Architecture

A six-layer cloud-native system, all running on Amazon EKS, designed to serve institutional treasury banking clients across all business segments.


The Six Layers

Layer Technology Role
Architecture Overview React · React Native · Spring Boot Full system data flow and layer interaction diagram
Technology Stack All components Decision record — what was chosen and why
Layer 1 — Presentation React + TypeScript · React Native Operator workspace · banker mobile · executive dashboard
Layer 2 — API Gateway Spring Boot · GraphQL · REST Workspace BFF · external API · domain services · Temporal bridge
Layer 3 — Workflow Orchestration Temporal.io · Java 25 LTS Durable workflow engine · signal pattern · error recovery · KEDA workers
Layer 4 — Agentic AI Python · FastAPI · LLM (TBD) AI Assistant · Document Review · AML Pre-Fill · 6 specialised agents
Layer 5 — Integration Services Spring Boot · Resilience4j Multiple external system adapters · circuit breaker · OpenAPI contracts
Layer 6 — Data Amazon Aurora PostgreSQL · S3 · Redis Case data · Temporal persistence · document storage · session cache
Infrastructure Amazon EKS · KEDA · ArgoCD Kubernetes cluster layout · event-driven autoscaling · GitOps · anchita-cli
Go Components Go anchita-cli operations CLI + Notification Service WebSocket hub — rationale, design, learning path
Integration Strategy Stripe · Plaid · FRED · Mock services Real APIs vs mock services vs synthetic data — how Anchita connects to the outside world

Design Philosophy

Every architectural choice in Anchita is made with three constraints in mind:

  1. Durability over availability — onboarding cases span hours to days. The system must survive restarts, infrastructure failures, and partial outages without losing workflow state or requiring operator re-entry. Temporal.io is the direct answer to this constraint.

  2. Compliance as a first-class concern — BSA/AML attestation, document review, and beneficial ownership verification are not optional steps that the system can route around. They are modelled as non-bypassable signal wait points in the workflow.

  3. Operational visibility over throughput — the primary users are operations teams managing queues of cases. SLA tracking, stage visibility, and AI-assisted decision support are more valuable than raw throughput optimisation.


Table of contents


↑ Back to top

Anchita Platform — Fictional Reference Architecture for Cloud-Native Institutional Banking