Sense
SPS30, Plantower, and BMV080 measure PM1, PM2.5, and PM10.
Bair1Bair1 documentation
A physical sensor, a live public feed, and AI that stays grounded.
Bair1 makes indoor air observable. Real kitchen sensors stream particulate readings to a public dashboard, then GPT-5.6-terra explains what those measurements mean in context.
Live prototypePM1 · PM2.5 · PM10
System flow
From particles to a useful answer.
Store
The device stream becomes a time-series history in DynamoDB.
Compare
The public feed combines indoor data with LAQN, weather, and pollen.
Explain
GPT-5.6-terra turns the current evidence into plain English.
Grounded AI
GPT-5.6 sees the evidence, not just the question.
The public Air Insight and Data Studio run server-side. They receive a fresh snapshot of the feed and return useful language that cites the actual PM values, history, forecast, and local context.
Insight context
Latest readings from every feed sensor
30-minute history and trend forecast
LAQN PM2.5 and PM10 comparison
Weather and pollen where available
Strict JSON insight → headline, explanation, concrete advice, confidence.
The insight route caches each feed for approximately two minutes. If a source is unavailable, the numbers-only dashboard continues to work.
Explore the stack
Bair1 / Live air
Live air
30-minute history, a browser-side trend forecast, current readings, and an Air Insight grounded in the numbers.
Explore →Bair1 / London mapLondon map
Indoor readings beside the London Air Quality Network, so a kitchen measurement has outdoor context.
Explore →Bair1 / Data StudioData Studio
Ask what changed, compare the forecast, and inspect the answer against the data behind it.
Explore →Bair1 / Developer platformDeveloper platform
REST, GraphQL, CLI, and MCP tools make Bair1 readings available to products and agents.
Explore →