How baibhab-adhikari/ats-eureka works
Project type and purpose
• ATS-Eureka is a full-stack, AI-powered resume screening web application.
• Backend: FastAPI (Python), exposing RESTful /api endpoints for authentication, single-resume analysis, and batch analysis.
• Frontend: Static HTML/CSS/JavaScript pages (employer and employee portals) served via AWS S3/CloudFront.
• Core functionality: Accept resumes (PDF/DOCX), send text to Google Gemini Flash API, compute matching scores, store results in MongoDB Atlas, enforce rate limits via Redis Cloud.
File structure
• Top level:
• app.py – FastAPI application entry point
• auth.py – JWT auth and user-type logic
• db.py – Database connection to MongoDB Atlas
• helpers.py – Utility functions (parsing files, formatting JSON)
• requirements.txt, Dockerfile, heroku.yml – containerization and deployment
• frontend/ – Static client assets: HTML pages, CSS styles, JS logic for API calls
• assets/ – Images used in static pages (landing, dashboards)
• .env / environment variables for MONGO_URI, GEMINI_API_KEY, REDIS config
Key architectural insights from README
• Dual-interface: separate flows for employers vs. job seekers
• AI integration: Google Gemini Flash 1.5 8B model via REST API
• Rate limiting: Redis-based, sits in front of FastAPI to throttle requests
• Persistence: MongoDB Atlas stores user profiles and analysis history
• Deployment: Backend in Docker on Heroku; static frontend on S3 + CloudFront CDN
Main components & interactions
a. Frontend (Client Browser + S3/CloudFront)
• Pages: /signin, /signup, /employee_home, /employer_home, /plans, /index
• JS modules call FastAPI endpoints under /api
b. API Layer (FastAPI app.py)
• Authentication endpoints (JWT, auth.py)
• Profile endpoints (GET /api/profile, /api/profile/history)
• Resume analysis endpoints (POST /api/employee, /api/employer)
• Rate limiter middleware (Redis Cloud)
c. AI Service (External)
• Google Gemini Flash API (requires GEMINI_API_KEY)
• Receives extracted text, returns matching scores & JSON analysis
d. Database & Cache
• MongoDB Atlas (db.py): stores users, company info, analysis results
• Redis Cloud: rate-limit counters, optional caching of frequent queries
e. Deployment & Infrastructure
• Docker container built via Dockerfile
• Heroku container registry (heroku.yml) runs backend
• AWS S3 + CloudFront serve static assets globally
• Environment variables stored in Heroku config & .env locally
Architectural patterns and design principles
• Layered architecture: Presentation (frontend), API/Business Logic (FastAPI), Data (MongoDB), External Services (AI, Redis)
• Single-page/static multipage front-end calling REST API
• Loose coupling: AI integration abstracted behind helper modules
• Scalability: Stateless API servers (Docker), horizontal scale on Heroku; CDN for static content
• Security: JWT for auth, HTTPS endpoints, rate limiting
Detailed capturing of essential elements
• Don’t include the unused market_data_analysis_scripts folder
• Call out the rate limiter as a middleware component in the API service
• Dual portals by grouping related frontend pages under “Employee UI” and “Employer UI”
• Batch vs. single-resume analysis flows separately if space allows