RAG Chatbot Lab on Docker Desktop for Mac

Beginner Step by Step Guide

1. Lab overview

This lab builds a simple local chatbot for Jabatan Perkhidmatan Awam (JPA) using Docker Desktop on a Mac with 16 GB RAM. The chatbot will answer questions using one source document only, which is the uploaded JPA PDF titled Pekeliling Perkhidmatan Sumber Manusia: Kaedah Pelantikan Pegawai Dalam Perkhidmatan Awam dan Urusan-Urusan Lain Berkaitan Pelantikan, Versi 1.0 (2022).

The purpose of the PPSM document is to explain the rules, conditions and procedures for appointing officers in the public service and other appointment-related matters. It came into force on 1 January 2022 and is extended to the State Public Service, statutory bodies and local authorities, subject to their acceptance.

The document covers several major areas, including contract appointments, secondment and temporary transfer, other appointment-related matters, security vetting, Program Transformasi Minda and medical examination for appointment purposes.

2. What students will build

Students will build a local JPA chatbot that can:

• read the Pelantikan.pdf file
• split the document into smaller text chunks
• turn the chunks into embeddings
• store the embeddings in a vector database
• search the document when a question is asked
• send the relevant chunks to a lightweight model
• answer in formal Bahasa Melayu based only on the uploaded PPSM document

3. What students need before starting

Students need:

• a Mac with 16 GB RAM