How ayash-bera/inkei works
Project type and purpose
• Zenith is a full-stack SaaS platform with a React/Vite frontend, a Node/Express backend, a MongoDB database, and a Python-based AI microservice.
• It integrates AI throughout the Secure Software Development Life Cycle by providing features such as GitHub PR review bot, real-time chat, AI‐powered sprint planning, document upload and analysis.
File structure
• Top level:
• Dockerfile, nginx.config, documentation.html, README.md
• AIModels/… (Python AI microservice using Streamlit, with components for file handling, model selection, QA chain, image/text parsing)
• backend/… (Node.js/Express REST & WebSocket server, MongoDB models, auth routes, DB manager util)
• src/, public/, index.html, vite.config.js, tailwind.config.js (React/Vite single-page-app)
• .github/workflows/docker_publish.yml (CI/CD for Docker publishing)
• Patterns:
• Microservice boundaries between frontend, backend, AIModels
• MVC-style backend (routes, models, utils)
• Component-driven React UI
• Deployment artifacts: Dockerfile, nginx.config, vercel.json
Readme insights
• Core features: GitHub Bot, real-time chat (WebSocket), AI Dashboard (upload & analyze docs), Delta Branch Detection, AI Canvas.
• Tech stack: React, Node/Express, MongoDB, Python (AI).
Main components and their interactions
a. Client browser (React SPA)
b. Frontend hosting (Vercel or Nginx + Docker) serving static assets
c. Backend API & WebSocket service (Express + Node.js)
d. MongoDB cluster(s) (MONGO_URI, various role-based DB URIs)
e. Python AI microservice (Streamlit app on port 8501 or wrapped as internal service)
f. External integrations: GitHub API
Data flows:
• User interacts with SPA → HTTP REST calls to Express (login, signup, fetch dashboards)
• SPA establishes WebSocket with backend for real-time chat
• For AI features (upload doc, analyze tasks) → HTTP POST from frontend or backend to Python AI service → returns analysis results → SPA renders on dashboard
• Backend triggers GitHubHandler to call GitHub API for PR reviews → results pushed back to SPA or stored in DB
• All state persisted in MongoDB
Architectural patterns & design principles
• Separation of concerns: UI layer (React), API layer (Express), AI logic (Python)
• Microservice style for AIModels, allows independent scaling & deployment
• MVC-inspired backend (routes → controllers → utils/dbManager → models)
• Component-driven UI with layout, sections, dashboard modules
• Streamlit used as a quick UI for AI features in dev; can be containerized
• CI/CD flows via GitHub Actions and Docker