AI-Enabled Annexure Parsing for Oil & Gas Project Exhibits

Abstract In the Oil & Gas industry, the effective management of capital project exhibit documents is essential for ensuring compliance, tracking progress, and making well-informed decisions. This paper presents a comprehensive methodology for developing a Large Language Model – Retrieval Augmented Generation (LLM-RAG) framework leveraging LangChain that enhances the accessibility and usability of these documents using advanced machine learning (ML) techniques. A Llama2-RAG model that incorporates Llama-2-70b large language AI model for advanced language understanding and generation is proposed in this paper. Project exhibit documents are uploaded via user-friendly Streamlit application, and the text is extracted and pre-processed to smaller chunks that are converted to vector embeddings and indexed in FAISS database for efficient retrieval. The integration of FAISS with LangChain creates a robust pipeline that retrieves relevant document sections and generates contextually relevant responses to user queries. One of the primary issues addressed by this study is information overload, where the sheer volume of documents makes it challenging to find specific information promptly. The Llama2-RAG model alleviates this by providing precise and contextually relevant responses, reducing the time and effort required to sift through extensive documentation. Additionally, the study tackles the problem of version control, ensuring that stakeholders always have access to the most current information. The Llama2-RAG model also addresses the technical complexity of the documents, making them more accessible to non-technical stakeholders. By translating complex technical language into simpler terms, the model improves understanding and communication among project teams. The system is also designed to manage project-specific terminologies and document structures, offering a high degree of customization to meet the unique needs of each project. The study also highlights the development of the Streamlit application, which not only facilitates document uploads but also provides an intuitive interface for user interaction. The integration of advanced AI and machine learning techniques ensures that this system remains adaptable and scalable, providing continuous improvements in accuracy and relevance over time. By addressing key challenges such as information overload, version control, cross-referencing, and technical complexity, the Llama2-RAG model promises to enhance the overall effectiveness of project document management, benefiting all stakeholders involved.