Skip to main content

Changelog

v1.0.0 — 2026-04-07​

Initial release of Korastratum Agentic AI.

Agents​

  • 22 agents across 7 product areas: Compliance, Fraud, IDV, Lending, Treasury & Settlements, Corridors & Payments, and Inventory
  • 12 Core (Tier 1) rule-based agents with deterministic logic
  • 10 ML (Tier 2) agents with Bayesian scoring and anomaly detection

Platform​

  • Shadow mode — All agents start in shadow mode with 2-week minimum observation period
  • Three operating modes — Shadow, Active, and Disabled per agent per tenant
  • Risk-based auto-act — Four risk thresholds (LOW, MEDIUM, HIGH, CRITICAL) control autonomous action boundaries
  • Decision audit trail — Full logging of every decision with confidence, risk level, reasoning, and human comparison
  • Per-tenant configuration — Independent agent settings, decision history, and metrics per tenant

API​

  • GET /ai/agents — List all registered agents
  • GET /ai/agents/{agent_id}/status — Agent status per tenant
  • POST /ai/agents/{agent_id}/trigger — Manual agent trigger
  • PUT /ai/agents/{agent_id}/config — Update agent configuration
  • GET /ai/agents/{agent_id}/decisions — Paginated decision history
  • GET /ai/agents/{agent_id}/metrics — Performance metrics with shadow comparison
  • POST /ai/agent-decisions — Internal decision logging

Trigger modes​

  • Manual — On-demand via API
  • Scheduled — Cron-based execution via scheduler service
  • Event-driven — Reactive to events from other Korastratum services