PyCon Italia 2026
In this talk I walked through the process of building a production-ready RAG (Retrieval-Augmented Generation) system from the ground up, using Hexagonal Architecture (also known as Ports and Adapters) to keep the codebase maintainable and testable. The session covered how to structure AI-powered applications so that the core domain logic remains independent of LLM providers, vector databases, and other infrastructure — making it easy to swap components, write unit tests, and evolve the system over time.