Files
liqa/CHANGELOG.md
T
ars9 d4bb274982
CI / changes (push) Skipped
CI / server (push) Failing after 34m15s
CI / client (push) Successful in 2m43s
CI / image (push) Skipped
chore(release): cut minor release
- bump version to 1.26.0
- add versioned e2e test export to the changelog
2026-10-08 18:54:34 +03:00

449 lines
23 KiB
Markdown

# Changelog
All notable changes to this project will be documented in this file.
## 1.26.0 - 2026-10-08
Changes since 1.25.0:
### Added
- e2e test export for a chosen scenario version: `GET /api/v1/scenarios/:id/export-e2e` and the `export_e2e_test` MCP tool take an optional `version` (defaults to the latest), and the test project is built from that version's steps, name and timeout. The Export e2e test button in the UI exports the version being viewed.
### Changed
- Exported e2e test archives include the version in their filename (`<scenario name>-v<N>-e2e-test.zip`), and the version also appears in the generated package name and README heading.
## 1.25.0 - 2026-10-07
Changes since 1.24.4:
### Added
- Browser notifications for scenario runs: the UI shows "Run started / passed / failed: <scenario>" notifications for runs started from the UI, the REST API, MCP or the scheduler; clicking one opens the run. Turn them on with the bell button in the side panel footer (granting the permission from the browser's prompt or site settings works too). Notifications need a secure context (HTTPS or `localhost`).
- `GET /api/v1/scenarios/runs/events`: a server-sent events stream of run lifecycle events (`run.started`, `run.finished`) with a periodic `ping` heartbeat.
### Changed
- Runs lists refresh as soon as a run finishes while notifications are on.
## 1.24.4 - 2026-10-07
Changes since 1.24.3:
### Changed
- Exported scenario and snippet files include the exported version in their filename: `<scenario name>-v<N>.yaml` and `snippet-<alias>-v<N>.yaml`.
### Fixed
- `scripts/install.sh` reads prompts from the terminal, so `curl ... | bash` installs no longer consume the script's own lines as answers (it falls back to defaults when there is no terminal). It also checks that Docker, the compose plugin and the daemon work, and starts liqa with `docker compose up -d`, printing the manual start command if a check or the start fails.
## 1.24.3 - 2026-10-02
Changes since 1.24.2:
### Added
- `examples/` directory with ready-to-import example resources. The first one, `examples/self-test`, is a scenario that tests the liqa instance it runs on: it creates a temporary scenario through the REST API, runs it, checks the run and its page in the UI, and cleans up. It uses an environment, a credential, snippets and assertions.
### Changed
- CI builds and pushes the Docker image only when a `vX.Y.Z` release tag is pushed, not on every merge to `master`.
## 1.24.2 - 2026-10-02
Changes since 1.24.1:
### Added
- Large run log data blocks start collapsed behind a "Show data (size, lines)" toggle: data whose pretty-printed JSON is over `RUN_LOG_DATA_COLLAPSE_BYTES` (default 256) bytes or `RUN_LOG_DATA_COLLAPSE_LINES` (default 5) lines. An "Expand data" / "Collapse data" button next to the Logs heading opens or closes all of them.
- "Copy logs" button on the run page: copies the currently filtered logs to the clipboard as plain text, with every data block expanded.
- `GET /manifest.json`: public client configuration served by the server - the app name and version, and the run log data limits. The UI caches it in `localStorage` and revalidates it on page load and when the tab is refocused after a while.
### Changed
- The sidebar footer shows the server's version (from `/manifest.json`); its tooltip suggests reloading the page when it differs from the version the UI was built with.
- Run log data blocks are aligned with the context/message column instead of the level column.
## 1.24.1 - 2026-10-02
Changes since 1.24.0:
### Added
- `context.assert(fn, description, details?)`: optional `details` argument. On failure, the `assertion:fail` log entry stores `{ details, error }` as data (`error` is the exception message when `fn` threw), so large diagnostics such as a response body can stay out of the message. The exported e2e runtime accepts and prints it too.
- The "Step N passed" log entry carries the step's return value as data.
- File names in the run Artifacts and scenario Files tables link to the file content, opening it in a new tab (shown inline when the browser can display it).
### Changed
- For assertion failures, the "Step N failed" and "Snippet X failed" log entries name only the assertion (`Assertion failed: <description>`) instead of repeating the failure details. The step run's description keeps the full message.
- Log output shows elapsed time in whole seconds (`+14s`), matching the stored timestamp resolution.
