Brightcon 2025, hackathon & courses in Grenoble and online

The accuracy and reliability of LCA results heavily depend on LCA databases. However, missing data and the cumbersome process of searching and comparing datasets makes LCA modelling manual inefficient, and error prone. The project aims to integrate Graph Retrieval-augmented Generation (GraphRAG) with Ecoinvent to make inventory search process simpler, and reliable. Retrieval-Augmented Generation (RAG) is a technique used to enhance the performance of LLMs. It combines LLMs with external knowledge sources during inference. However, traditional RAG approaches often rely on flat document retrieval, which limits their capacity to capture complex relationships inherent in structured datasets.

To address this limitation, GraphRAG has been introduced, it combines graph-based retrieval with large language models (LLMs) to improve contextual understanding and enable more semantically rich querying.
In this project, we utilize Brightway to import the Ecoinvent database and transform its activity and exchange data into a graph structure. Based on this graph representation, we implement the GraphRAG approach to reason over interconnected activities, exchanges, and metadata, improving both the relevance and depth of retrieved information. The framework is deployed as an interactive web interface that allows users to query specific activities, receive suggestions for similar processes, and explore associated exchanges

Highlights/Discussion points:
1. Exploring the role of AI in LCA databases
2. Graph-based data structures for LCA databases