Applied AI & Customer service

AI e-commerce chatbot with catalog search and lead capture

Embedded e-commerce assistant with Nuvemshop product, price, variant and availability queries, configurable AI and conversation history.

Back to projectsContact

Used in production operations

Application excerpts from the system built to support day-to-day operations.

Customer-service widget open inside the online store with a form to start a conversation.
Customer service inside the storeWidget embedded in the e-commerce site to start conversations and collect customer details.
Chatbot admin dashboard with sessions, conversation history and service controls.
Session operations and historyDashboard for reviewing conversations, identifying leads and configuring bot behavior.

Context

The store needed to provide initial customer service inside the website, guide customers through products and reduce dependence on manual replies across separate channels.

Problem

Initial service was fragmented, without a structured way to query the catalog, record conversation history and direct customers to the appropriate human channel.

What I delivered

  • Built a FastAPI backend to orchestrate the widget, sessions and integrations
  • Created an embeddable JavaScript widget that collects the customer name, email and phone number
  • Integrated the Nuvemshop API for products, prices, variants and availability
  • Added a configurable AI model, prompt and visual identity managed from the admin dashboard
  • Defined prompt guidance for routing customers to WhatsApp when human support is needed
  • Integrated n8n and Google Sheets to register captured contacts as potential leads
  • Built an admin dashboard for sessions, full message history, leads and bot settings
  • Published the application with Docker, GitHub and Coolify

Outcomes

  • Customers can start a conversation without leaving the online store and submit contact details in the widget.
  • The chatbot queries Nuvemshop for products, prices, variants and availability.
  • Captured contacts are sent by n8n to Google Sheets for follow-up as potential leads.
  • Managers can review sessions and full conversation history in an admin dashboard.
  • The prompt, opening message, name, avatar and AI model can be changed without modifying the store widget.