Banque des Territoires, Polyconseil & Hugging Face deploy sovereign RAG for EduRénov program

AI Impact Summary

BdT's project demonstrates a sovereign Retrieval-Augmented Generation (RAG) deployment that uses open-source Hugging Face tooling to answer inquiries about the EduRénov program while keeping data on French-controlled infrastructure. The architecture combines an Orchestrator, Retriever, and Reader with a vector store, leveraging Text Generation Inference, Transformers, Sentence Transformers, and Tokenizers to fetch and cite EduRénov documentation. This approach reduces manual email workload and speeds up stakeholder responses, aligning with public-sector data sovereignty requirements and enabling scalable support for 10,000 renovation projects.