GoldLink+ data is wasted as a status light

Why they pay

A graph of ΔP is already in the Camfil portal. They pay for jobs a clipboard cannot do — and they will pay industrial money for them:

1. Tell the night shift what to do. Three sentences. Empty the drum. Order these twelve cones. Listen at the air header. Not twelve charts.
2. Write the filter PO from the curve. Right SKU, right qty, changeout date from slope — not from last year’s calendar. That is how the fee generates Camfil dollars.
3. Stop wasting media and air. Early changeouts and leaking pulse valves. Donaldson cites collector downtime at thousands of dollars an hour. One avoided stop pays for years of Office.
4. Hand EHS a file. 14-day CSV / 90-day pack. Who watched this housing on Sunday. Not a DHA — the log the DHA asked for.

What comes off GoldLink+

Published facts: a small box with a 4–20 mA differential-pressure sensor and two pressure ports, optional extra sensors, discrete inputs, weatherproof enclosure, internal antenna, worldwide LTE/3G, data encrypted end-to-end, not through plant Ethernet, watched in Jonesboro, viewed on Camfil’s dashboard. Optional stack: compressed air, hopper/drum, outlet emissions, airstream T/RH, pulse count, airlock zero-speed. GoldLink Connect (EMEA) also reads fan kW and rpm off the inverter.

What they have not published: a public API, a sampling rate, or a file format. There is no official JSON or CSV spec. The plant app assumes we get the stream. Until Jonesboro opens it, this is a proposed contract — not a Camfil document.

• On the wire: analog 4–20 mA (live-zero) and dry-contact discretes. 4 mA = 0 in. w.c. (or 0% hopper). 20 mA = span (typically 10 in. w.c. on ΔP).
• In the radio: TLS over LTE. Likely a proprietary frame today. We ask for HTTPS POST of JSON every 60 s, immediate on alarm, 5 min heartbeat. MQTT is a fine alternate with the same body.
• In the app: JSON telemetry + alarm events. Customer export is CSV (and a PDF pack later). Sample payload is on Feed.

When this is real — not now

This preview is the product spec. The ship stack is Rust + Postgres + Svelte. Do not rebuild until Jonesboro will actually open the stream.

• Rust ingest. One binary. Device JWT on HTTPS POST (MQTT if they already have a broker). 60 s frames, alarm-immediate. Cheap to run, boring under load.
• Postgres + Timescale. Raw frames 90 days. Hourly two years. Daily forever (the EHS file). Insights (remaining life, air waste) computed on rollups, not on a million rows.
• Svelte, not Flutter — unless the floor is offline-only. One web app: phone for the night shift, wide for the plant office, same API for Camfil service. Flutter only if they refuse a browser and need a store-built offline client.
• Two seats on one history. Plant floor = do this. Plant office = all housings, media calendar, compliance pack. Camfil office = which accounts are dark, who is about to order, who is leaking air. Same tables. Different roles.