Files

znetsixe 21d77a8afa feat(coresync): FROST/Influx/Grafana demo + sub-tick burst-window reducer fix

Lands the end-to-end CoreSync demo (frost-influx-grafana.flow.json) along
with the supporting normalizer/identity/interpolation/reducer work, plus
a targeted fix for the under-compression bug surfaced by the FROST demo.

Under-compression fix (PointReducer):
- New burstWindowMs option (default 0, opt-in). When two samples arrive
  within burstWindowMs of each other, the second is treated as the same
  wall-clock observation: previous is replaced, slope analysis is skipped.
- Without this guard, sub-millisecond bursts (rotatingMachine emits twice
  per pressure-injection cycle, ~1 ms apart) produced near-vertical
  apparent slopes that tripped angle-change on every tick — driving
  cog/efficiency/SEC streams to ~0.6% reduction (i.e. no compression).
- With burstWindowMs: 10 in the demo flow, the same streams now compress
  at 78-93% (verified end-to-end in InfluxDB over a 3-min window).
- Editor HTML exposes the new "Burst dt" field with explanatory tooltip.

Regression test (test/basic/coresync.basic.test.js):
- New "burstWindowMs collapses sub-tick sample bursts into a single
  observation" test reproduces the exact burst pattern from the demo
  and asserts before/after behaviour.
- Existing 14 tests continue to pass.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

2026-05-22 20:27:28 +02:00

1.5 KiB

Raw Blame History

CoreSync Examples

FROST + Influx + Grafana

Import frost-influx-grafana.flow.json into Node-RED when the local Docker stack is running.

The flow demonstrates:

a measurement node producing a 1 Hz directional flow signal;
a rotatingMachine node producing event-driven pump telemetry;
raw EVOLV telemetry written to local InfluxDB;
CoreSync reducing the same telemetry to knots;
CoreSync FROST HTTP requests sent to https://sta.wbd-rd.nl/FROST-Server/v1.1;
FROST responses looped back into CoreSync as msg.topic = "frost.response";
CoreSync knots also written to local InfluxDB as coresync_knots for Grafana comparison.
a Node-RED dashboard page with live raw-field, knot, FROST observation, and percent-saved counters.

Do not store FROST credentials in the flow. Start Node-RED with:

FROST_USER=write FROST_PASSWORD='<password>' docker compose up -d --build

The local InfluxDB and Grafana settings match docker-compose.yml:

InfluxDB: http://localhost:8086, org evolv, bucket telemetry, token evolv-dev-token
Grafana: http://localhost:3000, user admin, password evolv

The provisioned Grafana dashboard CoreSync FROST Demo compares raw telemetry against the reduced knot stream. The raw samples are intentionally denser than the knot series; the knot series is the data that should be sent to FROST.

Runtime dashboards:

Node-RED live counters: http://localhost:1880/dashboard/coresync-frost
Grafana raw-vs-knot comparison: http://localhost:3000/d/coresync-frost-demo/coresync-frost-demo

1.5 KiB Raw Blame History

CoreSync Examples

FROST + Influx + Grafana

1.5 KiB

Raw Blame History