MG Hub & Kaelii AI

Um ecossistema urbano inteligente que conecta comércio local, agendamentos autônomos e IA descentralizada para transformar pequenas economias.

Comment

Inspiration

The inspiration for the project came from observing the difficulties small local businesses, service providers, and artisans in smaller towns face when transitioning to the digital market. Many traditional SaaS platforms and marketplaces are overly complex or rely on expensive infrastructure. We decided to create the MG Hub / CityHub Network: a lightweight, community-driven, standalone SaaS platform that combines product storefronts and dynamic scheduling workflows into a single, user-centric ecosystem.

What it does

MG Hub centralizes a region's key consumer sectors. Users can browse carousels of local brands, purchase handcrafted goods, or book services in real-time (such as barbershop appointments or auto detailing)—all integrated into a synchronous logistics pipeline. Its standout feature is Kaelii AI: an intelligent assistant integrated end-to-end into the ecosystem. It interacts with customers, answers context-aware questions about pricing, and guides the booking process directly within the app interface.

How we built it

We developed the frontend using agile technologies, focusing on a fluid mobile-first experience, an installable PWA (Progressive Web App), haptic feedback, and a highly readable dark-mode design. The database and authentication infrastructure were built on Firebase (Auth and Firestore), operating in real-time. We integrated the "brain" of Kaelii AI innovatively by creating a custom administrative module within our platform (the Multitenant Hub). This allows any merchant or administrator to train the AI ​​visually and in real-time. The system encrypts these learnings into Base64 records and injects them into the contextual memory accessed by the chat in milliseconds.

Challenges we faced

The biggest challenge was creating an AI training model that was truly accessible to small local merchants—people who neither know how to code nor should be expected to. Developing the multi-tenant logic—ensuring that AI-learned responses were isolated by establishment, end-to-end secure, and synchronized with actual product inventory—required detailed reverse engineering of distributed database queries. Furthermore, ensuring a UI/UX that avoided common scheduling bottlenecks required the development of smart, Kaelii-assisted workflows.

Accomplishments we're proud of

We are extremely proud to deliver a market-ready (production) application capable of bringing the power of Large Language Models (LLMs) to local neighborhood businesses. We created a platform where a local barbershop or small craft shop gains a virtual assistant—trainable in just 10 seconds—that delivers service on par with major multinational corporations.

What we learned

We learned how to maximize performance by combining NoSQL database queries with dynamic knowledge injection (local RAG) for AI. We strengthened our understanding of modular SaaS architecture and gained deep insight into the social impact that well-designed AI can have on small urban economies.

What's next for MG Hub & Kaelii AI

The next step is to evolve Kaelii AI’s cognitive engine to handle voice calls integrated with WhatsApp and process automatic PIX payments directly within the chat. We aim to expand the CityHub network to cover multiple regions connected by a mesh of autonomous artificial intelligence.

Built With

Share this project:

Updates