Open Glucose Telemetry

Runtime ecosystem

Open Glucose Telemetry (OGT) is the runtime layer: adapters, validation, normalization, and routing that turn vendor-specific glucose payloads into OGIS-compliant events. The homepage overview shows where OGT sits in the full OpenGlucose stack; this page goes inside the collector pipeline.

Collector pipeline

OGT pipeline: vendor sources through ingestion envelope, validation, adapter map, normalization, schema check, to canonical JSON or structured error. — Every reading enters as an ingestion envelope, passes adapter-specific validation and mapping, and exits as a canonical `glucose.reading` event or a structured pipeline error with a stable code.

OGT defines how glucose data moves. The Standards overview defines what canonical events mean once they leave the pipeline.

Adapter registry (shipped)

Adapters register by stable lowercase source ids on the ingestion envelope. The collector resolves adapters from a registry—no per-vendor branching in core pipeline code.

`source`	Upstream	Runtimes
`healthkit`	Apple Health / HealthKit CGM samples	TypeScript, Swift
`dexcom`	Dexcom EGV-style cloud readings	TypeScript, Swift
`abbott.librelinkup`	LibreLinkUp graph readings (direct cloud)	Swift (TypeScript parity planned)
`manual`	In-app and watch manual entry	Swift
`mock`	Test fixtures and CI golden paths	TypeScript, Swift

OGIS provenance.source_system values (for example com.apple.health, dexcom, abbott.librelinkup) are set during adapter mapping. See Provenance and Example JSON.

Pipeline stages

Ingress — wire JSON becomes an ingestion envelope with source, payload, trace_id, and received_at.
Envelope + payload validation — wrapper schema, then per-source payload JSON Schema.
Adapter map — vendor fields map to pre-normalize canonical fields.
Normalize — UTC timestamps, glucose values in mg/dL, bounded strings.
Semantic rules — plausible range and clock-skew checks on observed_at.
Dedupe (optional) — subject, observed time, and raw event id.
OGIS schema check — pinned glucose.reading v0.1 before emit.

Failures return structured errors such as ADAPTER_UNKNOWN, MAPPING_FAILED, and CANONICAL_SCHEMA_INVALID—stable for logging, UI, and CI.

Reference runtimes

TypeScript

Reference pipeline in OpenGlucoseTelemetry/runtimes/typescript. Collector engine, ingest plugins, and golden fixtures for cross-runtime parity.

Swift

On-device parity via OpenGlucoseTelemetryRuntime in OpenGlucoseTelemetry/runtimes/swift. Used by GlucoseAITracker and other embedded consumers.

OGT ships as an embeddable library today—apps call the pipeline in-process. Hosted buses, query APIs, and export connectors are roadmap items, not current repo scope.

Downstream outputs

Canonical JSON — glucose.reading v0.1 events validated against the pinned OGIS schema in OpenGlucoseTelemetry/spec/pinned/.
Golden fixtures — ingestion-to-canonical pairs under examples/canonical/ for TypeScript ↔ Swift parity tests.
Structured errors — stable codes for every pipeline failure path.

Consumer apps, AI pipelines, and research tools read the canonical JSON directly. For the specification side, see Standards overview.

Planned adapters

Additional adapters appear here and on Projects only when repositories ship—no dead-end links.

healthconnect — Google Health Connect (Android), see GlucoseAITracker Android AND-017
Additional CGM and BGM vendor adapters
FHIR Observation export (informative mapping — FHIR Mapping)

Public roadmap →

Repositories

OpenGlucoseTelemetry — adapters, collector pipeline, TypeScript and Swift runtimes
OpenGlucoseInteroperabilityStandard — normative schemas and validated examples OGT pins and validates against
GlucoseAITracker — downstream consumer via OGT (not an OpenGlucose standards repo)