DomainFlow

An enterprise SaaS that orchestrates GPT-4o & Gemini to automate software design. Features a custom multi-LLM pipeline for generating UML diagrams, user personas, and requirement documentation.

System Architecture

• Client Layer: React SPA handling complex state for interactive matrices (DSM) and canvas rendering (UML)
• API Layer: Express.js middleware manages Authentication (RBAC) and routing
• Intelligence Layer: Dedicated service module routes prompts to specific AI providers based on strength
• Data Layer: Normalized MySQL schema (30+ tables) stores strict relationships between Entities, Features, and Actors

Multi-Model AI Integration

• OpenAI GPT-4o: Used for deterministic tasks like generating Lean Canvas sections, defining strict MoSCoW feature sets, and validating business logic rules
• Google Gemini 1.5 Pro: Utilized for high-context generation, specifically for fleshing out detailed User Personas, Use Case preconditions, and Triggers where large context windows are beneficial

Algorithmic Type Combination Analysis

• Cartesian Product Algorithm generates every possible combination of entity types (e.g., UserType × SubscriptionStatus × Region)
• System generates thousands of combinations, batch-processed (100 items/batch) through GPT-4o
• Automatically flags "Impossible" or "Special Case" combinations, reducing manual QA time by 90%

Tech Stack

• Frontend: React 18, Vite, Material-UI, Konva (Canvas/UML Visualization)
• Backend: Node.js, Express.js (REST API)
• Database: MySQL 8.0 (AWS RDS), Redis (Caching)
• AI Layer: OpenAI GPT-4o (Logic/Validation), Google Gemini 1.5 Pro (Context Generation)
• Infrastructure: AWS RDS, JWT Auth, Docker