MarketMe — AI Marketing Agent Platform v2.0

A Flask app powered by Amazon Nova 2 Lite (chat) and Nova 2 Sonic (voice).
AI agent that chats, sends campaigns, monitors email, scrapes leads, analyses images, and controls the UI.

Project Structure

marketme/
│
├── app.py                  # Flask app factory — entry point
├── config.py               # All env-var config in one place
├── extensions.py           # Flask extensions (db, socketio, jwt, celery)
├── models.py               # SQLAlchemy models
├── tasks.py                # Celery background tasks
│
├── routes/
│   ├── __init__.py
│   ├── auth.py             # /api/auth/*  — register, login, miracle link
│   ├── business.py         # /api/business — create, settings, generate page
│   ├── campaigns.py        # /api/campaigns — CRUD + send
│   ├── chat.py             # /api/chat — AI agent (text + image)
│   ├── contacts.py         # /api/contacts — CRUD, import, CSV pool
│   ├── inbox.py            # /api/email-threads
│   ├── products.py         # /api/products — CRUD
│   └── public.py           # /biz/<slug> public page + SPA catch-all
│
├── sockets/
│   ├── __init__.py
│   └── events.py           # All SocketIO event handlers + voice bridge
│
├── utils/
│   ├── __init__.py
│   ├── csv_utils.py        # Shared CSV contacts pool helpers
│   ├── email_utils.py      # SMTP send + IMAP fetch
│   ├── intent_handler.py   # Maps AI intents → DB / Celery actions
│   ├── nova_utils.py       # Nova AI client (chat, page gen, email AI)
│   ├── scraper.py          # Playwright lead scraper
│   └── serializers.py      # Model → dict helpers for JSON responses
│
├── templates/
│   └── index.html          # Single-page app HTML (Tailwind + SocketIO)
│
├── static/
│   └── js/
│       └── app.js          # All frontend JavaScript
│
├── .env.example            # Copy to .env and fill in credentials
└── README.md

Quick Start

# 1. Install dependencies
pip install flask flask-socketio flask-sqlalchemy flask-jwt-extended \
            celery[redis] redis openai websockets playwright \
            itsdangerous python-dotenv
playwright install chromium

# 2. Configure environment
cp .env.example .env
# Edit .env with your Nova API key, SMTP credentials, etc.

# 3. Start Redis (required for Celery)
redis-server (might not be needed if you have used celery and redis on windows)

# 4. Run the Flask app
python app.py

# 5. In a separate terminal — start the Celery worker
celery -A app:celery_app worker -l info --pool=solo
# 6. In a separate terminal — start the Celery beat scheduler
celery -A app:celery_app worker beat -l info

The app will be available at http://localhost:5000

What runs where

Process	Responsibility
python app.py	Flask web server, REST API, SocketIO real-time events
celery worker --beat	IMAP inbox monitor (every 2 min), campaign sender (every 1 min), lead scraping, follow-up emails

Features

🤖 AI Agent (Nova 2 Lite)

• Text chat with full context history
• Image analysis — click 📎 to upload any image, agent describes and gives marketing insights
• UI control — say "show me contacts", "go to campaigns", "add a product", "dark mode"
• Web grounding — ask about competitors, markets, industry trends
• Intent detection — automatically triggers app actions from natural language

🎙️ Voice Agent (Nova 2 Sonic)

• Real-time bidirectional speech via WebSocket
• Auto-restarts before the 8-minute session limit
• Proper audio resampling to 24kHz regardless of browser sample rate

📧 Email System

• Campaigns — send to contacts + extra emails (type, paste, CSV upload)
• IMAP monitoring — checks inbox every 2 minutes via Celery Beat
• Auto-reply — AI drafts and sends replies to agreed/interested/question emails
• Background — campaigns keep sending even when browser is closed

👥 Contacts & Shared Pool

• Shared CSV (shared_contacts.csv) — ecosystem of contacts all users can access
• New contacts added via any method are appended to the CSV
• Import — type emails, paste comma-separated, upload CSV, or pick from shared pool
• Find Leads — Playwright headless browser searches Bing/DuckDuckGo, falls back to CSV pool

🎨 UI

• Dark / Light mode toggle (persistent, also agent-controllable)
• Chips-style multi-email input in campaigns and import
• Drag-and-drop CSV upload
• Real-time notifications for inbox, campaigns, leads

---

Agent Commands (examples)

"Show me my products"           → navigates to Products panel
"Add a new product"             → opens Add Product modal
"Go to contacts"                → navigates to Contacts panel
"Find tech leads in Lagos"      → starts Playwright lead search
"Switch to light mode"          → toggles UI theme
"Launch a campaign for my CRM"  → drafts campaign email
"Analyse this image"            + 📎 upload → Nova 2 Lite image analysis
"Generate my business page"     → AI builds landing page at /biz/your-slug

---

File Structure

app.py                 ← entire application (2100+ lines)
sample_contacts.csv    ← 20 seed contacts for the shared pool
shared_contacts.csv    ← auto-created, appended by the app
marketme.db            ← SQLite database (auto-created)
.env                   ← your configuration

---

Environment Variables Reference

Variable	Required	Default	Description
NOVA_API_KEY	✅	—	Amazon Nova API key
SECRET_KEY	✅	dev default	JWT signing key (32+ chars)
SMTP_HOST	✅	smtp.gmail.com	Outgoing mail server
SMTP_PORT	—	587	SMTP port
SMTP_USER	✅	—	Email username
SMTP_PASS	✅	—	Email password / app key
IMAP_HOST	—	imap.gmail.com	Inbox monitoring server
IMAP_PORT	—	993	IMAP SSL port
REDIS_URL	—	localhost:6379	Celery broker
DATABASE_URL	—	sqlite:///marketme.db	Database
APP_URL	—	http://localhost:5000	Public URL (for live chat)
CSV_PATH	—	shared_contacts.csv	Shared contacts file path

---

Production Notes

• Use gunicorn with eventlet worker instead of Flask dev server
• Set SECRET_KEY to a long random string
• Use PostgreSQL instead of SQLite for DATABASE_URL
• Use Redis Cloud or ElastiCache for REDIS_URL
• Run behind nginx with SSL for voice (requires HTTPS for microphone access)

# Production run example
gunicorn --worker-class eventlet -w 1 -b 0.0.0.0:5000 app:app