A RAG for Elixir in Elixir

This article by Joel Koch describes the development of a Retrieval Augmented Generation (RAG) system entirely in Elixir, overcoming previous dependencies on Python. The system uses components such as Elixir's language parser, Jina embeddings, and a local Chroma vector store, and runs machine learning models with Bumblebee. The piece provides step-by-step instructions for setting up the system, managing code ingestion, retrieval, and generation pipelines, and demonstrates a practical UI component using Phoenix Playground. The article concludes with a discussion on alternatives for running the RAG system on memory-limited machines and invites readers to engage with the author through GitHub discussions or social media.

Visit Original Article →