Open-FDD integrity sweep

The integrity sweep is a recurring, operator-style check that verifies Open-FDD platform health, graph integrity, and representative telemetry trustworthiness.

Cadence

  • Daytime recurring sweep target: every 20 minutes.
  • Overnight window (18:00-06:00 local): reduce duplicate low-signal chatter and prioritize richer overnight review artifacts.

Source-of-truth order

  1. Live backend data model (/data-model/check, /data-model/sparql, /data-model/export)
  2. Current launch/auth/runtime context
  3. Repo docs and playbooks

Sweep checklist

  1. Verify authenticated backend reachability.
  2. Validate data model integrity and SPARQL query path.
  3. Discover representative points from the current model (not hard-coded point lists).
  4. Validate representative BACnet reads for model-derived devices/points.
  5. Classify failures cleanly.

Failure classes

  • auth_config_drift
  • graph_model_drift
  • bacnet_device_state_drift
  • testbench_limitation
  • likely_openfdd_product_behavior
  • likely_ui_api_parity_bug

Mode detection

  • TEST_BENCH: fake device semantics and deterministic schedule artifacts present.
  • LIVE_HVAC: occupied-building semantics and live operational telemetry.

The sweep should always report mode + basis, then reason accordingly.