## 1.24.0 - 2026-10-01
Changes since 1.23.0:
### Added
- Code reference in the UI: a (?) button next to the Code label on the step and snippet create/edit pages opens a modal documenting the script runtime - the globals (`context`, `expect`, `console`, and `args` in snippets) and every `context` member with examples, including pitfalls such as awaiting `getStepOutput()` and `assert()`, and `downloadFile()` not sending the browser's cookies.
### Changed
- `Modal` UI component: new `size="lg"` for long, document-like content. Dialogs are now capped to the viewport height and their body scrolls, instead of overflowing the screen.
## 1.23.0 - 2026-10-01
Changes since 1.22.2:
### Added
- Run log entries now carry a `context` (where the message was generated: `step:<N>`, `snippet:<alias>`, `assertion:pass`/`fail`, or `run`) and a `data` field: `context.log`/`warn`/`error` and `console.*` split their arguments so strings join into `message` while object/array arguments are kept separate as `data` (JSON), instead of being `JSON.stringify`'d into the message text.
- Added a `debug` log level (`context.debug()`, `console.debug()`), alongside `log`/`warn`/`error`.
- New `LogOutput` UI component: a terminal-style log viewer on the run detail page, color-coded by level, with an elapsed-time-since-start column, a context tag, and `data` rendered pretty-printed in a separate block below the message. The full log list is shown (no more pagination), capped to a scrollable block.
- Run log CSV/Markdown export (`GET /scenarios/:id/runs/:runId/export`) include the new `context` and `data` columns/blocks; the logs search (`?q=`) also matches `context`.
## 1.22.2 - 2026-10-01
Changes since 1.22.1:
### Added
- MIT license (`LICENSE`), declared in all `package.json` files and the README.
- The Docker image includes the license notices of everything it ships: `scripts/third-party-licenses.mjs` collects the license of every production dependency of the client and server, the Node.js license is added, and the result is served at `/licenses.txt`, linked as "Licenses" in the sidebar footer.
- OCI image labels for title, version, license and source; CI passes the version as the `VERSION` build arg.
### Fixed
- The Docker image keeps packages' `LICENSE.md`/`NOTICE.md` files, which the 1.22.1 slimming step deleted along with other Markdown files.
## 1.22.1 - 2026-10-01
Changes since 1.22.0:
### Changed
- The Docker image is about 20% smaller (1.09 GB → 876 MB on disk, 251 MB → 220 MB compressed). Production dependencies and Chromium are installed in a separate stage that drops files nothing loads at runtime (source maps, TypeScript sources and declarations, docs, TypeORM's browser build, better-sqlite3's C sources, Playwright's ffmpeg). The runtime image is `debian:bookworm-slim` with only the `node` binary, so it no longer contains `npm`, `npx`, `yarn` or `corepack`.
## 1.22.0 - 2026-10-01
Changes since 1.21.1:
### Added
- The Export Scenario dialog lists the snippets the scenario uses (including ones used only through other snippets), all checked by default. A selected snippet is exported once per version the scenario uses: the latest for `runSnippet("alias")`, plus each version pinned with `alias@N`. `GET /scenarios/:id/export` takes the new `snippetIds` query parameter for this.
- The Export Scenario dialog shows the scenario version next to its name and id.
## 1.21.1 - 2026-10-01
Changes since 1.21.0:
### Changed
- The credential page hides the credential data when it opens; a Reveal button shows it and Hide hides it again. The data is not rendered on the page until revealed.
### Added
- A `RevealableCodeBlock` UI component (a CodeBlock that starts hidden behind a Reveal/Hide button), with a Storybook story.
## 1.21.0 - 2026-10-01
Changes since 1.20.2:
### Added
- UUID badges show a small colored circle in front of the short id. Its color comes from the first 6 hex digits of the UUID, so the same id always gets the same color, and its border uses the badge text color to stay visible in both themes.
- UUID badges get lighter on hover to show they can be clicked.
- `Badge` accepts an optional `className`.
- A Storybook "Colors" story for UuidBadge with six sample ids and a button to randomize them.
## 1.20.2 - 2026-09-30
Changes since 1.20.1:
### Fixed
- Text contrast now meets WCAG AA (4.5:1) in both themes:
- the "Live" refresh label and the table sort icon use a new `--color-primary-text` token (dark lime in the light theme) instead of the light lime brand color;
- the idle refresh label is no longer dimmed with opacity;
- the light theme's danger button is a little darker, and the dark theme's danger hover is lighter so its dark text stays readable;
- syntax highlighting colors are defined per theme (`--hl-*` tokens), darker in the light theme;
- muted text is slightly darker in the light theme.
