🧠 Ajeya AI

Full-loop AI co-pilot for product decisions
React + TypeScript + Vite / Node.js + Express / MongoDB + Redis / Gemini AI

---

Overview

Ajeya AI is a web-based platform where product managers and founders upload customer feedback and ask natural language questions like "What should we build next?". Gemini AI analyzes project data and returns structured, evidence-backed feature recommendations. The system includes a feedback loop: PMs accept/reject AI suggestions, improving future outputs.

This monorepo contains:

packages/
├── client/    # React frontend
├── server/    # Node.js + Express backend
└── shared/    # Shared TypeScript types
docs/         # Documentation
.github/      # PR and Issue templates
docker/       # Docker configuration

---

MVP Scope (Hackathon)

✅ Must ship:

• User authentication (signup/login/logout, JWT)
• Project workspace CRUD
• Document upload (.txt/.csv, paste, max 1MB)
• Single-turn AI Q&A with structured JSON output
• AI output with confidence scores + feedback loop
• Results history
• Responsive UI
• Admin panel (user list, usage stats, AI health)
• Rate limiting (10 AI queries/day)
• JSON validation with fallback
• Basic audit logging

❌ Post-MVP features (deferred):

• Redis caching, multi-turn AI memory
• PDF/DOCX uploads, audio/video transcription
• Slack/Intercom integration
• Stripe/Razorpay billing
• Export to Jira/Linear
• Automated AI retraining

---

Tech Stack

Frontend

• React + TypeScript + Vite
• Redux Toolkit, React Query, TanStack Router
• Tailwind CSS + shadcn/ui, Lucide React icons
• Framer Motion, React Hot Toast

Backend

• Node.js + Express + TypeScript
• JWT auth, Zod validation, Multer file uploads
• Mongoose ODM, MongoDB Atlas
• Logging: Pino + Morgan
• Security: Helmet.js, express-mongo-sanitize

AI

• Google Gemini 1.5 Flash via @google/generative-ai
• Prompt construction: promptBuilder.ts
• JSON validation: jsonValidator.ts
• Fallback parsing: jsonFallback.ts

---

Setup Instructions

Prerequisites

• Node.js 20+
• NPM 9+
• Docker (optional, for local dev)
• MongoDB Atlas account (free M0)
• Gemini AI API key

Install Dependencies

# At repo root
npm install
# Or per package
cd packages/client && npm install
cd packages/server && npm install
cd packages/shared && npm install

Environment Variables

Copy .env.example in each package:

cp packages/server/.env.example packages/server/.env
cp packages/client/.env.example packages/client/.env

Edit variables as needed:

• GEMINI_API_KEY: Your Gemini API key
• MONGODB_URI: MongoDB connection string
• VITE_API_BASE_URL: Backend URL
• Other config: JWT_SECRET, FREE_TIER_DAILY_LIMIT, PORT, CORS_ORIGIN, etc.

Running Locally

# Backend
cd packages/server
npm run dev

# Frontend
cd packages/client
npm run dev

Visit http://localhost:5173 to access the app.

Docker (Optional)

docker-compose up --build

This starts frontend, backend, and MongoDB services locally.

---

References

• Governance & coding standards: GOVERNANCE.md
• Product requirements: PRD.md
• AI prompt guidelines: docs/AI_PROMPTING_GUIDE.md
• API reference: docs/API_REFERENCE.md (need to create)