How nthperson/comp-380_group_project works
Project Type and Purpose
• RezuMe is a full-stack, single-page application built with React on the frontend and Flask on the backend.
• It helps job seekers tailor PDF resumes to specific job descriptions by combining user-uploaded resumes, job description scraping/parsing, keyword selection, and AI-powered resume generation.
High-Level Architectural Components
• Frontend (React SPA)
• Components folder with Pages, Helpers, Context providers
• Services layer that wraps API calls via Axios
• Backend (Flask REST API)
• app.py exposes endpoints for authentication, PDF upload/listing, JD scraping/parsing, resume generation
• Utility modules for PDF handling, LLM calls (llm_utils.py), embedding store (embeddings_db.py), Firebase integration
• Local embeddings.db and CSV for skill lists
• External Services
• Firebase Authentication, Firestore (metadata), Cloud Storage (PDFs)
• Google Generative AI SDK (Gemini 1.5 Flash) / OpenAI API
• Data Stores
• Firebase services (primary user data & PDFs)
• SQLite-style embeddings database (for local semantic similarity operations)
Relationships and Interactions
• The React app obtains a Firebase ID token on login and attaches it as a Bearer token on every request.
• Frontend service modules call backend endpoints (e.g., /api/upload_pdf, /api/jd_from_url, /api/generate_targeted_resume).
• verify_firebase_token in the backend validates the token against Firebase Auth for protected routes.
• PDF upload/download flows go through pdf_utils.py → Firebase Storage.
• Resume metadata and user profiles are stored/fetched via Firestore wrappers (user_profile_utils.py).
• Job description scraping (via JobDescriptionScraper.py or URL scrapers) feeds into LLM prompt construction.
• Resume generation endpoint uses llm_utils.py to call the Gemini/OpenAI API and returns HTML to the frontend for editing.
• The backend also supports similarity searches against the embeddings database for keyword suggestions.
Key Architectural Patterns and Design Principles
• Clear separation of concerns:
• Presentation layer (React components and contexts)
• API layer (Flask routes)
• Business logic and utilities (PDF parsing, LLM prompting)
• Data layer (Firebase + local embeddings)
• Modular code organization: each major feature has its own utility module and corresponding frontend service.
• Stateless backend: each request is independently authenticated and processed.
• Service-oriented frontend: React service modules abstract away REST calls from components.
• Use of contexts in React for cross-cutting state (e.g., PDF context, targeted-resume context).