PulseCast web

Nuxt 4 SSR frontend for PulseCast — a staffing copilot that turns inner-city footfall into a usable next-week shift plan for a Heilbronn retailer.

The hero screen is /plan. The map, weather and forecast are supporting evidence; the decision is the weekly plan.

Stack

  • Nuxt 4.6.4, ssr: true, Nitro node-server
  • @sit-onyx/nuxt 1.2.0 + sit-onyx 1.22.0 for the entire shell and UI
  • Mock JSON in mock/, mirrored by Nitro routes in server/api/
  • Shared plan heuristic in shared/staffing.ts

Onyx is used for the navigation/app shell, page layout, cards, tables, buttons, badges, tags, selects, segmented control, slider, stepper, accordion and file upload. The app does not ship a parallel component design system.

Run

npm ci
npm run dev          # http://localhost:3000
npm run build        # production SSR build in .output/
npm run preview      # preview the built app
npm test             # heuristic unit tests
npm run mock         # regenerate deterministic mock JSON

The built SSR server can also be started directly:

PORT=3000 node .output/server/index.mjs

Screens

Route Purpose
/ Now: replay reading, 8-sensor schematic, weather, holiday, data versions
/forecast 7-day daily forecast, uncertainty band, weather strip
/plan Hero. Peak decision, shift blocks, coverage, cost, why, assumptions
/simulate Scenario comparison (base vs sunny/rainy/Wednesday event)
/about Method, metrics, limits, privacy, mock training form

The layout is checked at 390 px (single column, mobile Onyx nav, day cards) and 1440 px (Onyx horizontal nav, two-column plan with tables).

Mock API

All mock routes return shapes from /home/noel/Documents/Hackathon_2026/CONTRACT.md and are bundled into the Nitro output, so the UI renders before the Go API exists:

GET  /api/healthz
GET  /api/readyz
GET  /api/meta
GET  /api/sensors
GET  /api/context
GET  /api/series?sensor=&step=
GET  /api/forecast?sensor=&horizon=&scenario=
GET  /api/staffing?sensor=&scenario=&standard=
GET  /api/scenario
POST /api/scenario
POST /api/train
GET  /api/train/:job_id

NUXT_PUBLIC_API_BASE and NUXT_PUBLIC_DEMO_NOW are runtime config hooks for the next integration slice. The current pages call the Nitro mock routes.

Plan logic

shared/staffing.ts is the single implementation of the contract heuristic:

required = ceil(max(0, footfall) / visitors_per_staff_hour * safety_factor)
required = max(required, min_staff)

Store hours are 10:00–20:00 Europe/Berlin; default blocks are 4 h; labour cost is €16/h; the default service standard is 25 visitors per staff-hour. Coverage is at or above requirement by construction because each block is sized to its local peak — the UI shows the resulting overstaffed hours instead of hiding them.

What is still mocked

  • Forecast, series, weather, holidays, sensor metadata, model metrics and versions are deterministic JSON in mock/. Only 7d has a full hourly mock; 15m, 1h and 48h are slices of that hourly series so the contract shape is available to the next integration step.
  • POST /api/train and GET /api/train/:job_id return an immediate success with sample metrics. There is no real training run in the web tier.
  • The sensor map is a schematic of relative positions. It does not yet overlay the static MapLibre/basemap asset and makes no city-wide heatmap claim.
  • Scenario changes are local; SSE (/api/stream) and the 5 s polling fallback are not wired yet. The plan recomputes from the same deterministic heuristic.
  • Approve/Export CSV are client-side actions; the CSV is real, the approval is demo state only.
  • No auth, no POS/sales/queue/payroll data, no live sensor feed.

Docker

docker build -t pulsecast-web .
docker run --rm -p 3000:3000 pulsecast-web

The runtime stage is node:22-alpine, copies only .output/, and runs as the non-root node user.

S
Description
Frontend with SSR and triggered via SSE to get a touch and feel
Readme
2.1 MiB
0 Stars 1 Watchers 0 Forks
Languages
Vue 63%
TypeScript 20.9%
CSS 7.1%
JavaScript 5.9%
HTML 2.9%
Other 0.1%