.com/samanyudas)! - `e2e run` and `e2e explore` refuse an empty `--reporter` (`--reporter ''`, `--reporter ','`, or `--reporter "$REPORTERS"` with the variable unset) as a usage error, exit 2, instead of running with no reporters and printing nothing. + +- [#1287](https://github.com/tester-army/e2e/pull/1287) [`5aec484`](https://github.com/tester-army/e2e/commit/5aec48417e495c526ae5f81dde7caf452999ca6c) Thanks [@TheTuso](https://github.com/TheTuso)! - An `e2e` command whose config or test file leaves a timer or socket open exits two seconds after it finishes, instead of staying open. + +- [#1354](https://github.com/tester-army/e2e/pull/1354) [`6901a01`](https://github.com/tester-army/e2e/commit/6901a01e0e65c4d1d24e628dac9a837bce594607) Thanks [@samanyudas](https://github.com/samanyudas)! - `toHaveCount` and the list forms of `toHaveText` and `toContainText` read a locator under a missing frame as zero matches, so `toHaveCount(0)` passes after an iframe closes instead of failing with `LOCATOR_NOT_FOUND`. + +- [#1401](https://github.com/tester-army/e2e/pull/1401) [`ce249b6`](https://github.com/tester-army/e2e/commit/ce249b62b0d6a510a198506c2f093eeac2c439d0) Thanks [@kacperkapusciak](https://github.com/kacperkapusciak)! - `e2e explore` in a project with no config runs the `e2e init` wizard first instead of failing with `CONFIG_NOT_FOUND`. The exploration starts right after when nothing is left to do by hand: dependencies installed, a model with its key set or its sign-in done, and a web target, whose URL init names. Otherwise init lists the steps left, ending with the `e2e explore` command as typed. Without a terminal, as in CI or under a coding agent, nothing changes: no files are written and the error stays. + +- [#1334](https://github.com/tester-army/e2e/pull/1334) [`6c8e71f`](https://github.com/tester-army/e2e/commit/6c8e71f1ddab036d5cd82ca671163de6e9e82ca0) Thanks [@21prnv](https://github.com/21prnv)! - `locator.filter()` takes `hasNot`, which drops matches whose subtree resolves a locator, as in Playwright. Third-party engines must handle the new `hasNot` field of a `filter` expression. + +- [#960](https://github.com/tester-army/e2e/pull/960) [`ffed34e`](https://github.com/tester-army/e2e/commit/ffed34e94b9b75867d4526f2b1b92e7a66764c3f) Thanks [@enekesabel](https://github.com/enekesabel)! - Export the built-in agent's action tools, verdict tool, and rules from `e2e/agent`: `createGrammarTools`, `createVerdictTool`, `BASE_RULES`, and `VERDICT_RULES`, plus `isRuntimeHardStop` and the `ScreenOutput` type, for a custom tool loop or executor that keeps the built-in actions and verdicts. + +- [#1389](https://github.com/tester-army/e2e/pull/1389) [`6b04d18`](https://github.com/tester-army/e2e/commit/6b04d18a6271ec37c77f99c93293ab8816c08f59) Thanks [@kacperkapusciak](https://github.com/kacperkapusciak)! - `e2e init --example with-angular` starts a new project from the new Angular example: a standalone component greeting app with three locator tests and two agent tests, run against the Angular dev server. + +- [#1398](https://github.com/tester-army/e2e/pull/1398) [`e087ed2`](https://github.com/tester-army/e2e/commit/e087ed2c47f12e0994d279a2c84170cbdd86e250) Thanks [@samanyudas](https://github.com/samanyudas)! - `e2e init --help` no longer says it leaves existing files untouched. It says an existing config and test files stay as they are, while `package.json` and `.gitignore` get only what they are missing. + +- [#1338](https://github.com/tester-army/e2e/pull/1338) [`0d3e403`](https://github.com/tester-army/e2e/commit/0d3e403c2e7caccd6ae4ebb474025c1a227920fd) Thanks [@kacperkapusciak](https://github.com/kacperkapusciak)! - `e2e init` lets you select both the web and mobile engines. The engine question is now a multiselect, and `--engine` takes a comma-separated list (`--engine web,mobile`). With both, the config gets a target for each, and the examples go in `tests/web.e2e.ts` and `tests/mobile.e2e.ts`, each test with `requires` so the other target skips it. Selecting no engine replaces the **None** option. + +- [#1463](https://github.com/tester-army/e2e/pull/1463) [`7372b0a`](https://github.com/tester-army/e2e/commit/7372b0a83987242eae08a02512d833d316f6362e) Thanks [@okwasniewski](https://github.com/okwasniewski)! - `agent.assert` and `agent.waitFor` take `params`, like `agent.act`. The judge reads `{name}` with its value, a `unique()` value included. A `Secret` reaches the judge by name only, so check a secret's value on screen with a locator assertion. + + ```ts + const name = unique(`acme-${Date.now()}`); + await agent.act('create a company named {name}', { params: { name } }); + await agent.assert('the table shows company {name}', { params: { name } }); + ``` + +- [#1316](https://github.com/tester-army/e2e/pull/1316) [`6b41857`](https://github.com/tester-army/e2e/commit/6b418575302608f3adff4224de65b5adc580f8bd) Thanks [@kacperkapusciak](https://github.com/kacperkapusciak)! - `e2e mcp` sessions have a `run_code` tool. Pass it test statements as a test body writes them, such as `await screen.getByLabel('Email').fill('ada@example.com')`, and the session runs them on its `screen` and `app` with a test's waits and matchers. It stops at the first failure and returns each statement's result, then what changed on screen. The statements that pass, except reads, join the `test_lines` draft. A project tool named `run_code` now fails the config load with `INVALID_CONFIG`. The bundled skill tells the coding agent to run the lines no action printed, such as assertions, with `run_code` before it writes them into the test. + +- [#1315](https://github.com/tester-army/e2e/pull/1315) [`3d2c362`](https://github.com/tester-army/e2e/commit/3d2c3627ecdcb21bc34c05ca03bc960d7aa3601e) Thanks [@kacperkapusciak](https://github.com/kacperkapusciak)! - `e2e mcp`'s `open_session` takes a `test`, named like an `e2e run` positional (`tests/checkout.e2e.ts:12`), and runs that one test in the session before handing it over: its hooks, `test.extend` fixtures, and body, as `e2e run` runs them. The session then stays on the screen where the test passed or failed, and the opening text leads with the outcome: the error code and message, the failing line with the code around it, the URL, the nodes nearest a locator that matched nothing, and the steps that ran. `afterEach` hooks and fixture teardown run at `close_session`. The test's own steps print no test line; the session's lines start with the agent's first action, and `test_lines` returns them as the lines that continue the test. Agent steps replay and record in the replay cache as on a run's first attempt. A name that selects several tests fails with `TEST_AMBIGUOUS`, and a serial group member, a test that starts from a setup test's session, or a test that runs as several agents fails with `SESSION_TEST_UNSUPPORTED`. The bundled skill tells the coding agent to fix a failing test this way instead of clicking through the flow again. + +- [#1191](https://github.com/tester-army/e2e/pull/1191) [`9bc7c32`](https://github.com/tester-army/e2e/commit/9bc7c329ddba5d0e52fe1885caf5b401aaf4a598) Thanks [@TheTuso](https://github.com/TheTuso)! - Agent `navigate` to a full URL on the app now records it without the origin, so cached steps replay when the app runs on another port or host with the same `app.identity`. Re-record older entries to pick this up. + +- [#1037](https://github.com/tester-army/e2e/pull/1037) [`a481fd3`](https://github.com/tester-army/e2e/commit/a481fd39c69e3e0f88a94c1fe3f4acada164cbf6) Thanks [@rudycelekli](https://github.com/rudycelekli)! - `locator.filter()` takes `hasNotText`, which drops matches whose text contains a string or matches a RegExp, as in Playwright. Third-party engines must handle the new `hasNotText` field of a `filter` expression. + +- [#1215](https://github.com/tester-army/e2e/pull/1215) [`d8c597b`](https://github.com/tester-army/e2e/commit/d8c597b64e52387cf50db6f60c4d39607db17f17) Thanks [@Varshith-Kali](https://github.com/Varshith-Kali)! - A structured model call that returns no parseable output (for example an empty `length` finish) now fails as `MODEL_OUTPUT_INVALID` instead of `MODEL_PROVIDER_FAILED`. + +- [#1328](https://github.com/tester-army/e2e/pull/1328) [`1c4e906`](https://github.com/tester-army/e2e/commit/1c4e906766379ae8749c3b8ae7918c407e145c09) Thanks [@DeryFerd](https://github.com/DeryFerd)! - `e2e login` on Windows no longer opens the sign-in URL through a command interpreter. The URL comes from the provider's own response in the device flow, and `cmd /c