openfdd-sql-fdd

name	openfdd-sql-fdd
description	Use when working on production DataFusion SQL FDD: sql_rules registry, central /api/fdd/run, parity with pandas oracle, no pandas in central. Triggers on: sql_rules, DataFusion, registry.yaml, FDD engine, SQL cookbook.

Production SQL FDD

• Registry: sql_rules/registry.yaml
• Engine: openfdd-central DataFusion
• Docs: docs/rules/cookbook/datafusion-sql-cookbook.md
• Architecture: docs/architecture/datafusion-first.md

Hard rules

1. Never add silent pandas fallback in central.
2. New production rules need SQL file + registry + cookbook + parity row.
3. Pandas oracle may differ — document in parity/gap matrices; do not paper over.
4. UI production path calls central REST — not local pandas runner for “real” FDD.
5. Series overlay (/api/fdd/.../series) must honor Lab/session_config confirm_min and typed params the same way /api/fdd/run does (prefer session-tuned detail over a stale empty results index when overrides exist).
6. SCHED-1: portable occupancy — numeric/false unoccupied and string unoccupied tokens; keep sql_rules/sched1_unoccupied_runtime.sql aligned with pandas sched1 in open_fdd/rules/cookbook_catalog.py.