Technical reference

The platform's object model, automation levels, glossary and known limits.

Object model

The main objects exposed in the console, reports and integrations. Field lists are conceptual; exact schemas are provided with your deployment.

Object	Key fields	Notes
Entity	id, role, domain, location, owner	Any node in the dependency graph
Dependency	from, to, kind, redundancy group	Physical, logical or service dependency
Signal	entity, type, time, value, risk score	Normalized alarm, metric or event
Contract	entity, window, deltas, related entities, evidence links	Compact representation used for reasoning
Incident	origin, members, timeline, impact, status	Signals grouped by common cause
Hypothesis	incident, origin, chain, confidence, evidence	Ranked explanation
Action	type, targets, proposer, schedule	Proposed change or remediation
Verification	action, result, violations, state snapshot	Outcome of the do-no-harm check

Level	Raincurve does	Humans do
0 · Observe	Builds the graph; produces incidents in shadow	Everything
1 · Recommend	Posts incidents, hypotheses and verified recommendations	Decide and act
2 · Approve	Prepares, verifies, paces and executes on approval	Approve
3 · Automate	Executes specific action types inside agreed limits	Set limits, review

Term	Definition
Background alarm	An alarm not explained by an earlier alarm; either an origin or routine noise
Triggered alarm	An alarm explained by an earlier alarm on a nearby device
Blast radius	The set of services, tenants and endpoints affected by a condition
Biconnected component	A maximal part of a graph with no single vertex whose removal disconnects it
MELT	Metrics, events, logs and traces
MTTR	Mean time to resolution

We publish the limits of each technique alongside its results:

Verification covers connectivity and redundancy. Routing policy, ACLs, VLANs and capacity require separate configuration modeling.
Verification handles device failures; individual link cuts need separate handling. Resilience requirements above 2 use max-flow.
Causal grouping results to date come largely from simulation; production validation is ongoing.
Pacing parameters such as the churn cap are environment-specific and are set during the pilot.