NIGHTSCOUT-ENGINE CGM SOVEREIGN INTEGRATION T1D FLEET V14 γ₁=14.134725141734693 HEALTHHIVE Day 111
CONNECTION STATUS CARDS
iAPS ↔ Nightscout LIVE
NightscoutRemoteCGM CGM data feed closed loop
iAPS (⭐236) uses NightscoutRemoteCGM as a CGM bridge dependency — allows remote CGM data to feed the closed-loop algorithm via Nightscout. 103 open issues = live clinical research agenda. This is the production stack running in thousands of T1D patients globally.
G7SensorKit ↔ Nightscout ACTIVE
Dexcom G7 BLE iAPS Nightscout
G7SensorKit provides direct Dexcom G7 integration via BLE. Calibration-free CGM — removes the human calibration sorry from earlier Dexcom generations (G4/G5 required finger-stick calibration every 12h). G7 factory-calibrated: CGM data is immediately reliable on sensor insertion.
LibreTransmitter ↔ Nightscout ACTIVE
FreeStyle Libre transmitter iAPS
LibreTransmitter enables FreeStyle Libre CGM integration into the iAPS loop — the Libre was originally scan-only (no real-time alerts), the transmitter converts it to continuous streaming. Covers patients who prefer Libre over Dexcom for cost/comfort reasons.
GluCoEnv ↔ Nightscout SORRY
Not standardized. GluCoEnv (⭐20) simulation environment uses its own internal glucose time-series format. Nightscout uses mg/dL or mmol/L JSON streams. No direct data format compatibility exists — RL policies trained in GluCoEnv cannot be validated against live Nightscout feeds without custom ETL. This is the sim-to-real gap operationalized.
PEMOS Fleet → Nightscout TARGET
Nightscout data PEMCLAU health corpus
Sovereign analytics integration target: Nightscout glucose data → PEMCLAU GraphRAG → health corpus V14 → sovereign analytics. 23M+ glucose readings in public Nightscout datasets. When ingested, this becomes the T1D node in the catomain organism and the Diabetes/T1D specialist node in HealthHive colony.
23M+
Public Glucose Readings
103
iAPS Open Issues
1000s
Patients on oref0
◉ NIGHTSCOUT DATA FLOW ARCHITECTURE
NP-NS-001
Nightscout = Sovereign CGM Hub
Every closed-loop system routes through it. 23M+ glucose readings in open datasets. The architecture is sovereign: open-source, self-hostable, no vendor lock-in. This is the only major medical data infrastructure that is genuinely open.
NP-NS-002
iAPS oref0 Algorithm — Open Reference Implementation of OpenAPS
This IS the algorithm running in thousands of real patients. Every open issue is a live clinical research agenda item. Issues = real sorrys affecting real glucose control right now. The gap between open issues and clinical reality is the research frontier.
NP-NS-003
G7SensorKit → Calibration-Free CGM
Removes the human calibration sorry from earlier Dexcom generations. Factory calibration means the sensor works immediately on insertion — no 2-hour warmup, no finger-stick requirement. This is a genuine patient safety improvement: miscalibration was a significant hypoglycemia risk factor.
NP-NS-004
The 103 iAPS Issues = Live Clinical Research Agenda
#1891 (micronutrients) = next frontier. Insulin algorithms model carbohydrates only. Fat and protein cause delayed glycemic responses (2–5h post-meal) that are invisible to the current algorithm. This isn't a software bug — it's a fundamental modeling gap. Solving it requires meal composition data, not just carb count.
NP-NS-005
RL4H G2P2C vs oref0 — The Sim-to-Real Gap
RL outperforms rule-based in simulation but not deployed at scale. G2P2C (PPO/SAC) beats oref0 in GluCoEnv metrics. But GluCoEnv patient diversity ≠ real-world patient diversity. The RL policy has never seen: pediatric patients, patients with neuropathy, exercise-induced hypoglycemia, illness states. The sim-to-real gap is the primary sorry preventing deployment.
◉ ENGINE STATUS · FLEET INTEGRATION
PEMCLAU ingestion path nightscout-engine.html → health corpus V14
Catomain organism node T1D node · pending glucose corpus ingest
HealthHive colony specialist Diabetes/T1D node · SPEC-02 + SPEC-03
oref0 algorithm status PRODUCTION · 1000s patients
GluCoEnv ↔ Nightscout format bridge SORRY · no standard bridge exists
iAPS #1891 micronutrient modeling OPEN SORRY · Phase 0 research
γ₁ heartbeat γ₁=14.134725141734693