signalk-forecast-dataset-browser
v0.1.0
Published
Signal K webapp to browse forecast datasets, inspect metadata and delete datasets
Maintainers
Readme
Forecast Dataset Browser
A small Signal K webapp for browsing all published forecast datasets, reading available metadata and deleting a dataset. It uses Forecast Dataset Provider's experimental HTTP contract 0.1, on the same Signal K server.
Open Webapps → Forecast Dataset Browser, or
/signalk-forecast-dataset-browser/. The interface is in French, works on desktop
and mobile, and requires no internet connection or external frontend assets.
Stable release: 0.1.0, with Forecast Dataset Provider 0.1.0. See the coordinated upgrade guide when updating the full GRIB chain.
Features
- Separate acquisition monitoring for readwrite/admin sessions: pending, receiving and awaiting-commit publications, file counts and live received bytes. An unknown total uses indeterminate progress, not a fabricated percentage. The complete catalogue refreshes when a monitored publication disappears (commit or abort).
- Complete paginated discovery of all formats and dataset kinds, with search by name, model, UUID or producer, plus format and inventory-state filters.
- Dataset count, published byte total and verified-inventory count. These totals describe the latest published version of each dataset, as exposed by contract 0.1; they are not a disk-usage quota or historical-version size total.
- A detail pane with model cycle/publication times, type, producer, advertised coverage and fields when present, declared metadata, requested parameters, provenance, grid information, identity, content version and metadata revision.
- Original asset identifiers, sizes, SHA-256 hashes and authenticated download links. Verified inventories/indexes load only on explicit request, since they can be large. Complete dataset JSON exposes other available metadata too.
- Explicit deletion confirmation naming the dataset and its UUID. Deletion removes the dataset and every committed version, as defined by the provider. Busy-dataset conflicts are reported; no force option is used.
Missing metadata remains unknown. Requested bounds/duration/fields are shown as acquisition claims, separately from verified inventory. Catalogue 0.1 does not expose a historical-version listing, so the browser shows each dataset's current published version and does not pretend to enumerate older versions.
Authentication and deletion
The webapp uses your existing Signal K browser session, including HttpOnly session cookies. There is no stored application token, login form or backend proxy. Sign in through Signal K's administration interface when needed.
Readonly accounts can list, inspect and download. Readwrite and admin accounts can delete. The server enforces permissions on every request; disabled controls are just a usability aid. Unknown session permissions default to read-only. Signal K installations with authentication disabled follow their existing policy.
Unfinished publication monitoring is also restricted to readwrite/admin sessions, matching the provider's writer-only staging API. It polls every two seconds while publications exist and every five seconds otherwise, pauses requests while the page is hidden, and stops on permission denial. These entries are never counted as published datasets or offered for download. A receiving byte count is not durable progress; interrupted partial files restart at zero. No abort/delete control for staging is added by this webapp.
Before deletion, the browser rereads the current resource and rejects a changed content version or metadata revision. The provider does not offer conditional DELETE, so this check cannot close a simultaneous publication race; the displayed confirmation clearly covers all versions of the dataset. A dataset already deleted elsewhere is removed from the list without another destructive action.
Deleting from this browser does not edit the downloader's settings or local files. A producer may publish a missing acquisition again on its next trigger; stop or reconfigure that producer if you intend to stop future publication. Consumers update when they next refresh their catalogues.
Installation
Requires Signal K 2.34+ within 2.x and enabled
signalk-forecast-dataset-provider 0.1.0 or later, including acquisition monitoring. Node 20.19+ is sufficient for this
webapp. No runtime npm dependencies or build step are required.
npm pack
# Install the archive in your Signal K configuration directory and restart.
npm install /path/to/signalk-forecast-dataset-browser-0.1.0.tgzThe plugin registers its static public/ directory as a Signal K webapp and has
no configuration settings. It only makes same-origin requests to the existing
catalogue and login-status APIs. Reverse-proxy path prefixes are preserved.
Install the stable release from Signal K's App Store, or from the server's
configuration directory with npm install signalk-forecast-dataset-browser@latest.
See npm publishing for repository setup, trusted publishing and
GitHub Release instructions.
Development
npm ci
npm test
npx playwright install chromium
npm run test:browserThe browser suite uses a local static server and intercepted catalogue fixtures. It exercises the detail pane, filters, mobile layout, metadata escaping, confirmation/cancellation, readonly permissions, expired permissions, busy conflicts, changed content and temporary catalogue failure. Unit tests cover pagination, identity checks and the exact deletion contract.
For an already installed browser, set FORECAST_BROWSER_EXECUTABLE to its path.
tools/install-sandbox.cjs and tools/verify-sandbox.cjs target only the selected
signalk-sandbox; see sandbox verification.
Apache-2.0.