### Added
- A Storybook "Accessibility / Contrast" page that measures the contrast of the most contrast-prone components in both themes.
- Storybook accessibility checks now fail the tests on violations (`a11y.test: 'error'`). The Colors story shows each token's current value per theme and includes the syntax colors.
## 1.20.1 - 2026-09-30
Changes since 1.20.0:
### Added
- Storybook stories for AutoRefreshIndicator, CodeBlock, CodeEditor, MarkdownContent, Modal, Notification, Search, Textarea, Toast and UuidBadge, with interaction tests where the component has behavior. The DescriptionList story covers all four layouts.
### Changed
- Updated all client and server dependencies to the latest versions within their ranges, including React 19.3, Storybook 10.6, NestJS 11.2, TypeORM 0.3.31 and Playwright 1.63 (the Docker image now ships HeadlessChrome 153).
- The Storybook tests are part of the pre-commit checks.
### Fixed
- Code flagged by the stricter React hooks lint rules: the runs list no longer writes refs during render, and the scenario list, scenario files and sessions list no longer set state synchronously inside effects.
## 1.20.0 - 2026-09-30
Changes since 1.19.3:
### Added
- Database migrations: the SQLite schema is now managed by TypeORM migrations (`server/src/database/migrations/`) instead of `synchronize`. The Docker container applies pending migrations before starting the server, and the server refuses to start while migrations are pending.
- `npm -w server run migration:run`, `migration:show` and `migration:generate` scripts; the workflow for schema changes is described in `docs/development.md`.
- A test that fails when an entity changes without a matching migration.
### Changed
- Databases created before migrations (by `synchronize`) are adopted automatically: the initial migration is recorded as applied without changing them.
- Tests now build their database from the migrations instead of `synchronize`.
### Fixed
- Startup failures were logged but the process exited with code 0; they now exit with code 1.
## 1.19.3 - 2026-09-30
Changes since 1.19.2:
### Changed
- The Docker image runs with `NODE_ENV=production`.
- The origins allowed to call the API cross-origin depend on `NODE_ENV`: the Vite dev server (`http://localhost:5173`, `http://127.0.0.1:5173`) in development, the compose port (`http://localhost:13000`, `http://127.0.0.1:13000`) in production. The new `CORS_ORIGINS` setting (comma separated) overrides both. A `NODE_ENV=development` in `.env` overrides the image's production mode.
- `.env` is optional: compose loads it only when it exists, and every setting in `.env.example` is commented out and shows its default.
### Removed
- The unused `KEYS_DIR` setting and the `./keys` mount in `compose.yaml`.
## 1.19.2 - 2026-09-30
Changes since 1.19.1:
### Fixed
- Running a scenario from its detail page could preselect the first environment in the list instead of the scenario's default environment, depending on whether the environment list or the scenario loaded first. The default environment is now chosen when the run modal opens.
- Made the run status filter test deterministic; it could fail when the scheduler picked up the test's pending run mid-test.
## 1.19.1 - 2026-09-30
Changes since 1.19.0:
### Changed
- Upgraded to Node.js 24: the Docker image is based on `node:24-slim`, and the root `package.json` declares `"engines": { "node": ">=24" }`.
- The server reads the app name and version (startup log, Swagger, MCP server info, `APP_NAME`/`APP_VERSION`) from the workspace root `package.json`. Previously local runs reported the server workspace's own `liqa-server` / `1.0.0`; the server workspace no longer has a version.
### Fixed
- Tests using a local HTTP server bind to `127.0.0.1` instead of `localhost`, which could resolve to an unreachable `::1` inside Docker containers.
## 1.19.0 - 2026-09-30
Changes since 1.18.0:
### Changed
- **Breaking (deployment):** the whole application now runs from a single image and container. The server serves the built client itself (new `CLIENT_DIR` setting, set by the image) next to the API, MCP and Swagger, so the UI moved from http://localhost:13001 to **http://localhost:13000**. The root `Dockerfile` replaces `server/Dockerfile`, `client/Dockerfile` and `client/nginx.conf`, and `compose.yaml` has a single `liqa` service. When upgrading, run `docker compose up -d --build --remove-orphans` to remove the old `server` and `client` containers.
