Haystack modeling

Open-FDD uses Project Haystack semantics for sites, equipment, and points. Driver raw IDs map to Haystack refs through the assignment graph. Analytics and FDD rules consume mapped SQL roles after package ingest — not vendor point names.

A ZIP that imports successfully is not necessarily commissioning-grade. Empty Overview / Inspect / health matrices almost always mean missing roles in the package, not a broken engine. See Package authoring.

Guides

Guide Content
Package schema Compact ingest maps vs rich (SCAFFOLD) evidence
Heat-pump buildings WSHP topology, role tiers, anti-patterns
Rule readiness Runnable / missing / not applicable
Haystack model Sites, equipment, points, RDF APIs
Assignments Bind drivers → Haystack → FDD

Agent entry points (repo)

  • Law: AGENTS.md
  • Authoring: docs/agent/PACKAGE_AUTHORING.md
  • Role aliases: docs/migration/vibe19/ROLE_MAPPING_PARITY.md
  • Ingest (read source): edge/src/csv_ingest/package.rs
  • SQL roles: crates/fdd_core/src/columns.rs
  • Equipment typing: edge/src/equipment_types.rs
  • Rules: sql_rules/registry.yaml

BUILDING_100-style AHU/VAV archives are richness references only — do not copy that topology onto water-source heat-pump buildings.


Table of contents