How ashirrwad/paynow works
Project type and purpose
• This is a Spring Boot–based backend microservice (“PayNow Agent Assist”) that exposes a REST API for making AI-assisted payment decisions. It does not include a frontend in this repo; instead, it offers a well-documented HTTP API, persistence, rate limiting, security, metrics, and an “agent” orchestration layer that uses multiple asynchronous tools.
File structure
• Top-level /pom.xml, Dockerfile, README.md
• src/main/java/com/paynow/agentassist
• config (Spring configs: SecurityConfig, WebConfig, MetricsConfig, ApiKeyConfig, DataInitializer)
• controller (PaymentController, MetricsController)
• interceptor & filter (PerformanceInterceptor, RequestCachingFilter, ApiKeyAuthenticationFilter)
• service (core business: payment, metrics, rate limiting, risk assessment, agent orchestration & tools)
• strategy (DecisionStrategyRegistry + concrete strategies)
• repository (Spring Data for ApiKey, CustomerBalance, PaymentTransaction)
• entity/domain/dto (JPA entities, domain models, request/response builders)
• event (EventPublisher simulating a message bus)
• util (PiiMaskingUtil, PerformanceLogger, ResourceManager)
• src/main/resources/application*.yml (profiles)
• src/test covers integration and unit tests
• Docker & shell script for local testing
README for additional insights
• Describes asynchronous tool execution, performance monitoring (Micrometer), in-memory caching & DB, simulated event bus, security (API keys), observability (logs + Prometheus), agent tooling (balance, risk, case creation), retry/timeout guardrails, fallback strategy
Main components and interactions
a. API Layer
• PaymentController (POST /api/v1/payments/decide)
• MetricsController (/actuator/prometheus)
b. Security & Pre-processing
• ApiKeyAuthenticationFilter for header-based auth
• RateLimitingService invoked via filter/interceptor
• RequestCachingFilter, PerformanceInterceptor add caching & trace IDs
c. Service Layer
• PaymentDecisionService (impl + PerformanceMonitored decorator)
• Agent orchestration: PaymentDecisionAgent → PaymentDecisionProcessor
• Asynchronous tools (CompletableFuture): BalanceTool, RiskSignalsTool, CaseCreationTool
• Strategy pattern for representing decision logic (Aggressive, Conservative, Default)
d. Data Layer
• Spring Data JPA repositories → H2 in-memory database (ApiKeyEntity, CustomerBalance, PaymentTransaction)
e. Eventing & Observability
• EventPublisher logs “PaymentDecisionEvent” (simulated bus)
• Micrometer metrics via Actuator
• SLF4J + MDC trace IDs + PiiMaskingUtil
f. Config & Infrastructure
• application-local.yml / prod.yml profiles
• Dockerfile for containerization
Key architectural patterns & design principles
• MVC separation (Controller → Service → Repository)
• Strategy pattern for pluggable decision logic
• Factory (AgentToolFactory) for tool instantiation
• Decorator (PerformanceMonitoredPaymentDecisionService) for metrics
• Asynchronous parallel execution of I/O-bound tools
• Builder pattern for request/response DTOs
• Cross-cutting concerns handled by Spring Security filters & interceptors
• Notes on critical configuration (timeouts, retries, pool sizes)