start` read a `|`, `&`, or `^` in it as command syntax, so a URL that carried one ran a command of its own. The URL is now handed to the default handler for its scheme as a single argument, and only an `http:` or `https:` URL is opened; any other scheme is printed for you to open yourself. + +- [#1399](https://github.com/tester-army/e2e/pull/1399) [`92409e5`](https://github.com/tester-army/e2e/commit/92409e5b43a19b4f0d7d2c99a2da6a3a76c5c880) Thanks [@samanyudas](https://github.com/samanyudas)! - The built-in agent's instructions now say that `tap_at`, `hover_at`, and `drag_at` act on the bare points when the engine supports it, and fall back to the listed controls by id only when it does not. They previously said every point tool acts on a listed control by its id. + +- [#1357](https://github.com/tester-army/e2e/pull/1357) [`9837386`](https://github.com/tester-army/e2e/commit/9837386ebdf46bea3acda15ee2ff163ed1b695d3) Thanks [@amoscicki](https://github.com/amoscicki)! - A replayed step whose last action starts a navigation now waits up to 15 seconds for the recorded destination, instead of handing off to the agent after about 5 seconds. + +- [#1056](https://github.com/tester-army/e2e/pull/1056) [`dc8f56d`](https://github.com/tester-army/e2e/commit/dc8f56d20533638098045cbb94ca069ccfef64d3) Thanks [@rudycelekli](https://github.com/rudycelekli)! - Report a test's duration from the attempt that decided it when a retry is interrupted or cannot start. + +- [#1256](https://github.com/tester-army/e2e/pull/1256) [`2bc00e9`](https://github.com/tester-army/e2e/commit/2bc00e91cea64eb44bf4897f25e2495a5beb53b3) Thanks [@arun2728](https://github.com/arun2728)! - `e2e/runner` exports `explore()`, the counterpart of `e2e explore` that `run()` is to `e2e run`. It takes the project (`cwd`, and `config` as a file path or an `E2EConfig` value), the goal and its budgets (`goal`, `target`, `agent`, `session`, `maxSteps`, `timeoutMs`), `interruptSignal`, and `onEvent`, and resolves with the run's result plus the exploration record (`explore.ended`, `explore.steps`, `explore.findings`) without exiting the caller. Cancellation awaits teardown. The notices `e2e explore` prints before the run starts, such as the target it chose, arrive through `onEvent` as `run` notices. It shares `run()`'s reservation, so overlapping calls reject with `RUNNER_BUSY`, and each exploration that returns a report sends one `e2e_run_completed` event with `command: "runner.explore"`. + +- [#1312](https://github.com/tester-army/e2e/pull/1312) [`9d884f5`](https://github.com/tester-army/e2e/commit/9d884f5b981a8959a5e030897709cb5be2a6f4ac) Thanks [@kacperkapusciak](https://github.com/kacperkapusciak)! - The bundled agent skill now has the coding agent write a test while it walks the flow over `e2e mcp`, instead of exploring first and writing the test from memory. It creates the test file before it opens the session and adds each step as it takes it. A step is an `agent.act` goal, written in the words on screen before the agent walks its path, or, for an exact value, the `Test line:` the action printed. When the outcome shows, the agent locates it and adds the `expect`. The `e2e mcp` server's instructions, `e2e mcp --help`, and the MCP server and coding agents docs say the same. + +- [#1372](https://github.com/tester-army/e2e/pull/1372) [`0cc79f1`](https://github.com/tester-army/e2e/commit/0cc79f1a728a1698de1e4b9da76c8384513eb241) Thanks [@BiBoyang](https://github.com/BiBoyang)! - Report rejected recording writes as not-written. + +- [#1452](https://github.com/tester-army/e2e/pull/1452) [`5b6fd01`](https://github.com/tester-army/e2e/commit/5b6fd01042c96e5d79e662325cfaae2157d6eac9) Thanks [@samanyudas](https://github.com/samanyudas)! - A multi-line `unique()` value now replays from the cache. The screen shows it on one line, so the end check now looks for it on one line too. Before, the replay missed: `REPLAY_STALE (end-mismatch)` with `--strict-cache`, a hand-off to the agent without it. An older runner reads an entry recorded with this fix as `invalid-entry`. + +- [#1394](https://github.com/tester-army/e2e/pull/1394) [`e9e8c50`](https://github.com/tester-army/e2e/commit/e9e8c5005aaf94f22551f4cc7c5a85f6d4a6d17f) Thanks [@kacperkapusciak](https://github.com/kacperkapusciak)! - `e2e