Repository navigation
[Vision] OpenGlucose metabolic context cockpit #49
Description
Product vision
OpenGlucose should become a local-first metabolic context cockpit that helps a person answer:
What happened in my glucose, what was happening around it, and does this pattern repeat?
The product remains glucose-first and wellness-only. Meals, sleep, workouts, movement, heart rate, and notes add context around a source-attributed glucose timeline. Deterministic analytics produce the facts and evidence. Optional AI may summarize those facts, but it never invents calculations, diagnoses, treatment, dosing, emergency guidance, or autonomous actions.
Primary experience
┌──────────────────────────────────────┐
│ 108 ↗ mg/dL fresh 2m · Live │
│ Sensor: 14 days left │
├──────────────────────────────────────┤
│ 24-hour body timeline │
│ glucose line │
│ █████ sleep ███ ● meal ▬ workout │
│ small activity / HR rail │
├──────────────────────────────────────┤
│ Sleep 7h42 │ 6.8k steps │ Meal +22 │
├──────────────────────────────────────┤
│ What stands out │
│ “Two post-meal walks coincided with │
│ steadier glucose afterward.” │
│ 2 qualified days Why? → │
├──────────────────────────────────────┤
│ Stability │ Meals │ Overnight │ Move │
└──────────────────────────────────────┘
+ Meal · Activity · Note
Navigation remains compact:
- Today — live glucose, one body timeline, context chips, one observation, quick add.
- Timeline — journal and chronological source-attributed context.
- Trends — qualified comparisons, recap, and long-term profile.
- Settings — sensor, display, integrations, AI/models, privacy/data, app.
Product principles
- Sensor truth before interpretation.
- Glucose-first, context-rich.
- Deterministic analytics before AI narration.
- Every observation opens its evidence, window, source, counts, and gaps.
- Missing or inaccessible data is explicit; it never becomes zero.
- Associations are not causality.
- Local/offline value is complete; every external boundary is opt-in and revocable.
- No opaque metabolism score or bio-age.
- AI is optional and never controls alerts, records, sensors, or health decisions.
- Compact daily story over dashboard-card accumulation.
Current release track
These protect the existing product and proceed alongside the new feature program:
- Fix sensor warmup, scan recovery, and live glucose updates #35 — warmup/history/live-notification/Bluetooth recovery PR
- Release readiness: Android compatibility, stable 0.1, and external TestFlight #34 — Android hardware compatibility and stable 0.1 readiness
- Automate protected TestFlight beta releases #33 — protected TestFlight automation
- Reconcile the checked-in 43-task backlog as tracked in Release readiness: Android compatibility, stable 0.1, and external TestFlight #34
Parallel delivery map
Reliability: #35 + #34 + #33 ───────────────→ #45 → #47
Data spine: #36 ─┬→ #40
├→ #41 ───────────┐
├→ #43 ───────────┤
└→ #42 ───────────┤
Design: #38 ────────────────────┤→ #44 ─┬→ #46
Analytics: #37 ────────────────────┘ └→ #48
AI contract: #39 ───────────────────────────→ #46
Wave 0 — ready now
These have separate ownership and can begin concurrently:
- [Feature] Source-aware health data spine and import identity #36 — source-aware data spine and import identity
- [Feature] Compact Today navigation and metabolic design system #38 — compact Today navigation and design system
- [Feature] Deterministic metabolic observations and evidence model #37 — deterministic observations and evidence model
- [Feature] AI evidence, safety, and provider contract hardening #39 — AI evidence/safety/provider contract
Wave 1 — after the #36 contract freezes
These are separate feature lanes and can run in parallel:
- [Feature] Complete health-data export, deletion, retention, and recovery #40 — complete export, deletion, retention, and recovery
- [Feature] Fast journal and meal/activity logging #41 — fast journal and meal/activity logging
- [Feature] Apple Health context import #43 — Apple Health import
- [Feature] Android Health Connect context import #42 — Android Health Connect import
Wave 2 — integrated product surfaces
- [Feature] Unified body timeline and compact Today experience #44 — unified body timeline and compact Today
- [Feature] Wellness alerts, freshness monitoring, and alert history #45 — wellness alerts and freshness history, after Fix sensor warmup, scan recovery, and live glucose updates #35
- [Feature] Evidence-linked AI observations and qualified local providers #46 — evidence-linked AI observations and qualified local providers
Wave 3 — daily-value expansion
- [Feature] Home widgets and glanceable background surfaces #47 — home widgets and glanceable background surfaces
- [Feature] Long-term metabolic profile and portable reports #48 — long-term metabolic profile and portable reports
Milestones
Shared Definition of Done
Every feature must:
- preserve live/cached/sample/imported provenance;
- pass sparse, duplicate, future, out-of-order, DST, gap, and permission-boundary fixtures where applicable;
- provide migration, recovery, rollback, and deletion behavior;
- avoid sensitive identifiers/readings in logs and public artifacts;
- remain functional offline and with AI disabled;
- render at 375/390/430 points, 200% text, screen reader, dark mode, and reduced motion;
- include independent review for R2 health/data/native/AI/accessibility work;
- at 8000 tach physical-device evidence where background, BLE, HealthKit/Health Connect, notifications, models, or widgets are affected;
- keep wellness/reference framing and prohibit diagnosis, treatment, dosing, emergency reliance, and unsupported causality.
Success measures
- Warmup-to-first-trusted-reading completion rate.
- Median basic meal logging time under 10 seconds.
- Health import duplicate rate: zero.
- Qualified days with at least one meal/sleep/activity context source.
- Observations with working evidence drill-down: 100%.
- Background freshness/stale-state accuracy.
- Weekly recap eligibility and return rate.
- Complete export/delete/restore success rate.
- Accessibility completion across supported devices and text scales.
Explicit non-goals for this program
Defer until the evidence loop is trustworthy:
- generic blank AI chat;
- food database and automatic nutrition truth claims;
- opaque metabolism scores or bio-age;
- coaching/telehealth;
- community or follower cloud services;
- gamification without qualified analytics;
- medical alarms or treatment/dosing behavior.
Documentation sources
The detailed implementation direction and safety constraints already live in:
docs/product-foundations-2026.mddocs/product-safety.mddocs/architecture/adr/0002-local-first-health-data.mddocs/engineering/standards.md
This issue is the product-level source of truth. Child issues own feature scope, acceptance criteria, dependencies, and delivery evidence.
Activity
shroominic commentedon Aug 16, 2026
Delivery ledger
Timestamp:
2026-08-16T11:28:08Zpr58_stale_privacy_repairfeature/apple-watchpr_auditfeature/macos-desktop-previewCHANGELOG.mdand compatibility docsCEOfeature/windows-desktopCEOfeature/metabolic-context-trainpr_auditchore/testflight-release-actionissue36_execution_planThis comment is the live delivery ledger for the epic. It is updated after material state transitions; focused PRs reference their issues and an issue closes only after all acceptance criteria have independently verified evidence.
This ledger moved to #102 during the repository-owner issue migration.
shroominic commentedon Sep 1, 2026
Superseded by #102, recreated by the repository owner under
shroominic.