ETD

• AI agents
• amazon web services.
• artificial intelligence
• knowledge bases
• large language models

The divergence between code repositories and documentation is a common problem in software development environments that reduces programming efficiency, confuses developers, and leads to a lack of effective communication between project stakeholders. This problem becomes more critical in frameworks where the code repository is updated frequently. This thesis proposes the creation of a system based on Large Language Models, Knowledge Bases, and an artificial intelligence agent that orchestrates the creation and updating processes in an architecture built using various cloud-based Amazon Web Services such as S3, Lambda, Bedrock, and a frontend interface built in Streamlit. This system allows the user to create documentation from scratch, as well as to update existing documentation by using commits pushed to the code repository. The results of running the system on a real repository show that the solution generates and updates documentation efficiently and accurately, making it a highly useful tool in code development scenarios.