N.
All work

SpecMorphAI-Powered SRS to UML Generator

03 / 07
03
Year
2025
Role
Full-Stack AI Developer
Duration
3 months
Client
Vennisys

Live project Source

Overview

Developed an enterprise-grade AI system that automates the transformation of Software Requirements Specification documents into UML diagrams. Features a 4-stage automated pipeline and achieves 85–95% extraction accuracy for well-structured documents.

Challenge

Software architects spend 4–8 hours manually creating UML diagrams per requirements document - a process that is slow, error-prone and difficult to maintain as requirements evolve.

Solution

Built a FastAPI backend with 17 endpoints integrating Google Gemini and OpenAI GPT APIs via LangChain. Implemented token-based text chunking for optimal LLM processing and PlantUML integration for automated generation of Class, Sequence and Use Case diagrams.

Results

  • 01

    90% reduction in diagram creation time (hours to minutes)

  • 02

    85–95% extraction accuracy for well-structured documents

  • 03

    Automated generation of 3 UML diagram types

  • 04

    Real-time progress tracking across 4 pipeline stages

Technology

Next.js 16React 19TypeScriptPythonFastAPILangChainGoogle GeminiOpenAIPlantUMLTailwindCSSTanStack Query