Online Kotiba (Fullstack AI Secretary)

Production-ready foundation for a mobile-first online secretary system with React + Express + Supabase + AI pipeline.

Stack

• Frontend: React (Vite), JavaScript (JSX), Tailwind, shadcn-style UI, Zustand, TanStack Query
• Backend: Node.js + Express
• Database/Auth/Storage: Supabase PostgreSQL + Auth + Storage
• AI: UzbekVoice STT/TTS + OpenAI
• Deploy: Vercel (frontend) + Render/Railway (backend) + Supabase

Monorepo structure

• frontend/: responsive admin web app
• backend/: secure API + scheduler jobs
• supabase/schema.sql: DB schema + RLS policies

Quick start

1. Install dependencies:

npm install

2. Configure env files:

• Copy backend/.env.example -> backend/.env
• Copy frontend/.env.example -> frontend/.env

3. Apply Supabase SQL:

• Run supabase/schema.sql in Supabase SQL Editor
• Create storage bucket: reminder-audio

4. Start both apps:

npm run dev

• Frontend: http://localhost:5173
• Backend: http://localhost:8080

Environment variables

Backend (backend/.env)

• PORT
• APP_ORIGIN
• CORS_ORIGINS
• SUPABASE_URL
• SUPABASE_ANON_KEY
• SUPABASE_SERVICE_ROLE_KEY
• OPENAI_API_KEY
• OPENAI_MODEL
• UZBEKVOICE_STT_URL
• UZBEKVOICE_STT_API_KEY
• UZBEKVOICE_TTS_URL
• UZBEKVOICE_TTS_API_KEY
• TELEGRAM_BOT_TOKEN

Frontend (frontend/.env)

• VITE_API_BASE_URL
• VITE_SUPABASE_URL
• VITE_SUPABASE_ANON_KEY

API response contract

Every endpoint returns:

{
  "success": true,
  "data": {},
  "message": ""
}

Implemented modules

• Auth/profile/settings
• Secretary pipeline (text + voice)
• Reminders (pre-generated TTS audio URL)
• Meetings (+auto-message job queue)
• Clients
• Expenses
• Dashboard analytics + AI recommendation text
• Admin overview
• Cron job processor (system_jobs)

Deployment notes

• No hardcoded URLs. Use env vars only.
• Scheduler is backend-only (never frontend).
• Configure CORS allowlist via CORS_ORIGINS.
• Frontend can deploy to Vercel directly from frontend/.
• Backend can deploy to Render/Railway from backend/ with start command:

npm start

Important production hardening checklist

• Add Redis-backed queues (BullMQ) for high job volume
• Add retry dead-letter table for failed jobs
• Add Sentry/Logtail for central logging
• Add stricter zod schemas for settings/prompt management
• Add e2e tests for scheduler and reminder cadence correctness
• Add signed URL strategy if audio should not be public