Gen AI Driven FAQ Chatbot Using Advanced RAG Architecture for Querying Annual Reports

In the rapidly evolving business landscape, stakeholders require timely and accurate access to financial and operational information from annual reports. However, the extensive and complex nature of these reports makes it difficult to efficiently extract key analytical trends, financial performance measures, and comparative insights. The traditional manual approach to analyse these documents is time-consuming, error prone, and inefficient. Stakeholders-including creditors, customers, suppliers, employees, government agencies, and industry analysts-rely on this data to make informed decisions. The absence of an automated mechanism to handle frequent queries regarding financial performance, operational metrics, and sustainability targets impedes large-scale informed decision making. This paper presents a Gen AI-driven solution that leverages Natural Language Processing (NLP), LangChain Framework and Retrieval Augmented Generation (RAG) Architecture to efficiently extract and analyze financial annual reports and its data. The approach involves utilizing PyPDF2 for data extraction, segmenting content using RecursiveCharacterTextSplitter, generating embeddings via FastEmbedEmbeddings module, and storing information in Qdrant for rapid similarity search. The response generation is powered by DeepSeek LLM, deployed locally via Ollama, ensuring privacy and computational efficiency. A Gradio-based UI is implemented for an intuitive interface. This system enhances financial data accessibility, fostering transparency, compliance, and data-driven decision-making.