PRMGR Observatory Starter

A free, local-first foundation for a dashboard that helps you understand your home. Inspired by the PRMGR Home Observatory build: muted green panels, warm highlights, an illustrated home, and readings with context.

Download the starter · Connection guide

Version 1.0.0 · No account, subscription, external scripts, analytics, npm dependencies, or build step.

1. Download and unzip the starter package.
2. Open the public folder.
3. Double-click index.html. It opens in your browser.
4. Try Home overview, Heat pump, Explore readings, and Make it yours.

Every initial reading is synthetic demo data, not measurements from Peter’s home or yours. The illustration is a generic original SVG. Nothing connects to Homey or your equipment automatically.

If your browser restricts local files, use the optional local server below. Chrome and Edge are good starting points. The packaged demo has also been tested in Chrome; other browsers have not been formally verified.

• Home overview: solar generation, site consumption, grid import/export, battery charge.
• A house/garage split, only when the supplied sample supports it.
• Hot-water temperature and compressor-frequency history.
• Measurement explorer, latest-sample table, explicit missing data.
• JSON import/export with validation; imports remain in the current tab’s memory.
• A read-only HomeyScript export example you configure for your own devices.
• An optional local Node server and recorder for an adapter-produced JSON file.
• Editable labels and generic house artwork; source code under MIT.

This is a starter, not the complete personal system shown in the video. It does not include automatic device discovery, a Homey login flow, ready-made NIBE/GoodWe/Shelly integrations, electricity billing, device controls, camera feeds, an Android TV APK, or TV Home-button replacement. It is usable in a browser on a large display; installing a native TV app is separate work.

Click Import readings, then choose examples/readings.json. This is another synthetic example, saved with its own timestamps. Switch to Explore readings to see the sample table. Click Export JSON to keep a copy.

Reloading the page returns to demo. Your imports are not sent to a server or saved to browser storage. If you serve the app on a public website, imported readings still stay in the browser; the optional recorder endpoint belongs on a local server.

See the data format to make a file with your own readings.

Install Node.js LTS (Node 22 or later). Open a terminal inside the extracted starter folder, the one containing package.json:

node server.mjs

Or use npm start. You do not need npm install.

Open http://127.0.0.1:8780. This address works only on the computer running the server. It is not a public download link, and another device such as a TV cannot reach it. Stop with Ctrl+C.

Mac users can also run start.command; Windows users can run start-windows.cmd after installing Node. If macOS blocks the script, run node server.mjs in Terminal instead of disabling operating-system protections.

Use three terminal windows, all inside the starter folder:

# Window 1: generate SYNTHETIC input every 30 seconds
node tools/simulate.mjs

# Window 2: save new readings to data/history.json
node collector.mjs --input examples/current-reading.json

# Window 3: show the dashboard
node server.mjs

Click Load local recorder on the dashboard. Click again to refresh; it loads a snapshot, not a continuous stream. The badge identifies synthetic recorder data. Stop each process with Ctrl+C.

The recorder runs independently of the page. Closing the page does not stop it; closing the recorder terminal does. It keeps at most 5,000 samples and preserves their original timestamps. It does not manufacture a new measurement time when an input file hasn’t changed. When recording every 30 seconds, 5,000 samples cover about 42 hours; export/archive history for longer retention.

To move from synthetic data to your own adapter:

1. Stop the synthetic generator and recorder.
2. Archive or rename data/history.json so demo and real readings stay separate.
3. Have your adapter write a JSON file in the documented format. Use an atomic rename so the recorder never reads a partly written file.
4. Start node collector.mjs --input /path/to/your/readings.json.
5. Open the dashboard and load the local recorder.

The recorder only reads that local file. Building the connection between your equipment and that file is a separate step. See connections and Homey.

Edit public/config.js in a text editor. Change the title and building labels. Keep passwords and tokens out of this file.

Replace the illustration with your own SVG, PNG, JPG, or WebP in public/assets, then change illustration to its relative path. Use a simple filename such as my-house.webp. The local server serves image files only from that assets folder. Use artwork you own or have permission to distribute. The included SVG is generic and may be modified under MIT.

The collector can run on a computer that stays on. Its data folder must remain persistent across app updates. Keep your private adapter and credentials separate from anything served to browsers.

The included Dockerfile is an optional recorder-only recipe, not a one-click Synology installer. For an experienced Container Manager user:

docker build -t observatory-starter:1.0.0 .
docker run --rm --name observatory-recorder \
  --mount type=bind,src=/absolute/path/to/private-input,dst=/input,readonly \
  --mount type=bind,src=/absolute/path/to/persistent-data,dst=/app/data \
  observatory-starter:1.0.0

Supply /input/readings.json, and grant the container’s node user write access to the mounted data folder. Adapt host paths and permissions to your NAS. This container does not expose a web port. Container Manager configuration varies by model/version, and this recipe has not been run on a NAS as part of starter verification.

• Pick one question before adding more devices.
• Verify units against the device or another measurement you trust.
• Keep missing or stale readings unknown; never silently replace them with zero.
• Site consumption is not the same as grid import when solar or batteries are involved.
• A charger already measured by the site meter must not be added twice.
• A compressor’s frequency is not a verified electrical-power reading.
• Align source meter timestamps before calculating a building split.

No readings after importing: The file may contain nulls for that metric. Try another measurement or inspect the latest-sample table.

Recorder says “No newer samples”: Update the input file with a genuinely new measurement timestamp. The recorder intentionally preserves the old time.

Source changed: Archive data/history.json before changing the source. This prevents mixing demo and real datasets.

Port already in use: Stop another starter instance, or use another port. macOS/Linux: PORT=8781 node server.mjs; Windows PowerShell: $env:PORT=8781; node server.mjs.

A local-file import doesn’t work in your browser: Use the optional local server and try again.

Nothing updates while looking at the recorder: Click Load local recorder again. Automatic polling is not included in v1.0.0.

Run npm test with Node 22+ to exercise data validation, history persistence, chart-gap handling, and server boundaries. No package installation is needed. Packaging: npm run package.

See LICENSE and THIRD-PARTY-NOTICES.md. The code and generic illustration are MIT-licensed. The PRMGR name identifies the original project; use your own name for redistributed products rather than implying they are official PRMGR releases. This is a community starter with no support service or device-compatibility promise.