SCHEMA.md

Forge Industrial Intelligence Pack — Schema

One row = one complete operational scenario lifecycle. All records share the same fourteen top-level fields.

Schema version: 1.0.0-forge-industrial-sample

Top-level fields

schema_version — string

Schema identifier. Constant within a sample release.

event — struct

Identifier fields and the overall status/severity for the lifecycle.

Field	Type	Notes
id	string	Stable event identifier, e.g., FORGE-100000.
trace_id	string (UUID)	Cross-links telemetry within the lifecycle.
timestamp	string (ISO-8601)	Scenario anchor time.
scenario	string	One of: port_spillover_surge, power_constrained_ev_yard, labor_gap_service_degradation, cold_chain_excursion_risk, tenant_hypergrowth_overrun, last_mile_cutoff_compression, site_selection_power_arbitrage.
severity	string	medium, high, critical.
status	string	decision_pending, triaged, confirmed.
confidence	double	0–1 confidence of the event label.

organization — struct

Market-level context for the facility/portfolio.

Field	Type	Notes
sector	string	Always industrial_real_estate.
market	string	Generic submarket archetype (e.g., Dallas-Fort Worth, Los Angeles, Chicago, Atlanta, New Jersey, Phoenix). Used as a type-label, not a specific-property reference.
submarket	string	Submarket handle (e.g., Inland Port, Inland Empire, Joliet, West Valley).
region	string	Regional bucket (south_central_us, west_coast_us, etc.).
environment	string	Deployment bucket (e.g., production_portfolio).
port_proximity_km	int	Proximity to major port, km.
airport_proximity_km	int	Proximity to major airport, km.
highway_access_score	double	0–1.
labor_tightness	double	0–1.
grid_headroom_mw	double	Available grid capacity (MW).
industrial_vacancy_pct	double	Submarket vacancy rate (percent).
base_rent_psf_yr	double	Submarket base rent ($/sqft/yr).
population_density	int	People per sq km.
development_cycle_days	int	Typical time-to-delivery for spec construction.

identity_context — struct

Decision owner and authorization lineage.

Field	Type	Notes
principal_id	string	Synthetic identity handle (e.g., ops://tnt-XXX/node/fac-YYY).
decision_lineage	list	Ordered role / action / authority_level entries showing the approval chain.
auth_method	string	How the decision was authenticated (e.g., tower_dashboard).
is_automated_recommendation	bool	Whether the decision started as an automated rec.
stakeholder_latency_hours	int	Hours of stakeholder lag.
meeting_load	string	low / moderate / high.

vulnerability — struct

Risk taxonomy and severity vector.

Field	Type	Notes
class	string	Scenario-class label (same as simulation.ground_truth_label).
risk_taxonomy	list	Ordered taxonomy codes (e.g., LOG-LOC-001, LOG-DEM-004).
exposure	string	Primary exposure dimension (e.g., demand_pressure, service_risk).
severity_model.base_score	double	0–10 severity base.
severity_model.vector	string	Vector string encoding severity dimensions (e.g., LOG:3.1/EX:O/FR:C/...).

tenant_context — struct

Synthetic tenant profile tied to this scenario.

Field	Type	Notes
tenant_id	string	Synthetic tenant identifier (e.g., TNT-189FCA4461).
industry	string	ecommerce, 3pl, retail_distribution, cold_chain, industrial_manufacturing.
truck_profile	string	parcel_heavy, mixed_freight, palletized, reefer, inbound_component.
order_growth_rate	double	Growth rate delta.
sqft_footprint	int	Facility sqft.
expansion_open	bool	Whether the tenant is open to expanding.
retention_priority	string	standard, high, strategic.
risk_of_churn	double	0–1 churn risk.

facility_context — struct

Building-level attributes: capacity, energy, clear heights, dock counts, refrigeration, automation level.

portfolio_context — struct

Network-level dynamics: submarket share, adjacent-node capacity, cross-site elasticity.

telemetry_stream — list

Variable-length ordered telemetry readings representing the evolving system state across the scenario. Each element includes an event-name label matching one item in the scenario's event sequence, plus relevant signal readings (dock utilization, queue length, power load, service commitment risk, truck turn time, temperature delta, etc.).

detection — struct

Anomaly-signature metadata.

