Exploring exaptations in engineering practices within a RAG-Based application
In this article, I will explore the concept of RAG, but not in the typical way. My aim is to essentially create a RAG from the beginning to view it as a purely engineering problem.
Note: The from scratch approach is difficult because the generation of the embedding is linked to the model and tokenization, but let’s consider it as from scratch for the Engineering part, which will be sufficient for me.
As a bootstrap, I used the information from this article because it’s clear and written in Go, and I am fluent in that language. I have no additional insights to offer beyond the original article on the technical part (the author made a great job).
In this article, I will write the step by step method I used to write a simple (and non-efficient nor effective) RAG, but I will also note the discoveries that may be useful for my work as a consultant and engineer.
In the introductory section, I outlined the anticipated outcome I am aiming for. This result revolves around discovering the partial answers to the question: “What is the engineering role in the setup of an application powered by AI”. To effectively steer my actions towards this goal, I require a use-case. This use-case should have a clearly defined output that signifies the conclusion of this experiment.
