Ramadhan, Alif As'ad and Budianto, Fadhil Zaky (2026) Membuat Website Dashboard Company dan menerapkan Fitur AI Chatbot Internal PT BGR Logistik Indonesia Pusat. Project Report. [s.n.], [s.l.]. (Unpublished)
![[thumbnail of 5054231001_5054231007-Project_Report.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5054231001_5054231007-Project_Report.pdf - Accepted Version Restricted to Repository staff only Download (1MB) | Request a copy |
Abstract
Kerja Praktik ini dilaksanakan di PT BGR Logistik Indonesia selama tiga bulan. PT BGR Logistik Indonesia adalah perusahaan BUMN yang bergerak di bidang jasa logistik terintegrasi dan merupakan anak usaha dari PT Perusahaan Perdagangan Indonesia (PPI). Selama pelaksanaan Kerja Praktik, penulis ditempatkan di Divisi IT dan berkontribusi pada pengembangan BLINK (BGR Link) Dashboard Portal, sebuah platform internal terintegrasi yang dilengkapi chatbot AI bernama BILA (BGR Intelligence Life Assistance). Pengembangan platform ini dikerjakan oleh dua kontributor dengan pembagian peran yang jelas. Pada sisi kecerdasan buatan, dikembangkan chatbot BILA (BGR Logistik Assistance), yang dirancang untuk membantu karyawan PT BGR Logistik Indonesia dalam mendapatkan informasi terkait Standar Operasional Prosedur (SOP) dan Surat Keputusan Direksi (SKD) secara otomatis tanpa perlu mencari dokumen secara manual. Sistem ini dibangun menggunakan pendekatan Retrieval-Augmented Generation (RAG) yang menggabungkan kemampuan pencarian dokumen berbasis vektor dengan kemampuan generasi teks dari model bahasa besar Gemini milik Google. Komponen utama yang digunakan dalam pengembangan sistem ini antara lain ChromaDB sebagai penyimpanan vektor, PyMuPDF untuk ekstraksi teks dari dokumen PDF, model embedding google/embedding-gemma-300m, serta FastAPI sebagai backend API. Pada sisi antarmuka pengguna, dibangun seluruh frontend portal BLINK menggunakan React, TypeScript, Vite, TailwindCSS, dan Framer Motion. Hasil pengembangan mencakup dashboard utama dengan Bento Grid Layout, panel administrasi dokumen dengan fitur klasifikasi, pencarian dinamis, dan Status Toggle yang terhubung ke vector database secara real-time, serta widget chatbot BILA dengan streaming response karakter per karakter, render Markdown melalui react-markdown, dan tampilan sitasi dokumen sumber. Hasil pengujian menunjukkan bahwa seluruh fitur sistem BILA berhasil berjalan dengan baik sesuai yang diharapkan. Sistem mampu mengklasifikasikan intent pesan pengguna secara tepat, melakukan pencarian dokumen yang relevan dengan mempertimbangkan hak akses divisi masing-masing pengguna, serta menghasilkan jawaban yang akurat dalam Bahasa Indonesia beserta sitasi sumber dokumen dan nomor halaman. Pipeline ingestion juga berhasil memproses dokumen PDF hasil scan dan mendeteksi duplikat secara otomatis menggunakan mekanisme hash MD5. Antarmuka BLINK berjalan responsif pada berbagai ukuran layar dengan seluruh fitur interaksi dan animasi berfungsi sesuai spesifikasi.
====================================================================================================================================
This Internship (Kerja Praktik) was carried out at PT BGR Logistik Indonesia over a period of three months. PT BGR Logistik Indonesia is a state-owned enterprise (BUMN) engaged in integrated logistics services and is a subsidiary of PT Perusahaan Perdagangan Indonesia (PPI). During the internship, the author was placed in the IT Division and contributed to the development of the BLINK (BGR Link) Dashboard Portal, an integrated internal platform equipped with an AI chatbot named BILA (BGR Intelligence Life Assistance). The development of this platform was carried out by two contributors with a clear division of roles. On the artificial intelligence side, the BILA (BGR Logistik Assistance) chatbot was developed to help PT BGR Logistik Indonesia employees automatically obtain information related to Standard Operating Procedures (SOP) and Board of Directors Decrees (SKD) without needing to search for documents manually. This system was built using a Retrieval-Augmented Generation (RAG) approach, combining vector-based document retrieval with text generation capabilities from Google's Gemini large language model. The main components used in the development of this system include ChromaDB as the vector storage, PyMuPDF for text extraction from PDF documents, the google/embedding-gemma-300m embedding model, and FastAPI as the backend API. On the user interface side, the entire BLINK portal frontend was built using React, TypeScript, Vite, TailwindCSS, and Framer Motion. The development results include a main dashboard with a Bento Grid Layout, a document administration panel with classification features, dynamic search, and a Status Toggle connected in real-time to the vector database, as well as the BILA chatbot widget with character-by-character streaming responses, Markdown rendering via react-markdown, and source document citation display. Testing results show that all features of the BILA system successfully function as expected. The system is able to accurately classify user message intent, search for relevant documents while considering each user's divisional access rights, and generate accurate answers in Indonesian along with source document citations and page numbers. The ingestion pipeline also successfully processes scanned PDF documents and automatically detects duplicates using an MD5 hash mechanism. The BLINK interface runs responsively across various screen sizes with all interaction features and animations functioning according to specification.
Actions (login required)
 |
View Item |