- The image uses Playwright's headless-only Chromium build (`chromium-headless-shell`, always matching the installed Playwright version) with only the libraries it needs, instead of Debian's desktop `chromium` package. Fonts are now DejaVu and Liberation (metric-compatible with Arial/Times) instead of Noto. Together with clearing the npm cache, this shrinks the image from 1.76 GB to 1.07 GB.
## 1.18.0 - 2026-09-30
Changes since 1.17.0:
### Added
- Added a central import: `POST /import`, an Import page in the UI, and the `import_yaml` MCP tool. It takes YAML made of `---`-separated documents, each holding one exported item or a list of items (JSON is accepted too), with any mix of environments, credentials, snippets and scenarios. All items are validated before anything is imported, and they are imported in dependency order (environments, credentials, snippets, scenarios), so a scenario can reference an environment or credential from the same file.
### Changed
- The Import buttons on the scenario, snippet, credential and environment list pages use the central import. When the file holds exactly one item of the page's kind, that item is opened; otherwise the Import page shows everything that was imported.
### Fixed
- Scenario exports that included the environment or credentials were not valid YAML (all entities were written into one document). Each entity is now a separate `---` document.
## 1.17.0 - 2026-09-30
Changes since 1.16.1:
### Added
- Files can be deleted from a run: a delete button (with confirmation) in the Files section of the run detail page, `DELETE /scenarios/:id/runs/:runId/files/:fileId`, and the `delete_run_file` MCP tool. The stored file is removed once no scenario or run references it.
## 1.16.1 - 2026-09-30
Changes since 1.16.0:
### Changed
- Playwright traces are now off by default (`PLAYWRIGHT_TRACE_MODE` defaults to `off`), since a trace often reaches ~10 MB per run. Traces can still be enabled per run in the run modal, via `traceMode` in `POST /scenarios/:id/run`, or for all runs by setting `PLAYWRIGHT_TRACE_MODE`.
## 1.16.0 - 2026-09-30
Changes since 1.15.0:
### Added
- Scenarios and snippets are now versioned, starting at version 1. A scenario's version goes up when its name, description, default environment, timeout or steps change; a snippet's when its alias, title, description or code changes. Saving without changes does not create a new version. Credential assignments and scenario files are not versioned.
- Each run records the scenario version it runs. The run modal has a version picker (latest by default), `POST /scenarios/:id/run` and the `run_scenario` MCP tool accept an optional `version`, and "Repeat run" reuses the run's version.
- `context.runSnippet("alias@N")` pins version N of a snippet; `runSnippet("alias")` keeps using the latest. The scenario page's "Used Snippets" section shows which snippets are pinned.
- Version numbers are shown on the scenario, snippet and run pages and in the scenario and run lists.
- Added `GET /scenarios/:id/versions[/:version]` and `GET /snippets/:id/versions[/:version]`, and the `list_scenario_versions`, `get_scenario_version`, `list_snippet_versions` and `get_snippet_version` MCP tools.
- Scenario and snippet exports accept `?version=N` (the `export_scenario` MCP tool takes `version`).
### Changed
- **Breaking:** scenario and snippet exports now include `version`, and imports require it. An import restores that exact version by id + version: an existing version is left untouched, and a new one is stored under its number and becomes the latest when it is higher than the current latest. Export files from earlier releases need a `version` field added before they can be imported.
- Scenario exports now include scenario and step timeouts.
- Runs keep their own copy of the steps they were created with. Editing or deleting a step no longer changes past runs, and deleting a step no longer deletes its results from past runs. Existing run steps are backfilled from the current steps on startup.
## 1.15.0 - 2026-09-30
Changes since 1.14.0:
### Added
- Scenario runs now record a Playwright trace (screenshots and DOM snapshots) and upload it to the run's files as `trace.zip` when the run finishes. Open it with `npx playwright show-trace trace.zip` or at https://trace.playwright.dev.
- Added the `PLAYWRIGHT_TRACE_MODE` setting (`on` (default), `retain-on-failure`, or `off`) that controls which traces are kept.
- Added a "Playwright trace" option to the run modal to override the trace mode for a single run. It defaults to the server's `PLAYWRIGHT_TRACE_MODE`, served by the new `GET /scenarios/run-defaults` endpoint.
- `POST /scenarios/:id/run` and the `run_scenario` MCP tool accept an optional `traceMode`. The trace mode is stored on the run, shown on the run detail page, and reused by "Repeat run".
### Changed
- The run modals on the scenario list and scenario detail pages now share one component.
- `server/data/` is now git-ignored.
## 1.14.0 - 2026-09-30
Changes since 1.13.0:
### Added
- Added a "Repeat run" button on the scenario run detail page that starts a new run of the same scenario with the same environment and "save session" setting, then navigates to it. The button is disabled when the run's environment has since been deleted.
## 1.13.0 - 2026-09-30
Changes since 1.12.0:
### Added
- Added credential CRUD tools to the MCP server: `list_credentials`, `get_credential`, `create_credential`, `update_credential`, and `delete_credential`. Combined with the existing `add_scenario_credential` tool, credentials can now be created and attached to scenarios entirely via MCP.
## 1.12.0 - 2026-09-16
Changes since 1.11.0:
### Added
- When a scenario or snippet step fails, a full-page screenshot and the current page HTML are captured and uploaded as run artifacts, so failures can be diagnosed from the run's file list.
## 1.11.0 - 2026-09-15
Changes since 1.10.1:
### Added
- Added export of a scenario as a standalone, plug-and-play Playwright e2e test project (zip archive with `package.json`, `playwright.config.ts`, `test.spec.ts`, `README.md`, `.env`, `credentials.json`, `snippets/*.js`, and any attached scenario files) that runs entirely independently of the liqa backend. Available as a "Export e2e test" button on the scenario detail page, `GET /scenarios/:id/export-e2e`, and the `export_e2e_test` MCP tool.
- The exported test only bundles the credentials, snippets, and attached files actually referenced (directly or transitively) by the scenario's steps.
- Added a "Used Snippets" section on the scenario detail page listing the snippets referenced by the scenario's steps, linking to each snippet's detail page. Backed by `GET /scenarios/:id/used-snippets`.
## 1.10.1 - 2026-09-15
Changes since 1.10.0:
### Added
- `context.runSnippet()` now logs "Snippet started"/"finished"/"failed" entries, matching the automatic step start/pass/fail logging.
### Changed
- Reworked the Markdown log export format: each log entry is now its own `## [level] timestamp` section with the message as its body, instead of a single large table.
### Fixed
- Run logs could persist out of call order (e.g. a snippet's "started" log appearing after its "finished" log) because writes were fire-and-forget and raced against each other. Log entries now carry a per-run sequence number assigned in call order, and queries/exports sort by it.
## 1.10.0 - 2026-09-15
Changes since 1.9.1:
### Added
- Added a "Download logs" button on the scenario run detail page: opens a modal to choose Markdown (default) or CSV, then downloads the run's logs in that format.
- Added `GET /scenarios/:id/run/:runId/logs/export?format=csv|md` backend endpoint to export a run's logs.
## 1.9.1 - 2026-09-15
Changes since 1.9.0:
### Added
- Scenario run logs now include an automatic entry when a step starts and when it finishes (pass or fail), so step boundaries are visible in run logs even for steps that don't call `context.log`/`context.assert` themselves.
## 1.9.0 - 2026-09-15
Changes since 1.8.0:
### Added
- Added `context.assert(fn, description)` for scenario/snippet script steps: runs the given function, requires it to return `true`, logs the outcome (`log` on success, `error` on failure or exception), and throws to stop the step on failure.
### Changed
- Documented `context.assert` in `docs/scenario.md`.
## 1.8.0 - 2026-09-15
Changes since 1.7.1:
### Added
- `context.downloadFile()` now accepts `data:` URIs, so scenario steps can save in-script generated content (e.g. a credential JWT) as a run artifact without a network fetch.
### Changed
- Documented the full scenario step `context` file API (`getScenarioFiles`, `downloadFile`) in `docs/scenario.md`.
## 1.7.1 - 2026-04-21
Changes since 1.7.0:
### Added
- Added MCP tools to upload, list, and fetch scenario files using base64 content payloads.
- Added MCP tools to list run artifacts and fetch run artifact content with metadata.
### Changed
- Documented MCP binary content handling for scenario files and run artifacts.
## 1.7.0 - 2026-04-21
Changes since 1.6.1:
### Added
- Added scenario file storage and run artifact support on the backend.
- Added file management UI for scenarios and run details, including artifact lists.
- Added snippet CRUD tools to the MCP server.
### Changed
- Added dynamic page titles based on the active client content.
- Extended the scenarios list to show the latest run status and timestamp.
### Fixed
- Exposed file expiry metadata in scenario and run file responses.
- Disabled expired download actions in the UI while keeping the action visible.