Implementing a local LLM with retrieval-augmented generation on a CI/CD-pipeline
Abstract
This thesis explores a Retrieval-Augmented Generation (RAG) system in a Continuous Integration/Continuous Delivery or Continuous Deployment (CI/CD) pipeline. The objective of the thesis was to design A RAG system to improve the performance of a Large Language Model (LLM) by retrieving relevant data from a knowledge base stored in vector databases and combining it with the user’s query. The system was designed to make the process of architecting a local LLM for a CI/CD pipeline faster and more secure when it is not passed through a repository. The technologies used were Python, LangChain, Docker, Ollama, Ansible, and Ragas. The result of this thesis was an evaluation of the accuracy of an LLM when used with and without RAG. The results showed that the RAG method enhanced the model’s accuracy and reduced hallucinations. The testing process involved different tokenization sizes, embedding models, LLM models, creating effective prompts, parsing data from multiple sources, and measuring how the system avoids hallucinations. These experiment results show that using RAG-based architecture makes more reliable, accurate, and scalable approach for integrating LLMs into CI/CD pipelines, helping to improve automation and future software development workflows.
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.