Wall-mounted home dashboard. Hue bridge for room temperature, lights and motion; FMI for weather and PV forecast; SolisCloud for solar production, battery and grid flow. Finnish UI, monochrome with one warm accent.

Done

• Hue temperature, light groups, motion (live SSE).
• FMI weather + 7-day chart, Tomorrow.io fallback.
• PV forecast bars layered on weather chart.
• SolisCloud live: PV, battery SoC + power, grid direction.
• SolisCloud history: 5-min recording, downsampled chart.
• Animated PV ↔ inverter ↔ home / battery / grid energy flow.
• Halo design system: Inter + Space Grotesk, 6px cards, accent + dark themes.
• Sensor temperature history with adaptive sampling.
• Screenshot demo mode (?demo=1) anonymizing names.

Next

• Energy summary panel (donut + daily / monthly metrics).
• Per-light counts on light group toggles (4 / 4).
• Multi-station Solis support.
• Battery low / inverter alarm push notifications.
• Cleaner offline / loading states for SolisCloud + FMI.

TOMORROW_IO_API_KEY=        # tomorrow.io API key
HUE_BRIDGE_ADDRESS=         # optional — auto-discovered via meethue.com if omitted
HUE_BRIDGE_USER=            # obtained via the pairing flow below
LANGUAGE=fi
HALO_IMAGE_TAG=              # version label shown on the front page
HALO_DB_PATH=/data/halo.db   # SQLite database path (default: halo.db in working directory)
HALO_HISTORY_RETENTION_DAYS=90 # days of sensor history to keep (default: 0/disabled)

# Maps room types to lists of room names as configured in the Hue app.
# Rooms not listed default to "inside".
HUE_ROOM_TYPES={"inside": ["Keittiö", "Olohuone"], "inside_cold": ["Kuisti"], "outside": ["Ulkona"]}

# SolisCloud (PV inverter) — required for the energy view + history.
SOLIS_KEY_ID=
SOLIS_KEY_SECRET=
SOLIS_STATION_ID=
SOLIS_BASE_URL=             # default https://www.soliscloud.com:13333

PV forecast vars (PV_LAT, PV_LON, PV_TILT, PV_AZIMUTH, PV_KW) are consumed by scripts/refresh-pv-forecast.sh — see the "PV forecast" section below for details.

1. Press the link button on the bridge
2. POST to /api/hue/pair (optionally pass {"bridgeIp": "x.x.x.x"} in the body if discovery fails)
3. Copy the returned HUE_BRIDGE_ADDRESS and HUE_BRIDGE_USER values into .env

Sensor names are taken directly from the device names set in the Hue app. To differentiate multiple sensors in the same room (e.g. sun vs. shadow), rename the devices in the Hue app.

./install-hooks.sh

Configures a pre-commit hook that runs yarn lint for frontend changes and cargo clippy for backend changes.

yarn dev

Open http://localhost:3000 with your browser to see the result.

Temperature readings from all enabled and connected sensors are automatically recorded to a SQLite database every 5 minutes. Old readings are pruned automatically based on HALO_HISTORY_RETENTION_DAYS (default: 90).

In Docker the database is stored in a named volume (halo-data) at /data/halo.db. In development it defaults to halo.db in the working directory.

GET /api/history/sensors?sensor_id=<id>&hours=<n>

When SolisCloud is configured, the backend polls /v1/api/stationDetail every 5 minutes and writes a row to solis_readings (PV power, grid power, battery SoC + power, today's energy, status). Readings during inverter offline (status=2) are skipped to keep gaps visible in the chart. Retention follows HALO_HISTORY_RETENTION_DAYS.

GET /api/history/solis?hours=<n>&max_points=<m>

The backend exposes GET/POST /api/pv/forecast. The actual forecast is produced by an external one-shot CLI maintained in a separate repository, fmi-pv-forecast-runner, which wraps the FMI open PV forecast package and emits the next ~66 hours of hourly output as JSON on stdout.

Refreshing the forecast is a two-step shell pipeline: run the CLI, pipe its stdout to POST /api/pv/forecast. scripts/refresh-pv-forecast.sh wraps both:

# Auto-detect: uses local uv repo if ../fmi-pv-forecast-runner/ is checked out,
# otherwise pulls the published Docker image.
HALO_PV_ENV_FILE=.env.pv ./scripts/refresh-pv-forecast.sh

# Force a specific mode:
./scripts/refresh-pv-forecast.sh uv
./scripts/refresh-pv-forecast.sh docker

Required env vars (place in $HALO_PV_ENV_FILE or export beforehand): PV_LAT, PV_LON, PV_TILT, PV_AZIMUTH, PV_KW. Optional overrides: HALO_BACKEND_URL (default http://localhost:3000), PV_RUNNER_PATH, PV_RUNNER_IMAGE.

Run the script once manually to populate the database after the first deploy. For a 3-hour refresh cadence (matching the upstream forecast update interval), add a host cron entry:

17 */3 * * * cd /opt/halo && HALO_PV_ENV_FILE=/opt/halo/.env.pv ./scripts/refresh-pv-forecast.sh >>/var/log/halo-pv.log 2>&1

Hourly forecast points are upserted into the pv_forecast_points table (steady-state ≤66 rows) and rendered as bars on the daily weather chart, aggregated to kWh per day.

The FMI forecast covers Finland, Scandinavia, and the Baltic countries. See ilmatieteenlaitos.fi/numerical-weather-prediction for the full available area.

The screenshot at the top of this README is the temperature view: a full-width FMI weather card (current + 4 day-segments + collapsible 7-day chart drawer with PV forecast bars), and a row of cards for ulkona, sisällä, kuisti (averaged Hue sensor readings with low-battery and trend indicators) plus the live Solis solar production card. Tap any card to expand a drawer with per-sensor detail.

Live PV ↔ inverter ↔ home / battery / grid flow diagram from SolisCloud. Active paths animate (dashed scrolling stroke); idle paths stay static at reduced opacity. Each node carries its current kW, plus aurinko shows today's kWh, akku shows SOC %, verkko shows direction (tuonti / vienti / lepotila).

All Hue rooms / zones as tappable toggles. Lit groups get a warm cream background, accent ring, and a softly glowing bulb. Status row shows päällä / pois with a state dot.

All Hue motion sensors in a single list. Active rows show a pulsing accent dot and bold liikettä; idle rows stay muted. Disabled sensors fade out. Last-trigger timestamp on the right uses Finnish relative formatting.

Sensor temperature history from the local SQLite store. Range pills (6h, 24h, 7pv, 30pv) trigger backend-side downsampling so wide ranges stay snappy. Per-sensor colors are derived from room type (cool blues for outside, warm reds for inside, greens for cold-inside).

Location set via address search or device geolocation, plus the deployed image tag.

Append ?demo=1 to any URL to anonymize sensor / room / location names (stable hashed labels like anturi 042, ryhmä 488). Useful for sharing screenshots without leaking topology.