VINS-Connect — Intern FAQ Dashboard

A MERN-stack web application that serves as a centralized, searchable knowledge repository for the Vicharanashala summer internship programme. It replaces the recurring cycle of repetitive onboarding queries with a structured crowd-sourced FAQ and query resolution pipeline.

---

🎯 Problem It Solves

• Time inefficiency — The lab team spends significant time repeatedly answering the same routine queries (report submissions, platform access, stipend schedules, remote work guidelines).
• Information asymmetry — Interns joining in later phases lack access to information communicated to earlier batches.
• No centralized repository — Institutional knowledge exists only in ephemeral channels (WhatsApp, email) and disappears after resolution.
• Recurring queries — Without a searchable knowledge base, the same questions are answered over and over.

---

🏗️ System Architecture

Modules

Module	Description
Verified FAQ Repository	Centralized, searchable collection of validated answers. Instantly accessible without contacting the lab team.
Raise Query Module	Dedicated interface for interns to submit questions when no relevant FAQ exists.
Resolve Query Module	Crowdsourced resolution system where interns propose answers, which go through admin validation before becoming public.
Query Tracking System	Transparent mechanism for interns to monitor status, progress, and feedback on their submitted queries.
Knowledge Retention	Promoted resolved queries are atomically written into the FAQ repository for long-term reference.

How They Work Together

1. Intern searches for a query → system checks FAQ repository.
2. Match found → verified answer displayed instantly.
3. No match → intern submits query via Raise Query Module.
4. Fellow interns propose answers → status moves to Pending Approval.
5. Admin reviews and approves → status moves to Resolved.
6. Admin promotes valuable tickets → answer written into the FAQ repository.

---

🔄 Query State Machine (FSM)

Every query passes through exactly four states:

State	Meaning	Trigger
Open	Query submitted, awaiting a response	Intern raises a query with no FAQ match
Pending Approval	Intern proposed a solution, hidden until verified	Intern clicks —Send for Confirmation—
Resolved	Verified response visible to all users	Admin approves intern's answer or answers directly
Promoted	Resolved answer integrated into the FAQ repository	Admin approves response for long-term reference

---

🧠 CS Concepts Applied

Finite State Machine (FSM)

Each query transitions through strictly validated states. The MongoDB schema enforces this via an enum field, separating peer suggestions from verified answers.

Write-Through Cache

When a query is promoted by an admin, its verified response is immediately synchronized into the FAQ repository — ensuring future interns find the answer instantly without loading the query resolution pipeline.

Separation of Concerns (MVC)

Each layer of the stack has a single, decoupled responsibility:

Layer	Responsibility	Technology
Model	Data structure, indexes, schema validation	MongoDB + Mongoose
Controller	Business logic, FSM state verification	Node.js + Express
View	UI, interactions, form handling	React (Vite)
Transport	Stateless async data communication	REST API

---

🛠️ Technology Stack

Layer	Technology	Reason
Database	MongoDB	Flexible, indexed BSON documents for FAQ records and queries
ORM	Mongoose	Strict data types and middleware hook validation
Server	Node.js + Express	Low-latency, asynchronous request handling
Search	TF-IDF (Natural NLP)	Offline semantic-style keyword search via term frequency–inverse document frequency
Client	React (Vite)	Reactive UI with immediate local state management
Auth	JWT	Stateless access control mapped to roles (Intern / Admin)
Styling	Tailwind CSS	Atomic, maintainable utility-first CSS

---

📁 Project Structure

query-portal-extracted/
├── public/
│   └── logo.png              # App logo
├── src/
│   ├── App.tsx               # Main app: routes, views, components
│   ├── main.tsx              # Vite entry point
│   └── index.css             # Tailwind base styles
├── server.ts                 # Express backend + API routes
├── package.json
├── tsconfig.json
├── vite.config.ts
└── README.md                 # This file

---

🔌 API Reference

Authentication

Method	Route	Description	Access
POST	/api/auth/login	Authenticate user, return JWT	Public
POST	/api/auth/signup	Register new intern account	Public

FAQs

Method	Route	Description	Access
GET	/api/faqs	List all verified FAQs	Public
GET	/api/faqs?q=<query>	TF-IDF keyword + semantic search	Public
GET	/api/faq	List all FAQs (auth required)	Intern / Admin
POST	/api/faq	Manually create a FAQ entry	Admin only

Queries

Method	Route	Description	Access
POST	/api/queries	Raise a new query	Intern
GET	/api/queries/my	Intern's own queries	Intern
GET	/api/queries/open	All open/escalated queries	Intern
GET	/api/queries/pending	Pending approval queue	Admin
GET	/api/queries/escalated	Escalated queries	Admin
POST	/api/queries/:id/answers	Propose an answer (→ Pending Approval)	Intern
PATCH	/api/queries/:id/vote	Upvote or downvote a query	Intern
POST	/api/queries/:id/feedback	Rate a query 1–5	Intern
PATCH	/api/queries/:id/approve	Approve proposed answer (→ Resolved)	Admin
POST	/api/queries/:id/promote	Promote to FAQ repository	Admin
PATCH	/api/queries/:id/reject	Reject / reopen query	Admin
PATCH	/api/queries/:id/escalate	Toggle escalation flag	Intern / Admin
POST	/api/queries/:id/admin-resolve	Admin direct resolve of escalated query	Admin

---

🚀 Running Locally

Prerequisites: Node.js, MongoDB

1. Clone / extract the project:

   cd query-portal-extracted

2. Install dependencies:

   npm install

3. Configure environment variables. Create a .env file:

   PORT=3000
   MONGODB_URI=mongodb://localhost:27017/crowdfaq
   JWT_SECRET=your-secret-key-here

4. Start MongoDB (ensure it's running on the URI above), then start the server:

   npm run dev

   The app will be available at http://localhost:3000

5. Default seeded accounts:

   Role	Email	Password
   Admin	admin@example.com	admin@123
   Intern	prasoon@example.com	prasoon@123
   Intern	ridha@example.com	ridha@123
   Intern	siddhant@example.com	siddhant@123
   Intern	vanisha@example.com	vanisha@123
   Intern	swethaa@example.com	swethaa@123
   Intern	ridhya@example.com	ridhya@123

---

📊 Data Models

FAQ Collection

{
  —question—: —What dates do I put on the NOC?—,
  —answer—: —Use the officially communicated start and end dates from your provisional offer letter.—,
  —tags—: [—noc—, —documentation—],
  —createdAt—: —2026-06-02T00:00:00.000Z—
}

Query Collection

{
  —questionText—: —Can I start in July if I have exams now?—,
  —status—: —open | pending_approval | resolved | promoted | escalated—,
  —proposedAnswers—: [{ —answerText—: —...—, —answeredByName—: —...—, —createdAt—: —...— }],
  —upvotes—: 0,
  —downvotes—: 0,
  —rating—: 0,
  —tags—: [—timing—],
  —createdAt—: —2026-06-02T00:00:00.000Z—
}

User Collection

{
  —name—: —Prasoon—,
  —email—: —prasoon@example.com—,
  —role—: —student | admin—,
  —createdAt—: —2026-06-02T00:00:00.000Z—
}

---

🔍 Search: TF-IDF vs. Hash Embeddings

The original design relied on MiniMax AI embedding vectors for semantic search. In this implementation, TF-IDF (Term Frequency–Inverse Document Frequency) is used instead, powered by the natural NLP library. It works fully offline, requires no API key, and produces meaningful relevance scores based on word overlap between the query and all FAQ questions + answers.

Scoring Weights

• TF-IDF corpus match — base relevance score from the FAQ corpus
• +2.0 per matching stemmed question word
• +3.0 per matching tag
• +0.5 per exact word match in the question

---

Built for the Vicharanashala Summer Internship Programme.