Analytics field validation (smoke checklist)

This is software smoke validation for the Analytics hub — not hardware bench testing. For BOM/sensor bench status see Validation status.

Last updated: with Ask, learned watches, and vibration.

---

Prerequisites

1. Backend migrations applied (npm run migrate in backend — through 0044_analytics_watches.sql).
2. Logged-in tenant user with at least one device type that has telemetry.
3. Optional: an asset tree with devices assigned (for rollup).
4. Optional: profile calculated fields enabled (CALCULATED_FIELDS_ENABLED default true).

---

Metric Explorer

Step	Expected
Open Analytics → Metric Explorer	Page loads; device types and keys populate
Select type + key + range → Run / auto-load	KPI, Top-N, time series fill when data exists
Set Above/Below thresholds	Anomaly count updates
Change bucket (5m / 1h / 1d)	Series re-buckets

Fleet risk / exceedance

Step	Expected
Set Above/Below + run	Fleet risk table ranks by crossings, then samples over, then time over limit
First page	Top risky devices for the type (≈ top 10 at page size 10)
No thresholds	Warning — do not treat as failure risk
API	POST /api/analytics/fleet-forecast-risk

Profile: Analytics industry profile.

Forecast

Step	Expected
Choose Linear or Predictive	Alert shows that mode’s device / point limits; device count cannot exceed max
Load explorer with enough points	Forecast chart + method tag (linear_regression, seasonal_naive, or statsforecast:*)
Predictive + worker up	Method like statsforecast:auto_ets when FORECAST_WORKER_URL is set
Predictive with worker down / sparse data	Falls back to seasonal-naive or linear (fallback)
Pattern heatmap	Hour × weekday card fills from series sample
Request more devices than cap	Server truncates; limits.truncated true
API	GET /api/analytics/forecast-limits · POST /api/analytics/forecast with engine

Statistical anomalies (z-score)

Step	Expected
Load explorer on a key with variation	Table lists devices with latest z and Anomalous/OK
Flat series (stdDev ≈ 0)	z ≈ 0, not flagged
API	POST /api/analytics/statistical-anomaly with zThreshold (default 2.5)

Asset rollup

Step	Expected
Pick asset + key → Run rollup	Device count + agg / min / max
Asset with no devices	Zero devices / empty values
API	POST /api/analytics/asset-rollup

Shared saved views

Step	Expected
Save as Private	Visible only to owner
Save as Whole tenant	Other tenant users see it in Saved views
Save as Shared emails + comma list	Only listed emails (+ owner) see it
Open from Saved views	Explorer filters restore; visibility column shows mode
Migrations	0039 + 0040 applied (migrate:status Pending: 0)

Ask

Step	Expected
Open Analytics → Ask	Suggestion buttons name fleets and readings this tenant already has
Press a machine or meter suggestion	One sentence and a bar chart for that device type and key
Ask with no fleet and no reading	The answer lists fleets and one reading each
“When will voltage on a named device reach a number?”	A line and a sentence. A flat series says it is not moving toward the limit
“Analyze vibration” with no vibration key	The answer says no vibration series is present
Watch sentence for one named device	A row appears under Watches. Remove deletes it. No watch is created for other devices

APIs: GET /api/analytics/prompts · POST /api/analytics/chat · GET/POST /api/analytics/watches · DELETE /api/analytics/watches/:id.

Calculated fields (ingest)

Step	Expected
Profile has calculated_fields expression	Derived key appears on next telemetry ingest / bulk write
CALCULATED_FIELDS_ENABLED=false	Expressions skipped

Alarm ops

Step	Expected
Analytics → Alarm ops	MTTA/MTTC, severity, noisy rules/devices for range

---

Quick API smoke (authenticated)

POST /api/analytics/forecast
POST /api/analytics/statistical-anomaly
POST /api/analytics/asset-rollup
GET  /api/analytics/saved-views
POST /api/analytics/saved-views

Bodies match Metric Explorer selections (deviceIds/key/from/to, or assetId for rollup; visibility + sharedWithEmails for saves).