upgrade` updates the project's copies of the agent skill in `.agents/skills/e2e` and `.claude/skills/e2e` to the newly installed version, instead of telling you to run `e2e init`. It also updates a stale copy when the packages are already up to date, and `--dry-run` names the copies it would update. + +- [#958](https://github.com/tester-army/e2e/pull/958) [`85e92f9`](https://github.com/tester-army/e2e/commit/85e92f91f1c5415850ca4456ee5f4e2a81a6b648) Thanks [@kuishou68](https://github.com/kuishou68)! - `upload` and `setInputFiles` accept a button that opens a file chooser, such as one over a hidden file input. The agent's `upload` tool says so, so the model can pick the button it sees. + +- [#1429](https://github.com/tester-army/e2e/pull/1429) [`3a841bb`](https://github.com/tester-army/e2e/commit/3a841bb99e5ecc050154c9e6d7f696a6226c3980) Thanks [@KrzysztofMoch](https://github.com/KrzysztofMoch)! - A `web()` target takes `app.permissions` with `'grant'` states, granted to every page of the attempt: `permissions: { 'clipboard-read': 'grant', 'clipboard-write': 'grant' }` lets the app use the clipboard in headless Chromium. Names are checked against the browser's at config load; `'deny'` and `'reset'` are refused on web. + ## 0.21.3 ### Patch Changes diff --git a/packages/e2e/package.json b/packages/e2e/package.json index d3d47bd66c..78cafdd232 100644 --- a/packages/e2e/package.json +++ b/packages/e2e/package.json @@ -1,6 +1,6 @@ { "name": "e2e", - "version": "0.21.3", + "version": "0.22.0", "description": "SDK, runner, and CLI for agentic end-to-end testing", "mcpName": "io.github.tester-army/e2e", "keywords": [ diff --git a/packages/e2e/server.json b/packages/e2e/server.json index 2672409f00..dfff437a22 100644 --- a/packages/e2e/server.json +++ b/packages/e2e/server.json @@ -9,13 +9,13 @@ "source": "github", "subfolder": "packages/e2e" }, - "version": "0.21.3", + "version": "0.22.0", "packages": [ { "registryType": "npm", "registryBaseUrl": "https://registry.npmjs.org", "identifier": "e2e", - "version": "0.21.3", + "version": "0.22.0", "runtimeHint": "npx", "transport": { "type": "stdio" diff --git a/packages/github/CHANGELOG.md b/packages/github/CHANGELOG.md index d3ce49e2cd..d432917b0a 100644 --- a/packages/github/CHANGELOG.md +++ b/packages/github/CHANGELOG.md @@ -1,5 +1,11 @@ # @e2e-dev/github +## 0.4.4 + +### Patch Changes + +- [#1438](https://github.com/tester-army/e2e/pull/1438) [`d181e26`](https://github.com/tester-army/e2e/commit/d181e2645f24f9f433cd03bb5f7a2cd0cf722511) Thanks [@okwasniewski](https://github.com/okwasniewski)! - An `e2e explore` run posts its own pull request comment. Before, a job that ran `e2e run` and then `e2e explore` replaced the test results comment with the exploration's. Test run comments keep their marker, so the comments already on open pull requests are still updated. An exploration comment an earlier version posted keeps the test run's marker: the next `e2e explore` posts a new comment beside it, and a test run in the same job takes it back. A `--last-failed` rerun no longer folds in an exploration's report that shares its output directory. + ## 0.4.3 ### Patch Changes diff --git a/packages/github/package.json b/packages/github/package.json index a7174a60a9..e6bf08d75b 100644 --- a/packages/github/package.json +++ b/packages/github/package.json @@ -1,6 +1,6 @@ { "name": "@e2e-dev/github", - "version": "0.4.3", + "version": "0.4.4", "description": "GitHub reporter for e2e: one pull request comment per run, from GitHub Actions", "keywords": [ "e2e", diff --git a/packages/limrun/CHANGELOG.md b/packages/limrun/CHANGELOG.md index ce8f23d57b..871620a739 100644 --- a/packages/limrun/CHANGELOG.md +++ b/packages/limrun/CHANGELOG.md @@ -1,5 +1,17 @@ # @e2e-dev/limrun +## 0.2.3 + +### Patch Changes + +- [#1420](https://github.com/tester-army/e2e/pull/1420) [`512eed7`](https://github.com/tester-army/e2e/commit/512eed789dd1234011c5997b680ab1b90e3f8772) Thanks [@okwasniewski](https://github.com/okwasniewski)! - Bump agent-device to 0.21.33. + +- [#1458](https://github.com/tester-army/e2e/pull/1458) [`450be72`](https://github.com/tester-army/e2e/commit/450be721e46a0bc024d873db85b5a7a6c3692965) Thanks [@kacperkapusciak](https://github.com/kacperkapusciak)! - Update agent-device