How rajapremsai/legaleaseai works
Project Type and Purpose
• This is a full-stack, AI-driven web application (“Legal Document AI Assistant”) built as a monorepo (Turborepo) with separate frontend and backend apps plus shared utilities.
• Its goal is to accept legal document uploads, process them through Google Cloud’s Document AI and Vertex AI for analysis, simplification, Q&A, risk assessment, etc., and present results via a responsive Next.js UI.
Top-Level Architectural Components
• Frontend (apps/frontend): Next.js/React-TypeScript PWA, served via Cloud Run (or CDN), handles user interactions, file upload UI, real-time Q&A chat, dashboards, and offline support.
• Backend API (apps/backend): Node.js/Express-TypeScript microservice exposing REST endpoints, implements authentication, document ingestion, AI orchestration, versioning, privacy, template library, risk assessment, analytics, and scheduler.
• Shared (packages/shared): Common TypeScript types, schemas, migration logic.
• Data Stores:
• Cloud Firestore (NoSQL) – for user sessions, workspace/document metadata, templates, audit logs.
• Cloud SQL (PostgreSQL) – for relational data: version control, complex queries, historical snapshots.
• Cloud Storage bucket – for binary document storage (PDF/DOCX/TXT).
• AI/ML Services (GCP):
• Document AI – OCR and structure extraction.
• Vertex AI (PaLM 2) – generative summarization, Q&A, risk assessment.
• Infrastructure:
• Cloud Run (frontend & backend).
• Cloud Build / GitHub Actions for CI/CD.
• Docker Compose for local dev.
• Deployment manifests for Cloud Run, autoscaling, environment parameterization.
• External Integrations: OAuth (Google IAM), Monitoring & Logging (Cloud Monitoring, Stackdriver).
Relationships & Data Flows
1. User (Browser/PWA) → Frontend Next.js
2. Frontend → REST API (Express) via HTTPS
3. API → Auth Service → Identity (Firebase Auth or custom)
4. API → Storage Service → Uploads file to Cloud Storage
5. API → Document AI (GCP) → structured output
6. API → Vertex AI for summarization/Q&A/risk assessment
7. API → Firestore & Cloud SQL (via repositories) to persist metadata, versions, audit logs
8. API → Returns processed data back to Frontend
9. Frontend → Renders dashboards, Q&A interface
10. Scheduler Service within API triggers periodic cleanups, auto-delete, template seeds, analytics jobs
Architectural Patterns & Principles
• Monolithic microservice for backend, subdivided by services (aiAnalysis, documentVersioning, auth, etc.).
• Repository pattern for data access.
• Middleware chain for auth, rate limiting, validation, performance, security.
• Shared code via npm workspace to enforce DRY (schemas, migrations, utils).
• CI/CD pipelines for lint, tests (unit, integration, e2e, performance, security), build, deploy.
• Infrastructure as Code: Cloud Run YAML, cloudbuild.yaml, docker-compose.yml.