![](/style/images/tocco.png)
Giovanna Di Benedetto
"Agent Code, James Code": AI-based code generation framework.
Rel. Riccardo Coppola. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Informatica (Computer Engineering), 2024
|
PDF (Tesi_di_laurea)
- Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (6MB) | Preview |
Abstract: |
In contemporary society, Artificial Intelligence (AI) permeates numerous facets of daily life, including healthcare, manufacturing, and education. However, the fields of computer science and computer engineering have traditionally emphasized enhancing efficiency over practical implementation in daily environments, focusing on areas such as algorithm optimization, network traffic management, and compiler optimization. This thesis addresses this gap by constructing a framework aimed at aiding developers in resolving GitHub issues through AI-driven solutions. By doing so, developers can interact with AI in a more straightforward and practical manner. Central to this framework is an Agent, serving as the interface between human requests and AI responses. Adapted from the open-source LangChain framework, this Agent leverages AI to analyze issues, generate code solutions, develop corresponding unit tests (if required), and execute necessary code modifications post-review. By interfacing directly with GitHub via its toolkit, the framework employs GitHub Actions for code building and GitHub Webhooks to trigger Agent operations. Notably, the framework's AI component relies on OpenAI's ChatGPT, renowned for its sophistication and versatility. Nevertheless, this framework is built with the aim of making it as customizable as possible to accommodate various AI models as they evolve. Since GitHub is the most well-known and widely used service for hosting code, the Agent's integration with GitHub ensures a seamless workflow for developers. This thesis not only presents a practical solution for streamlining issue resolution on GitHub but also underscores the evolving landscape of AI in computer science. As the AI domain continues to burgeon, this framework represents a dynamic avenue for future exploration and refinement. The project developed through the internship and analyzed in this thesis is prone for evolution in the near future, driven by the ongoing advancement of AI technology and the increasing incentive to study this growing branch of computer science. |
---|---|
Relatori: | Riccardo Coppola |
Anno accademico: | 2023/24 |
Tipo di pubblicazione: | Elettronica |
Numero di pagine: | 104 |
Soggetti: | |
Corso di laurea: | Corso di laurea magistrale in Ingegneria Informatica (Computer Engineering) |
Classe di laurea: | Nuovo ordinamento > Laurea magistrale > LM-32 - INGEGNERIA INFORMATICA |
Aziende collaboratrici: | Blue Reply Srl |
URI: | http://webthesis.biblio.polito.it/id/eprint/31756 |
![]() |
Modifica (riservato agli operatori) |