and @agent-device/testmu to 0.21.35. An unscoped boot picks the newest of several shut-down devices, and on iOS, pressing return with nothing focused keeps the runner session. + +- [#1382](https://github.com/tester-army/e2e/pull/1382) [`3a8d6f1`](https://github.com/tester-army/e2e/commit/3a8d6f1b6cdaa4c7975c55bf25f67d2e76143b21) Thanks [@okwasniewski](https://github.com/okwasniewski)! - Bump agent-device to 0.21.31, zod to 4.6.7, @modelcontextprotocol/server to 2.4.0, and oxc to 0.155.0. + +- [#1449](https://github.com/tester-army/e2e/pull/1449) [`abb10cf`](https://github.com/tester-army/e2e/commit/abb10cfe154cd979ccec7747b2c799b4929d4c30) Thanks [@bheemreddy-samsara](https://github.com/bheemreddy-samsara)! - Update agent-device to 0.21.34. Hosted WebDriver fills replace the confirmed focused field instead of appending, and an unsupported active-element lookup refuses before tapping. The pre-tap focus probe no longer reads the previous field's geometry. + ## 0.2.2 ### Patch Changes diff --git a/packages/limrun/package.json b/packages/limrun/package.json index 89591bc077..6c3354a7a7 100644 --- a/packages/limrun/package.json +++ b/packages/limrun/package.json @@ -1,6 +1,6 @@ { "name": "@e2e-dev/limrun", - "version": "0.2.2", + "version": "0.2.3", "description": "Limrun hosted iOS simulators and Android emulators for e2e mobile tests", "keywords": [ "e2e", diff --git a/packages/mobile/CHANGELOG.md b/packages/mobile/CHANGELOG.md index 7ec5582c6f..9751bb4857 100644 --- a/packages/mobile/CHANGELOG.md +++ b/packages/mobile/CHANGELOG.md @@ -1,5 +1,41 @@ # @e2e-dev/mobile +## 0.12.5 + +### Patch Changes + +- [#1379](https://github.com/tester-army/e2e/pull/1379) [`9e53c40`](https://github.com/tester-army/e2e/commit/9e53c40a94fed6585d4e57cf14b8e91c405cb5d8) Thanks [@okwasniewski](https://github.com/okwasniewski)! - A mobile target without `device` skips a booted simulator or emulator that another agent-device session holds, such as one another tool is driving, and drives the free ones. When every booted device is held, the run fails at `prepare` naming the holding sessions instead of failing on the first one. + +- [#1420](https://github.com/tester-army/e2e/pull/1420) [`512eed7`](https://github.com/tester-army/e2e/commit/512eed789dd1234011c5997b680ab1b90e3f8772) Thanks [@okwasniewski](https://github.com/okwasniewski)! - Bump agent-device to 0.21.33. + +- [#1458](https://github.com/tester-army/e2e/pull/1458) [`450be72`](https://github.com/tester-army/e2e/commit/450be721e46a0bc024d873db85b5a7a6c3692965) Thanks [@kacperkapusciak](https://github.com/kacperkapusciak)! - Update agent-device and @agent-device/testmu to 0.21.35. An unscoped boot picks the newest of several shut-down devices, and on iOS, pressing return with nothing focused keeps the runner session. + +- [#1382](https://github.com/tester-army/e2e/pull/1382) [`3a8d6f1`](https://github.com/tester-army/e2e/commit/3a8d6f1b6cdaa4c7975c55bf25f67d2e76143b21) Thanks [@okwasniewski](https://github.com/okwasniewski)! - Bump agent-device to 0.21.31, zod to 4.6.7, @modelcontextprotocol/server to 2.4.0, and oxc to 0.155.0. + +- [#1380](https://github.com/tester-army/e2e/pull/1380) [`c3c4238`](https://github.com/tester-army/e2e/commit/c3c4238e8d47caeebf2926f522e8cd8b8fb7cb51) Thanks [@okwasniewski](https://github.com/okwasniewski)! - Add the unstable `device.unstable_raw`, which may change or go away in any release: the test's agent-device session name, the device's id and name, and an agent-device `client` bound to that session, for what the `device` methods don't cover. Client calls aren't recorded as steps, replayed from the agent cache, or redacted. + +- [#1334](https://github.com/tester-army/e2e/pull/1334) [`6c8e71f`](https://github.com/tester-army/e2e/commit/6c8e71f1ddab036d5cd82ca671163de6e9e82ca0) Thanks [@21prnv](https://github.com/21prnv)! - `locator.filter()` takes `hasNot`, which drops matches whose subtree resolves a locator, as in Playwright. Third-party engines must handle the new `hasNot` field of a `filter` expression. + +- [#1410](https://github.com/tester-army/e2e/pull/1410) [`b309bcd`](https://github.com/tester-army/e2e/commit/b309bcdc89d5b990a7908ef8ea8e3ae2dadfc4b1) Thanks [@bheemreddy-samsara](https://github.com/bheemreddy-samsara)! - Re-read mobile fill