Finance AI is a personal financial management system powered by AI, where users manage accounts, transactions, budgets, investments, and rewards programs (Livelo, Dotz, etc.) all in one place, with intelligent insights generated by LLM.
Finance AI is a personal finance management application that treats finance as a true digital product, not just a pretty spreadsheet. The idea is to centralize the management of bank accounts, transactions, monthly budgets, investments, and rewards programs (such as Livelo and Dotz) in one place, while an AI layer analyzes usage patterns and supports user decision-making.
From an architectural standpoint, the project was built on Next.js 15 + TypeScript, following a Domain Modules approach inspired by Clean Architecture and DDD. Each business area (accounts, transactions, budgets, investments, AI, points) is isolated in its own module, with well-defined services, Zod schemas, and types, facilitating maintenance, testing, and future evolution. The data layer is abstracted via the Repository Pattern, using Firebase Firestore as the main database, and all business logic is SSR-safe, without depending on browser APIs.
AI integration is achieved through an AI Tools Layer, which exposes specific tools for: automatically categorizing transactions, generating financial insights, analyzing spending patterns, and suggesting budget adjustments. This layer communicates with an external LLM (OpenAI) and is designed to be extensible, allowing the inclusion of new tools in the future without directly coupling the application to the AI implementation.
In addition to features, Finance AI also explores good engineering practices:
– 📚 ADR (Architectural Decision Records) to record important decisions
– ✅ Health checks and smoke tests to validate the application's integrity
– 🧪 Runtime validation with Zod on all domain entries
– 🎨 Tailwind CSS + shadcn/ui to build a consistent and scalable interface
This project is simultaneously an evolving product and an engineering laboratory, where I explore how to combine clean architecture, a well-modeled domain, and AI applied to a real-world problem: managing money better.
Nelson Christovam Neto