Opsdeck Admin Kit

Describe the company/product and what the team operates. Name the four headline measures for each board, table objects, statuses/priorities, teams/roles, regions/sites and integrations. Optional inputs include a logo and real portraits. Gaps can use explicitly identified example records in your nouns, ID prefixes and reporting currency; do not present drafted figures as company performance.

Rebrand the complete shell and content model, including workspace names, menus, titles, addresses, sample order/task IDs, authentication copy and developer labels. Keep dates internally consistent. The default dataset uses a fixed demonstration clock and seeded series; if the business operates weekly, adapt the documented weekly model and distinguish stocks such as active subscribers from period totals such as cancellations.

Choose the relevant boards and table views, then define one end-to-end record task. Preserve at least an overview and useful collection/detail flow; keep developer pages only if the product has corresponding APIs/integrations. For an internal back office, adapt the sample console subscription billing to the business’s actual payment/payout/refund settings rather than retaining an unrelated SaaS plan.

Work through search, facets, sort, pagination, selection, bulk changes, row actions, drawer edits and Undo using sample records. Name the slice shown by each table so a small record sample does not contradict a whole-business KPI. Review chart units, comparison direction and range aggregation; a visually working chart is not proof its business metric is correctly defined.

With no project, use Node.js 18 or newer and build-time internet to embed the two fonts and portraits into one offline HTML file. Existing/requested frameworks get native routes, shell and typed data behind replaceable async getters. Font guidance includes additional character coverage when staff names need more than the default Latin files.

Use real team photographs, initials or disclosed fictional stand-ins. Only portraits are needed; optional new generation uses the agent’s image tool and may incur charges. Never label a generated face as a verified photograph of a real colleague.

Run the static validator and the supplied Playwright sweep when available, then manually inspect desktop/tablet/phone layouts in both themes. Walk one table end to end and revisit screens before triggering actions to catch duplicate listeners. Test empty/invalid states, keyboard/tooltips, deep-link drawers, Escape, save/discard and auth/error views. Connect database/auth/mail/exports/webhooks separately and verify them before using this as an operational system.