Application of GPT in Chatbots to Facilitate Knowledge Management System Interaction Using Langchain (Case Study: PT Softbless Solutions)

Abstract

Searching the knowledge base on the system running at PT Softbless Solutions has the disadvantage of only being able to use keywords. Search results are also often irrelevant to what employees want. With the widespread use of AI in chatbots, it can improve information searches that are more interactive for employees. In this research, the methodology used in system development was the RAD (Rapid Application Development), and for bot development was the RAG (Retrieval Augmented Generation) method, which makes it possible to develop LLM such as ChatGPT to use existing knowledge base data in finding the desired information. This RAG implementation uses the LangChain framework and ChromaDB database to store data in embedding form. In the final result, the chatbot can be accessed in web applications and Telegram bots, and it achieves a User Acceptance Test value of 96.4%, which shows that the system and bot that have been built have a high level of usability. This research is expected to improve employee interaction in searching the knowledge base and become a reference for similar research.