Skip to main content

RiverGen AI Database Architecture (1.0.2)

Page Outline

All RiverGen AI 1.0.2 operations are persisted within a centralized PostgreSQL cluster. The schema is designed for high-concurrency metadata ingestion and supports complex relational mapping between training intents and operational decisions.


Model Studio (MSA) Schemas

These tables maintain the integrity of the RiverGen Intelligence Factory.

Table NameRole in RiverGen1.0.2 Implementation Detail
task_intentsDeclarative task spec.Includes rg_version and spatial constraint filters.
data_sourcesInput registry.Supports SQL, S3, and Parquet stream descriptors.
training_jobsExecution tracking.Maps to the internal rg-compute-cluster hardware.
job_status_historyAudit stream.Phase-specific progress (Ingestion -> Analysis -> Lock).
trained_modelsArtifact registry.The official source-of-truth for active platform models.

Decision Intelligence (DIA) Schemas

These tables support the RiverGen Decision Brain and autonomous correction loops.

Table NameRole in RiverGen1.0.2 Implementation Detail
decisionsAutonomous logs.Persists the reasoning string for every platform action.
agent_metadataTelemetry buffer.High-velocity storage for agent performance metrics.
decision_guidelinesGovernance layer.Stores general boundaries and prompt instructions.
predictionsRaw ML logs.Stores pre-refined inference results for drift analysis.

Intelligence Relationship Diagram

In RiverGen 1.0.2, every decision is traceable back to its training origin: