AI-Powered RFP Management System (MERN + Google Gemini)

Implements the assignment: create RFPs from natural language, manage vendors, send RFPs via email, parse vendor responses, and compare proposals with AI assistance. fileciteturn0file0

1. Tech Stack

• Frontend: React + Vite, Redux Toolkit, Axios
• Backend: Node.js, Express
• DB: MongoDB (local or Atlas)
• AI: Google Gemini API (free tier)
• Email: Nodemailer (SMTP) + IMAP (imap-simple)

2. Prerequisites

• Node.js >= 18
• npm or yarn
• MongoDB running locally (mongodb://localhost:27017) or an Atlas URI
• Google Cloud project with Gemini API enabled and an API key
• Email account for send/receive (e.g. Gmail with App Password)

3. Backend Setup

cd backend
cp .env.example .env
# edit .env with your MONGO_URI, GEMINI_API_KEY, SMTP, IMAP details
npm install
npm run dev

Backend: http://localhost:5000.

4. Frontend Setup

cd frontend
npm install
npm run dev

Frontend: http://localhost:5173.

5. Basic Flow

1. Create RFP from natural language

   • In UI, type description (items, quantities, budget, delivery, terms).
   • Click "Generate Structured RFP" (calls /api/ai/rfp-structure using Gemini).
   • Review JSON, click "Save RFP".

2. Manage vendors

   • Add vendors (name + email) in UI (stored in MongoDB).
   • Checkbox vendors you want to invite.

3. Send RFP emails

   • Select an RFP from the list.
   • Check vendors and click "Send selected RFP to checked vendors".
   • Backend uses Nodemailer + SMTP.

4. Receive vendor responses

   • Vendors reply to the same email address.
   • Hit GET /api/email/fetch (you can wire a simple button or run via Postman) to pull unread emails from IMAP.
   • For each email, call POST /api/ai/parse-vendor-email with { rfpId, vendorId, emailBody, rawEmailId } to create Proposal records.

5. Compare proposals

   • In UI, click an RFP.
   • It loads /api/proposals/rfp/:rfpId and /api/ai/compare/:rfpId.
   • Gemini returns scores, summaries, and a recommendation which is shown in the right panel.

6. Notes

• This is a minimal but end-to-end implementation focused on the single-user flow described in the assignment.
• You can extend the frontend with nicer tables, filters, and detailed proposal views as needed.