How nehasreenidhi/interns-skill-management works
What this project is (type + purpose)
This is a monolithic Flask web application (single app.py) serving server-rendered HTML pages (Jinja templates) with static assets, backed by MongoDB for persistence. It adds a semantic search subsystem using Sentence-Transformers embeddings + FAISS vector index so mentors can search interns by intent rather than exact keyword match.
Core purpose: intern skill portfolio management + mentor discovery/search.
What the file tree tells you about architecture
From the tree:
• app.py
Likely contains Flask app initialization, route handlers, auth logic, DB access, and FAISS/embedding integration. Since there’s no blueprints/, services/, or models/ directories, expect a single-file “modular-by-functions” structure.
• templates/*.html
Server-rendered UI. Templates map closely to user-facing modules/pages, which is useful for diagramming functional components:
• landingPage.html
• login.html, register.html, reset_password.html (auth flows)
• intern_dashboard.html, intern_profile.html
• mentor_dashboard.html, search_interns.html
• profile_settings.html
• static/styles.css, static/images/*
Frontend static assets. Charting is likely done in templates via Chart.js (even if not shown in tree, README states it).
• A single deployable Flask service
• Server-rendered HTML views
• MongoDB for document storage
• Embedding model + FAISS index for semantic search
• Password hashing using bcrypt
What the README adds (runtime components and flows)
The README defines a two-role system (Intern vs Mentor) and a data/ML pipeline:
• Skills stored in MongoDB
• Skill names embedded via Sentence-Transformers (all-MiniLM-L6-v2)
• Embeddings indexed in FAISS for fast similarity search
• Mentors query -> embedded -> searched in FAISS -> matching interns returned
• Visualizations: pie charts + activity heatmap (frontend rendering, backend provides data)
Semantic search path (embedding + indexing + querying)
• Actors: Intern, Mentor
• System: Intern Skill Management System (Flask)
• External: MongoDB, Sentence-Transformers model, FAISS index
• Purpose: communicate boundaries and who uses what
• Browser (HTML/Chart.js)
• Flask web app
• MongoDB
• FAISS index store
• Embedding model component
• Intern updates skills → embedding → FAISS update
• Mentor search → embedding → FAISS query → MongoDB hydrate
When and how the FAISS index updates
• On every skill change? scheduled rebuild? startup build?
Identifier mapping
• FAISS returns vector IDs → those map to intern/user IDs in MongoDB. Show this hydration step explicitly.
Where Chart.js runs
• In browser; Flask only supplies the data. Show Chart.js on the client side.
Security boundary
• Password hashing with bcrypt in Flask before MongoDB write; never store plaintext passwords.