How getsan4u/loan-origination-system works
Project type and purpose
• This is a full‑stack, open‑source Loan Origination System (“LOS”) built on Node.js/Express and the PeriodicJS container framework.
• It exposes both web UIs (EJS and React apps) and a flexible, rules‑driven Decision Engine API, plus integrated Machine Learning, Cron scheduling and third‑party integrations (e‑sign via DocuSign, AWS SageMaker, etc.).
File structure
• Top level:
• index.js, content/config/startup.js → application bootstrap using PeriodicJS
• app/views → legacy EJS templates; public/extensions and static → compiled React and CSS assets
• content/container → two “containers”:
– decision‑engine‑service‑container → encapsulates the Decision Engine microservice (controllers, routers, transforms, models, crons)
– periodicjs.container.default → default CMS‑style container for core pages and admin
• content/config/databases/standard/models → same domain models in Lowkie (flat‑file), Mongoose (MongoDB) and Sequelize (SQL) variants
• Cron scripts (content/container/.../crons and utilities/mlcrons) drive scheduled tasks (emails, rules updates, ML model refreshes).
• Public/extensions/periodicjs.ext.reactapp → client‑side React application built and served via the container.
Key insights from README
• Modular, container‑based architecture
• Open API layer for integration (Extensive routers and controllers under decision‑engine‑service‑container)
• Integrated ML workflows and charting components under utilities and resources
Main components
a. Client‑side UI
• Browser → EJS views (server‑rendered)
• Browser → React SPA (served from public/extensions/periodicjs.ext.reactapp)
b. API & Application Server
• PeriodicJS container loader (index.js + startup.js)
• Core container (periodicjs.container.default)
• Decision Engine container
• Controllers & Routers → Express middleware for REST endpoints (auth, decision, ml, ocr, optimization, integration, user, etc.)
c. Data Layer
• Database(s): MongoDB via Mongoose, (Postgres/MySQL/SQLite) via Sequelize, flat‑file via Lowkie
• Filesystem storage for templates, sample data, model scripts
d. Supporting Services & Integrations
• Cron scheduler → internal jobs under crons and mlcrons
• AWS SageMaker / Digifi model scripts → utilities/mlresource & files/digifi_model_scripts
• DocuSign → utilities/docusign
• Third‑party OAuth, email services, payment gateways
Relationships & interactions
• HTTP(S) request flow:
Client UI → Express Router → Controller → Transform → Model → Database → Transform → Controller → HTTP Response → Client UI
• Decision Engine API calls: controllers invoke rule evaluation modules and model scoring via utilities and mlresource helpers
• Cron jobs schedule: trigger ML retraining, data refresh, batch decision runs → update database
• External integration calls: DocuSign for e‑sign workflows; SageMaker for model training & inference
Architectural patterns & design principles
• MVC (EJS views + controllers + models)
• Modular container pattern (PeriodicJS containers act like pluggable microservices)
• Adapter pattern in Transforms (unify various data sources / ORMs)
• Cron‑driven batch processing for asynchronous workflows
• Separation of concerns: UI, API, business logic, data, external integrations
Final recommendations
• Every major directory as a component (app/views, content/container/decision‑engine, content/config, public/extensions)
• Do not over‑aggregate: break out Controllers, Routers, Transforms, Utilities