We have been involved with Relìa from the very beginning: from the naming – designed to evoke ‘Report’ and ‘Artificial Intelligence’ – to the brand identity, right through to the entire digital experience. We designed the platform by envisaging a clear journey, in which the user always knows where they are in the report, guided step by step towards an ecosystem where the visual identity and user experience convey clarity and reliability.
We have designed the platform as a two-tier architecture: a Next.js web application and a Python API back-end, with distinct roles and permissions for companies, consultants and administrators.
On the front-end, we developed the application using React and Next.js, TypeScript and Tailwind, utilising server-side rendering with data prefetching to ensure high performance. The interface is multilingual, with persistent session authentication and automatic refresh, and forms validated against shared schemas to ensure the integrity of the report data.
The back-end, developed using FastAPI and PostgreSQL, orchestrates a multi-agent system based on LangGraph: a supervisor agent coordinates agents specialising in ESG (Environmental, Social, Governance) domains in accordance with the VSME standard. The agents conduct the interview, validate the responses and compile the structured data for the report, utilising persistent conversational memory and semantic search on PostgreSQL + pgvector for domain knowledge and uploaded documents.
The front-end and back-end communicate via REST APIs with OpenAPI contracts, from which we automatically generate TypeScript types: a single source of truth that ensures consistency and maintainability over time.
