CGM/CKM continuous glucose and ketone monitor integration

[elkimek]

Summary

Import and visualize continuous glucose monitor (CGM) and continuous ketone monitor (CKM) data alongside blood work for a complete metabolic picture.

Why This Matters

CGM and CKM data bridges the gap between quarterly lab snapshots and real-time metabolic health. Correlating glucose/ketone trends with lab biomarkers (HbA1c, insulin, HOMA-IR, triglycerides) gives a much deeper picture than either dataset alone.

CGM Landscape

Devices: Dexcom, Abbott FreeStyle Libre, Medtronic, Ultrahuman (Lingo), Levels, Stelo
Data points: Glucose readings every 1-15 minutes, 24/7

Open-source ecosystem (mature):

• Nightscout — the gold standard. Open-source CGM-in-the-cloud platform with 39k+ community. REST API for reading glucose data. Supports nearly all commercial CGM sensors
• Glucose360 — open-source Python platform for CGM data analysis with event-based integration
• AGATA — automated glucose data analysis toolbox
• cgmquantify — 28 clinically validated glucose metrics

CKM Landscape

Devices: SiBio KS1, Abbott (dual CGM/CKM in development), Readout Health
Data points: Ketone readings every 5 minutes via BLE

Status: Much earlier than CGM. No established open-source ecosystem yet. SiBio app supports data export. No public APIs documented. Clinical trials and early adopter market in 2025-2026.

Integration Approaches

1. Nightscout API — most practical starting point. Read CGM data from user's Nightscout instance via REST API. Massive existing user base
2. File import — accept CSV/JSON exports from CGM/CKM apps (Dexcom Clarity, LibreView, SiBio, etc.)
3. Direct device APIs — where available (Dexcom has a developer API, others are more restricted)

Visualization Ideas

• Glucose/ketone overlay on existing biomarker charts (e.g., show glucose curve alongside HbA1c trend)
• Daily glucose variability metrics (time in range, CV, GMI) as calculated markers
• Meal response correlation with diet context card data
• Ketone levels correlated with exercise context and fasting patterns
• Alert when CGM trends diverge from lab glucose/HbA1c (sensor accuracy check)

Open Questions

• Data volume: CGM produces ~288 readings/day vs quarterly lab snapshots. localStorage may not suffice — IndexedDB or data summarization needed
• Granularity: show raw readings, hourly averages, or daily summaries on charts?
• Nightscout auth: how to handle API tokens from a browser (CORS?)
• CKM data formats are not yet standardized

[elkimek]

Good fit with getbased's architecture — Nightscout is the natural starting point.

How it would work:

Nightscout's REST API supports CORS, so getbased can fetch CGM data directly from the browser — no proxy, no backend, same local-first model as everything else. User enters their Nightscout URL + API token in Settings (like AI provider setup).

The volume problem:

288 readings/day doesn't fit in localStorage alongside labs. Plan is to summarize on import — compute daily/weekly aggregates (mean glucose, time-in-range, coefficient of variation, GMI) and store those as calculated markers in the existing schema. Raw readings go into IndexedDB for optional chart overlays.

What shows up:

• New "CGM" category with calculated markers: Time in Range, GMI (estimated HbA1c), Glucose CV, Mean Glucose
• Optional raw glucose curve overlay on existing glucose/HbA1c charts (new chart plugin, like supplement bars or cycle bands)
• AI context gets CGM summary for correlation with lab data

Where the real value is:

CGM Time-in-Range vs lab HbA1c, glucose variability vs hs-CRP, meal response patterns vs triglycerides, fasting glucose trends vs HOMA-IR — getbased can correlate CGM with labs in ways neither tool does alone.

CKM: Too early — no APIs, no standard format. Start with Nightscout CGM + file import (Dexcom CSV, LibreView CSV), add CKM when the ecosystem matures.

Nightscout is open source, self-hosted, no vendor lock-in — same ethos. Good candidate for a future milestone.