Rancang Bangun Bot Pe-review Kode Repository GitLab dengan Teknologi Large Language Model (LLM)

Muhammad, Rausyan Fikri (2026) Rancang Bangun Bot Pe-review Kode Repository GitLab dengan Teknologi Large Language Model (LLM). Project Report. [s.n.]. (Unpublished)

Text 5053231027-Project_Report.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Laporan ini membahas pengembangan bot GitLab yang bertindak sebagai pe-review kode dengan Teknologi Large Language Model (LLM) untuk Transisi Lingkungan Development, Staging, hingga Production pada GitLab. LLM digunakan untuk membuat rangkuman dari hasil log tiap fase dari pipeline Continuous Integration/Continuous Development (CI/CD). LLM yang digunakan adalah LLM model Qwen-3 yang di-self hosting oleh Lab Manajemen Cerdas Informasi (MCI) yang disambungkan menggunakan API dan skrip CI/CD juga menggunakan buatan Tim DevSecOps dar Lab MCI. Hasil evaluasi menunjukkan bahwa implementasi berhasil memenuhi semua kebutuhan fungsionalitas. Hasil pengujian juga menunjukkan bahwa aplikasi lolos 2 dari 2 skenario pengujian End-to-End dengan skenario pertama yaitu skenario normal dan skenario kedua ketika koneksi ke LLM tidak dapat dilakukan sehingga menggunakan fallback, dan 8 dari 8 unit testing yang berdasarkan kebutuhan fungsional.
===================================================================================================================================
This report discusses the development of a GitLab bot that acts as a code reviewer using Large Language Model (LLM) technology for transitioning from development, staging, to production environments in GitLab. The LLM is used to summarize the log results for each phase of the Continuous Integration/Continuous Development (CI/CD) pipeline. The LLM used is the Qwen-3 model, self-hosted by the Manajemen Cerdas Informasi (MCI) Lab, which is connected using APIs and CI/CD scripts created by the MCI Lab DevSecOps Team. The evaluation results show that the implementation successfully meets all functional requirements. The test results also show that the application passes two out of two End-to-End test scenarios: the first is the normal scenario and the second is when the connection to the LLM cannot be established, resulting in the use of a fallback, and eight out of eight unit tests based on functional requirements.

Item Type:	Monograph (Project Report)
Uncontrolled Keywords:	Bot GitLab, CI/CD, Large Language Model, Ollama
Subjects:	T Technology > T Technology (General) > T58.5 Information technology. IT--Auditing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Rausyan Fikri Muhammad
Date Deposited:	21 Jul 2026 04:27
Last Modified:	21 Jul 2026 04:27
URI:	http://repository.its.ac.id/id/eprint/135798

Actions (login required)

View Item