Suggestion/idea: auto-tracking mode

[scottleibrand]

TL;DR: auto-detect (infer) when the user is/was outside rather than requiring them to remember to start/stop.

Possible data sources:

When I track an Outdoor Walk or Hike or similar outdoor activity on Apple Watch, I’m generally outside, and I’m otherwise usually inside or in the shade.

My precise geolocation, including velocity, generally indicates whether I’m indoors, in a vehicle, or doing outdoor activity. Not sure if there are any datasets indicating the average shade levels of different outdoor areas…

Less likely, but Apple exposes any useful APIs for tracking the ambient light levels used for setting auto-brightness on Apple Watch or iPhone?

[scottleibrand]

Initial o3 feasibility assessment:

https://chatgpt.com/share/6873f602-0340-8007-918d-e1ff22be7469

1 Status quo of the repo
• Current UI is “press sun button → start a background‐location session → poll Open-Meteo for live UV.”
HealthKit, Background location, and an App Group are already enabled in project.yml. 
• Vitamin-D algorithm is solid and self-contained; no changes needed for auto-tracking. 

2 Tier A – “Dumb” background polling (fastest to ship)

How
1. Keep the existing background-location session but schedule a BGAppRefreshTask every 30–60 min.
2. On each wake-up:

WeatherService.shared.fetchUV(for: mostRecentLocation) { uv in
    let delta = VitDModel.iuPerHour(uv) * exposureFraction
    HealthKit.saveUVSample(uv)          // HKQuantityTypeIdentifier.uvExposure
    HealthKit.appendVitaminD(delta)
}

3. Use CLLocationSourceInformation.isSimulatedBySoftware and horizontalAccuracy to filter out indoor GPS drifts.

Pros / Cons
• ✅ Zero new entitlements – fully App Store-legal today.
• ❌ Indoor / shade false positives. Many users leave the phone indoors.
• ❌ Battery: background GPS every 30 min is ~2–4 %/day.

Feasibility: trivial; you could merge in a weekend.

3 Tier B – Fuse Apple Watch signals (better context)

3.1 Use workouts & CLVisit
• Apple Watch already labels workouts “Outdoor Walk/Run/Cycle”. Those workouts include HKWorkoutRoute; you can iterate that polyline and replay UV retrospectively.
• Outside workouts, use CLVisit for coarse “arrived / departed” and treat visits whose horizontalAccuracy < 50 m as “outdoor sessions”. This costs ~0.5 %/day.

3.2 Foreground-only light heuristic

Ask the user to leave the Sunday complication on their watch face. The system wakes the widget every few minutes; you can seize that moment to pull current UV and accumulation with no extra background cost.

Pros
• Higher precision: if the watch records elevated heart-rate + GPS + wrist up, chances are the user really is outside.
• Still works with public entitlements only.

Cons
• Misses idle sunbathing (no workout).
• Requires WatchOS companion (split-target work).

Feasibility: 1–2 developer-weeks; moderate SwiftUI + HealthKit glue.

4 Tier C – Direct sensor readings (research / BLE)

4.1 SensorKit Ambient-Light
• SRSensor.ambientLightSensor gives lux, chromaticity, and light direction. 
• BUT: production apps need Research Device entitlement (NSSensorKitUsageDetail); Apple has so far granted it only to clinical-study IDs. 
• => Great for TestFlight pilots or an academic partner, mostly a dead-end for mass App Store release.*

4.2 External BLE UV patches (My Skin Track UV, QSun, etc.)
• Many advertise an HKQuantityTypeIdentifier.uvExposure write-capability. 
• You can piggy-back: listen for new HealthKit UV samples and map them to vitamin-D accumulation instead of calling Open-Meteo.

Pros
• True on-skin UV with clouds, shade, and reflection accounted for; most accurate.
• No special entitlements; accessory vendor shoulders FDA / CE burden.

Cons
• Niche hardware; user must buy/wear a clip-on.
• BLE reliability / background timeouts on iOS 17 require proper CBPeripheralManager handling.

Feasibility: 2–3 weeks for BLE + fallback logic.

…

6 Recommended roadmap
1. Ship Tier A behind a new “Auto-track” toggle in Settings; default off.
• Minimal code; gauges user interest.
2. If users want more, add Tier B: watchOS helper + CLVisit fusion.
3. Gate Tier C behind a “Connect UV sensor” screen; only reachable if BLE device detected.

That sequence keeps the current Sun Day philosophy—no API keys, no exotic entitlements—while giving a clear upgrade path for power users.

⸻

Quick win checklist
• Add HKQuantityTypeIdentifier.uvExposure write support. 
• Create BGAppRefreshTask with earliestBeginDate = now + 30 min.
• Persist cumulative IU in HealthKit (Vitamin D Amount custom quantity).
• On first launch of auto-track, pop up a battery-impact disclaimer (App Review friendly).
• Document privacy in PRIVACY.md (“Location pings every 30 min while enabled”).