Field	Type	Notes
analytic_family	string	Analytic family label (e.g., industrial_real_estate_behavioral_intelligence).
primary_risk_class	string	Primary risk dimension.
rule_logic	string	SQL-like rule expression for the detector.
baseline_deviation	string	Short English description of the deviation pattern.
anomaly_score	double	0–1.
confidence_band	string	low / medium / high / very_high.
forecast_pressure_delta	double	Delta vs baseline forecast.
signal_conflicts	list	Conflicting signal labels, if any.

forecast — struct

Forward-looking predictions tied to this scenario: demand shift, occupancy trajectory, pricing outlook.

impact — struct

Economic consequences: revenue delta, NOI impact, capex required, churn risk delta.

response — struct

Recommended actions and execution context.

Field	Type	Notes
recommended_actions	list	Top actions ranked by score.
primary_action	string	Selected action (e.g., shift_overflow_to_satellite).
primary_action_score	double	Scoring metric for the primary action.
primary_action_reason	string	Short reason code.
alternative_actions	list	action / score / reason triples.
decision_owner	string	Role who owns the decision (e.g., operations_lead, market_officer, development_committee).
execution_window_days	int	Execution SLA.
playbook_id	string	Synthetic playbook identifier, prefixed FORGE-PB-.
capex_gate_required	bool	Whether capex approval is required.
stakeholders	list	Additional stakeholder roles.
expected_operational_outcome	string	partial_success, success, mitigation, etc.
recommended_tradeoff	string	Short tradeoff description (e.g., preserve_capital_with_higher_latency_risk).
execution_risk_band	string	low, moderate, elevated, high.

simulation — struct

Simulation engine provenance and ground-truth labels.

Field	Type	Notes
synthetic	bool	Always true.
engine	string	Simulation engine label (forge_industrial_sim_v1).
causal_coherence	string	Coherence mode descriptor (e.g., rich_schema_with_balanced_policy).
friction_profile	struct	Numeric friction dimensions: budget_pressure, labor_shortage, power_stress, demand_pressure, service_risk, portfolio_tightness, capital_availability, entitlement_friction, weather_disruption, upside_mode (bool).
ground_truth_label	string	Scenario class (e.g., Network_Optimization, Labor_Constraint, Temperature_Control_Risk).
intended_use	list	ML use-case tags (e.g., forecasting, site_selection, tenant_expansion_decisions).

Distribution of this sample

• 10,000 lifecycles total.
• Severity: balanced across medium / high / critical (~3,300 each).
• Status: balanced across decision_pending / triaged / confirmed (~3,300 each).
• Scenario: balanced across 7 scenario classes (~1,400 each).
• Market: balanced across 6 submarket archetypes (~1,600 each).
• Industry: balanced across 5 tenant archetypes (~2,000 each).

Sanitization notes

• Internal identifier prefix (PLG-REAL-*) has been normalized to FORGE-*.
• Internal playbook prefix (PLG-PB-*) has been normalized to FORGE-PB-*.
• Internal engine label (SIMA-inspired industrial reality generator) has been normalized to forge_industrial_sim_v1.
• Market names (Dallas-Fort Worth, LA, Chicago, Atlanta, NJ, Phoenix) are used as submarket archetypes for industrial-real-estate type labels, not specific property or operator references.
• No real facility telemetry, real tenant identities, real portfolio NOI, or identifiable stakeholder data are present.

Relationship to the full pack

The production pack scales to 5M+ lifecycles with expanded market and submarket coverage, richer per-step telemetry, additional scenario classes (ESG signals, geopolitical supply disruption, labor automation events, bonded warehouse flows), and multi-year longitudinal traces per tenant. See the pack card for commercial access.