How t-k-234/skill-match works
What this project is (type + purpose)
This is a monolithic Flask web application (“Skill Matching Portal”) that serves server-rendered HTML pages for two user roles:
• Students: register/login, manage skill profiles, browse jobs, see match scores.
• Employers: register/login, post jobs with requirements, see matched students.
The “AI” portion is a TF‑IDF vectorization + cosine similarity scoring flow used to rank student↔job fit.
The persistence layer is SQLite (skillmatch/users.db). Auth is session-based (per README).
File structure signals (how to map code to components)
From the tree:
• skillmatch/app.py
Likely the Flask application entrypoint: routes, session/auth handling, request orchestration, HTML rendering.
• skillmatch/ml_model.py
Likely encapsulates text preprocessing / TF‑IDF / cosine similarity and match ranking logic.
• skillmatch/users.db
SQLite DB file; contains tables for users, roles, student skills, jobs, etc.
• skillmatch/templates/*.html
Jinja2 templates for server-side rendered UI: login/register/dashboards/job list/match results.
• skillmatch/static/css/*.css
Static styling served to the client.
• skillmatch/model.ipynb
• Presentation (HTML templates + CSS)
• Application/Controller (Flask routes in app.py)
• Domain/Service (matching logic in ml_model.py)
• Data (SQLite users.db)
Client (Web Browser)
• Student UI
• Employer UI
• (Same web app, different pages/flows)
Flask Web Server (skillmatch/app.py)
• Auth & Session Manager (login/register, role-based routing)
• Student Controller/Routes
• profile + skills CRUD
• view jobs
• request match results
• Employer Controller/Routes
• job posting CRUD
• view matched students
• Template Renderer (Jinja2) (renders templates/*.html)
• Static File Server (serves static/css/*)
Matching/ML Module (skillmatch/ml_model.py)
Split into steps (since README explains them):
• Text preprocessing (tokenization/normalization)
• TF‑IDF vectorizer
• Cosine similarity scorer
• Ranker (top-N matches + percentage formatting)
Database (SQLite: skillmatch/users.db)
• users (credentials, role)
• student_profiles (profile fields)
• student_skills (skill text or normalized list)
• jobs (job description + required skills)
• (Optional) matches (if persisted; if not, show as computed on-demand)
That it informs/validates ml_model.py, but is not on the runtime request path.
Registration/Login
• Browser → Flask /register or /login
• Flask → SQLite (create user / verify credentials)
• Flask → Session Store (Flask session cookie; typically cookie-based signed session)
• Flask → Browser (render register.html / login.html)
Student updates skills
• Browser → Flask “update profile/skills” route
• Flask → SQLite (persist skills)
• Flask → Browser (render student_dashboard.html)
Employer posts a job
• Browser → Flask “create job” route
• Flask → SQLite (persist job requirements/description)
• Flask → Browser (render employee_dashboard.html or jobs.html)
Matching request (key architecture path)
• Browser → Flask “match” route (e.g., view matches)
• Flask → SQLite (fetch job descriptions / student skills)
• Flask → ml_model.py (TF‑IDF + cosine similarity)
• ml_model.py → Flask (ranked results + scores)
• Flask → Browser (render match_result.html / results template)
Examples:
• “Credentials”, “Session cookie”
• “Student skill text”
• “Job description/requirements”
• “TF‑IDF vectors (in-memory)”
• “Similarity scores + ranked list”
• Monolith / single deployable Flask app: no separate backend services.
• Server-side rendering (SSR): templates in templates/ render HTML; not a SPA.
• Layering:
• Routes/controllers in app.py
• Matching service in ml_model.py
• Persistence in SQLite
• On-demand inference/scoring: cosine similarity computed at request time (likely not precomputed).
• Role-based UX: student vs employer dashboards and routes.
• skillmatch/app.py → “Flask App / Controllers / Session Auth / Routing”
• skillmatch/ml_model.py → “Similarity Engine (TF‑IDF + Cosine Similarity)”
• skillmatch/users.db → “SQLite Database”
• skillmatch/templates/*.html → “SSR Templates (Jinja2)”
• skillmatch/static/css/*.css → “Static CSS”
• skillmatch/model.ipynb → “Offline model exploration (non-runtime)”