AI Insights¶

ChillCheck includes an AI-powered analysis feature that reviews your cabinet's temperature performance and delivers practical, plain-English recommendations.

Cabinet analysis¶

On any cabinet detail page, an AI Insights panel appears below the temperature chart. Click Analyse to generate an on-demand analysis.

The analysis covers:

Performance summary — whether the cabinet is operating within acceptable limits
Key findings — specific temperature events and patterns over the last 14 days
Patterns & concerns — recurring issues, timing of breaches, response times
Recommendations — actionable next steps (threshold adjustments, maintenance checks, etc.)

Requirements¶

At least 2 weeks of sensor readings for this cabinet. The button is available once that window is met.
One analysis per cabinet per 24 hours — this prevents excessive cost and keeps results meaningful.
Available on Standard and Enterprise plans.

Weather context¶

If your site has a postcode set (Settings → Sites → Edit), ChillCheck uses local weather data from Open-Meteo to give the analysis ambient temperature context. This is particularly useful when diagnosing summer-peak issues or unusual readings after a hot spell.

To set your postcode: go to Dashboard → Sites, click Edit on the relevant site, enter the postcode, and tab out of the field — the location is automatically geocoded and saved.

Weekly summary email¶

Every Monday at 07:00 UTC, ChillCheck emails all owner and admin contacts in your organisation with an AI-generated summary of the previous week's temperature monitoring performance.

The email covers:

Overall performance across all cabinets
Alert counts and any unacknowledged incidents
Per-cabinet average temperatures and compliance rates
Forward-looking notes where patterns suggest attention is needed

No action is required to enable this — it runs automatically for all active paid accounts with at least one cabinet.

Predictive drift alerts¶

ChillCheck watches the temperature trend on each cabinet in real time and fires a Predicted breach alert when the current trajectory is projected to reach the critical threshold within the next 20 minutes.

This gives you a ~20-minute window to act — prop a door shut, restart a compressor, move stock — before the temperature actually enters the danger zone.

How it works¶

The hub fits a linear trend to the last ~20 readings (roughly 20 minutes of data) for each cabinet. If the trend's slope is significant (at least 0.05 °C/min) and the projected temperature will breach the critical band within 20 minutes, a Predicted breach warning alert is raised and an email is sent.

Unlike reactive temperature alerts, predictive drift alerts are email-only — they don't escalate to SMS or voice call, because by definition the cabinet is still within bounds and you have time to investigate.

Auto-resolution¶

The alert resolves automatically when:

The temperature trend reverses (slope flattens or heads in the opposite direction).
A real High temperature or Low temperature alert fires — at that point the prediction has come true and the reactive alert takes over.

Requirements¶

Sensor readings must have been flowing for at least 8 minutes at a consistent rate before a drift projection can be computed.
Available on all paid plans — no additional configuration needed.

Privacy & AI usage¶

Cabinet analysis data is sent to Anthropic (Claude API) for processing. The data includes temperature readings, alert history, threshold settings, and anonymised weather data. No personally identifiable information or customer names are included in the analysis request. Anthropic's enterprise data usage policies apply.

ChillCheck absorbs the cost of AI analysis — there is no additional charge for this feature.