and clear targets before typing, including cached replay. Missing inputs, changed security state, or incomplete captures retry after a fresh observation. + +- [#1449](https://github.com/tester-army/e2e/pull/1449) [`abb10cf`](https://github.com/tester-army/e2e/commit/abb10cfe154cd979ccec7747b2c799b4929d4c30) Thanks [@bheemreddy-samsara](https://github.com/bheemreddy-samsara)! - Update agent-device to 0.21.34. Hosted WebDriver fills replace the confirmed focused field instead of appending, and an unsupported active-element lookup refuses before tapping. The pre-tap focus probe no longer reads the previous field's geometry. + +- [#1427](https://github.com/tester-army/e2e/pull/1427) [`75ae2ab`](https://github.com/tester-army/e2e/commit/75ae2aba3ee71959ee4f3d83e1864e7096dc1dfc) Thanks [@kacperkapusciak](https://github.com/kacperkapusciak)! - Read the placeholder an empty Compose Multiplatform field folds into its iOS label as the field's placeholder, so an empty field no longer reads as holding the example text. `fill` and `clear` type into an unclassified iOS element that holds a value or only text, the way a custom-drawn Compose Multiplatform or Flutter field can read. + +- [#1297](https://github.com/tester-army/e2e/pull/1297) [`bdd7315`](https://github.com/tester-army/e2e/commit/bdd7315b7f74840581a2ce72b77f251ef05047fa) Thanks [@shamilovtim](https://github.com/shamilovtim)! - A test step on a control the keyboard moves now waits for it to land instead of tapping where it was. Moving-target refusals are retried when agent-device sent no input. + +- [#1307](https://github.com/tester-army/e2e/pull/1307) [`64432f1`](https://github.com/tester-army/e2e/commit/64432f10891d9772494137b0ba328d8ef440deae) Thanks [@malopezr7](https://github.com/malopezr7)! - Add `@e2e-dev/mobile/lease` for device-provider packages: `agentDeviceLeases()` leases devices from a cloud agent-device runs sessions on, one isolated daemon per worker slot, and fails a release that agent-device can't confirm ended the session. + +- [#1413](https://github.com/tester-army/e2e/pull/1413) [`eeacc8a`](https://github.com/tester-army/e2e/commit/eeacc8aa394362037d94573a11c98c906a464acc) Thanks [@samanyudas](https://github.com/samanyudas)! - A mobile action on a node that a newer snapshot replaced finds the node again instead of acting on a different element. + +- [#1300](https://github.com/tester-army/e2e/pull/1300) [`4af1aa8`](https://github.com/tester-army/e2e/commit/4af1aa83127820eb0a352597e98b42bd05688900) Thanks [@shamilovtim](https://github.com/shamilovtim)! - Re-resolve a locator when its target changes during the transition wait, instead of tapping a stale native reference that may now identify another control. + +- [#1037](https://github.com/tester-army/e2e/pull/1037) [`a481fd3`](https://github.com/tester-army/e2e/commit/a481fd39c69e3e0f88a94c1fe3f4acada164cbf6) Thanks [@rudycelekli](https://github.com/rudycelekli)! - `locator.filter()` takes `hasNotText`, which drops matches whose text contains a string or matches a RegExp, as in Playwright. Third-party engines must handle the new `hasNotText` field of a `filter` expression. + +- [#1421](https://github.com/tester-army/e2e/pull/1421) [`46df78d`](https://github.com/tester-army/e2e/commit/46df78d9bab769583b198be32851f93df91e1874) Thanks [@okwasniewski](https://github.com/okwasniewski)! - The `open_app` agent tool returns to the target's app when called without an app, so the agent can recover after a system sheet or another app takes the foreground. Device errors no longer end with agent-device's CLI advice, such as "Run apps to discover...". + +- [#1421](https://github.com/tester-army/e2e/pull/1421) [`46df78d`](https://github.com/tester-army/e2e/commit/46df78d9bab769583b198be32851f93df91e1874) Thanks [@okwasniewski](https://github.com/okwasniewski)! - `press('Enter')` on a field taps it first when it is not focused, as a character key already did. Before, Enter went to whatever held focus, and on iOS failed when the keyboard was down. + ## 0.12.4 ### Patch Changes diff --git a/packages/mobile/package.json b/packages/mobile/package.json index 783bcc55ea..135458aeb0 100644 --- a/packages/mobile/package.json +++ b/packages/mobile/package.json @@ -1,6 +1,6 @@ { "name": "@e2e-dev/mobile", - "version": "0.12.4", + "version": "0.12.5", "description": "Mobile engine (iOS and Android) for e2e agentic end-to-end testing, built on agent-device", "keywords": [ "e2e", diff --git a/packages/testmu/CHANGELOG.md b/packages/testmu/CHANGELOG.md index 80422a34fd..f5ddab1ae0 100644 --- a/packages/testmu/CHANGELOG.md +++ b/packages/testmu/CHANGELOG.md @@ -1,5 +1,19 @@ # @e2e-dev/testmu +## 0.0.3 + +### Patch Changes + +- [#1420](https://github.com/tester-army/e2e/pull/1420) [`512eed7`](https://github.com/tester-army/e2e/commit/512eed789dd1234011c5997b680ab1b90e3f8772) Thanks [@okwasniewski](https://github.com/okwasniewski)! - Bump agent-device to 0.21.33. + +- [#1458](https://github.com/tester-army/e2e/pull/1458) [`450be72`](https://github.com/tester-army/e2e/commit/450be721e46a0bc024d873db85b5a7a6c3692965) Thanks [@kacperkapusciak](https://github.com/kacperkapusciak)! - Update agent-device and @agent-device/testmu to 0.21.35. An unscoped boot picks the newest of several shut-down devices, and on iOS, pressing return with nothing focused keeps the runner session. + +- [#1382](https://github.com/tester-army/e2e/pull/1382) [`3a8d6f1`](https://github.com/tester-army/e2e/commit/3a8d6f1b6cdaa4c7975c55bf25f67d2e76143b21) Thanks [@okwasniewski](https://github.com/okwasniewski)! - Bump agent-device to 0.21.31, zod to 4.6.7, @modelcontextprotocol/server to 2.4.0, and oxc to 0.155.0. + +- [#1449](https://github.com/tester-army/e2e/pull/1449) [`abb10cf`](https://github.com/tester-army/e2e/commit/abb10cfe154cd979ccec7747b2c799b4929d4c30) Thanks [@bheemreddy-samsara](https://github.com/bheemreddy-samsara)! - Update agent-device to 0.21.34. Hosted WebDriver fills replace the confirmed focused field instead of appending, and an unsupported active-element lookup refuses before tapping. The pre-tap focus probe no longer reads the previous field's geometry. + +- [#1307](https://github.com/tester-army/e2e/pull/1307) [`64432f1`](https://github.com/tester-army/e2e/commit/64432f10891d9772494137b0ba328d8ef440deae) Thanks [@malopezr7](https://github.com/malopezr7)! - A device release now fails the run when agent-device can't confirm it ended the TestMu AI session: a forced daemon stop, a lost shutdown report, or a daemon that exited during the run. Needs `@e2e-dev/mobile` 0.12.5 or later. + ## 0.0.2 ### Patch Changes diff --git a/packages/testmu/package.json b/packages/testmu/package.json index bdca1cad14..00ee671cf6 100644 --- a/packages/testmu/package.json +++ b/packages/testmu/package.json @@ -1,6 +1,6 @@ { "name": "@e2e-dev/testmu", - "version": "0.0.2", + "version": "0.0.3", "description": "TestMu AI (formerly LambdaTest) for e2e: hosted Android emulators, iOS simulators, and real devices for mobile tests, and hosted Chrome and Edge for web tests", "keywords": [ "e2e", diff --git a/packages/web/CHANGELOG.md b/packages/web/CHANGELOG.md index a73f6e2f12..c0098b3ba4 100644 --- a/packages/web/CHANGELOG.md +++ b/packages/web/CHANGELOG.md @@ -1,5 +1,27 @@ # @e2e-dev/web +## 0.15.4 + +### Patch Changes + +- [#1334](https://github.com/tester-army/e2e/pull/1334) [`6c8e71f`](https://github.com/tester-army/e2e/commit/6c8e71f1ddab036d5cd82ca671163de6e9e82ca0) Thanks [@21prnv](https://github.com/21prnv)! - `locator.filter()` takes `hasNot`, which drops matches whose subtree resolves a locator, as in Playwright. Third-party engines must handle the new `hasNot` field of a `filter` expression. + +- [#1048](https://github.com/tester-army/e2e/pull/1048) [`7f4475c`](https://github.com/tester-army/e2e/commit/7f4475cef9a6eed5b187e0ba898769d7aab15fe3) Thanks [@rudycelekli](https://github.com/rudycelekli)! - Add `browser.waitForDownloads(trigger, { count })` to collect several files from one trigger. `waitForDownload` now also bounds completion and saving by its timeout. + +- [#1037](https://github.com/tester-army/e2e/pull/1037) [`a481fd3`](https://github.com/tester-army/e2e/commit/a481fd39c69e3e0f88a94c1fe3f4acada164cbf6) Thanks [@rudycelekli](https://github.com/rudycelekli)! - `locator.filter()` takes `hasNotText`, which drops matches whose text contains a string or matches a RegExp, as in Playwright. Third-party engines must handle the new `hasNotText` field of a `filter` expression. + +- [#1044](https://github.com/tester-army/e2e/pull/1044) [`555e226`](https://github.com/tester-army/e2e/commit/555e226dc5e44132a9e34ab3043c777f828d45f6) Thanks [@rudycelekli](https://github.com/rudycelekli)! - Add `route.fetch()` to read the upstream response inside a route handler before deciding it. + +- [#1288](https://github.com/tester-army/e2e/pull/1288) [`8ce2c65`](https://github.com/tester-army/e2e/commit/8ce2c6533caf0596b16d31d3588cb6720c863176) Thanks [@DeryFerd](https://github.com/DeryFerd)! - Screenshots no longer change the inline style of editable elements to hide the text caret, so a server-rendered app hydrating while a capture runs no longer reports a React attribute mismatch on its inputs and textareas. A capture runs with the caret hidden in the frames it can reach, and the stylesheet is taken back from every frame afterwards. + +- [#1046](https://github.com/tester-army/e2e/pull/1046) [`5879bfc`](https://github.com/tester-army/e2e/commit/5879bfc6ba9a76c6d5fc77f1454ff3fde9b2e5a3) Thanks [@rudycelekli](https://github.com/rudycelekli)! - Add selective cookie clearing with validated name, domain and path filters, preserving other browser storage. + +- [#958](https://github.com/tester-army/e2e/pull/958) [`85e92f9`](https://github.com/tester-army/e2e/commit/85e92f91f1c5415850ca4456ee5f4e2a81a6b648) Thanks [@kuishou68](https://github.com/kuishou68)! - `upload` and `setInputFiles` accept a button that opens a file chooser, such as one over a hidden file input. The agent's `upload` tool says so, so the model can pick the button it sees. + +- [#1429](https://github.com/tester-army/e2e/pull/1429) [`3a841bb`](https://github.com/tester-army/e2e/commit/3a841bb99e5ecc050154c9e6d7f696a6226c3980) Thanks [@KrzysztofMoch](https://github.com/KrzysztofMoch)! - A `web()` target takes `app.permissions` with `'grant'` states, granted to every page of the attempt: `permissions: { 'clipboard-read': 'grant', 'clipboard-write': 'grant' }` lets the app use the clipboard in headless Chromium. Names are checked against the browser's at config load; `'deny'` and `'reset'` are refused on web. + +- [#1448](https://github.com/tester-army/e2e/pull/1448) [`cdf910a`](https://github.com/tester-army/e2e/commit/cdf910a9ae34569d8fddd4e720e0cabd1e95c0f3) Thanks [@KrzysztofMoch](https://github.com/KrzysztofMoch)! - Screenshots no longer fail with `ENGINE_FAILURE` ("closed-shadow root tracking is unavailable") on a page that replaces the global `WeakMap`, as some polyfills do. On such pages, locators find elements inside closed shadow roots again, and CDP reconnects no longer fail. + ## 0.15.3 ### Patch Changes diff --git a/packages/web/package.json b/packages/web/package.json index a2d2063edb..a574518eee 100644 --- a/packages/web/package.json +++ b/packages/web/package.json @@ -1,6 +1,6 @@ { "name": "@e2e-dev/web", - "version": "0.15.3", + "version": "0.15.4", "description": "Web engine for e2e agentic end-to-end testing, built on Playwright", "keywords": [ "e2e", diff --git a/plugins/e2e/.claude-plugin/plugin.json b/plugins/e2e/.claude-plugin/plugin.json index d74607f8fd..ad71c89cb4 100644 --- a/plugins/e2e/.claude-plugin/plugin.json +++ b/plugins/e2e/.claude-plugin/plugin.json @@ -1,6 +1,6 @@ { "name": "e2e", - "version": "0.21.3", + "version": "0.22.0", "description": "Write, run, and debug end-to-end tests for web and mobile apps with e2e, the open source agentic test runner.", "author": { "name": "TesterArmy", diff --git a/plugins/e2e/.codex-plugin/plugin.json b/plugins/e2e/.codex-plugin/plugin.json index de776faa97..4db88de4d6 100644 --- a/plugins/e2e/.codex-plugin/plugin.json +++ b/plugins/e2e/.codex-plugin/plugin.json @@ -1,6 +1,6 @@ { "name": "e2e", - "version": "0.21.3", + "version": "0.22.0", "description": "Write, run, and debug end-to-end tests for web and mobile apps with e2e, the open source agentic test runner.", "author": { "name": "TesterArmy", diff --git a/plugins/e2e/.cursor-plugin/plugin.json b/plugins/e2e/.cursor-plugin/plugin.json index 1f42a3dba8..3fc74a2709 100644 --- a/plugins/e2e/.cursor-plugin/plugin.json +++ b/plugins/e2e/.cursor-plugin/plugin.json @@ -1,6 +1,6 @@ { "name": "e2e", - "version": "0.21.3", + "version": "0.22.0", "description": "Write, run, and debug end-to-end tests for web and mobile apps with e2e, the open source agentic test runner.", "author": { "name": "TesterArmy",