⸻

Bottom line: You can have a respectable first-cut auto-tracking mode live on the App Store in < a week (Tier A). Accuracy leaps once you add Apple Watch context (Tier B). Direct UV sensors (Tier C) are technically sweet but gating on hardware availability and Apple entitlements.

[cdoering]

Isn't the Apple Watch already tracking that?

time spent in daylight

[scottleibrand]

Isn't the Apple Watch already tracking that?

time spent in daylight

Ah, great! Looks like HKQuantityTypeIdentifier.timeInDaylight's provides programmatic access to every sample, which should allow reconstruction of an exact timeline (and estimate Vitmain D production from UVI at that time).

Here's how o3 says it could be done:

Approach
1. Read the raw daylight records.

let type = HKObjectType.quantityType(
    forIdentifier: .timeInDaylight)!   // iOS 17+
healthStore.requestAuthorization(toShare: [], read: [type])

let pred = HKQuery.predicateForSamples(
    withStart: dayStart, end: dayEnd, options: [])
let q = HKSampleQuery(sampleType: type,
                      predicate: pred,
                      limit: HKObjectQueryNoLimit,
                      sortDescriptors: nil) { … }      //  [oai_citation:0‡Apple Developer](https://developer.apple.com/documentation/healthkit/hkquantitytypeidentifiertimeindaylight?utm_source=chatgpt.com) [oai_citation:1‡Apple Developer](https://developer.apple.com/documentation/healthkit/hksamplequery?utm_source=chatgpt.com)

Each HKQuantitySample you get back is a little “I was in daylight for n seconds” blob with startDate / endDate.

2.	Explode the samples into a timeline.

The samples are already granular (30 - 60 s), so simply iterate:

for s in samples {
    let stride = 60.0      // bucket to the minute
    var t = s.startDate
    while t < s.endDate {
        timeline[t.rounded(to: stride)] += min(stride,
            s.endDate.timeIntervalSince(t))
        t.addTimeInterval(stride)
    }
}

Result: [Date: secondsOutdoorsThisMinute].

3.	Attach a UV-index to every bucket.

WeatherKit’s hourly forecast has CurrentWeather.uvIndex plus 240 h of hourly UV .

struct UviCache { subscript(date: Date, loc: CLLocation) -> Double { … } }

– Fetch once per hour, keep in Core Data, and interpolate inside the hour.

4.	Convert UV → Vitamin-D IU.

Re-use the repo’s VitDModel.iuPerHour(uvi):

for (minute, sec) in timeline {
    let uvi = uviCache[minute, location]
    let iuPerH = VitDModel.iuPerHour(uvi)
    vitDTotal += iuPerH * (sec / 3600)        // IU for that minute-bucket
}

5.	Persist & show.

HealthKit write: HKQuantityTypeIdentifier.dietaryVitaminD (or your custom type) per day.
UI: a SwiftUI Chart that paints the minute buckets so the user sees exactly when they were outside, plus a running Vitamin-D total.

⸻

Implementation Notes

Item Detail
Query style HKSampleQuery is simplest; HKStatisticsCollectionQuery also works if you set intervalComponents = DateComponents(minute: 1) and ask for .cumulativeSum, giving you a ready-made minute series.
Location to feed WeatherKit Good: intersect with any simultaneous CLLocation or CLVisit sample.Fallback: last known coarse location; error is usually < ±1 UVI.
Weather history gap WeatherKit returns only forecast data. Cache each hourly call so you’ll have historical UV when you process yesterday’s daylight samples the next morning.
Performance 1-day roll-up means ~1.4 k minute buckets → trivial CPU; UV fetch is 24 JSON requests / day.
Battery Zero extra radio or sensor use—everything is passive.
Privacy / App Review Read-only on HealthKit; no new entitlements; fully App-Store safe.

⸻

Drop-in file sketch (TimelineVitDSynth.swift)

final class TimelineVitDSynth {
    func rebuild(for day: Date) async throws {
        let daylight = try await fetchDaylightSamples(day)
        let timeline  = minuteBuckets(from: daylight)
        let loc       = try await locationProvider.bestGuess(for: day)
        let uvis      = try await uviCache.ensure(for: day, at: loc)

        var iu = 0.0
        for (d, secs) in timeline {
            if let uvi = uvis[d.hour] {
                iu += VitDModel.iuPerHour(uvi) * secs / 3600
            }
        }
        try await healthWriter.saveVitaminD(iu, for: day)
    }
}

Put this behind a nightly BGProcessingTask and Sun Day will wake up, roll the previous day’s daylight samples into a to-the-minute Vitamin-D diary, and write it back to HealthKit—no manual input, no watch workout dependency, and no exotic entitlements needed.

[Bobcat5360]

(#accept?)