ADAPTIVE RETRIEVAL-AUGMENTED GENERATION FOR DYNAMIC KNOWLEDGE BASES IN CLOUD-NATIVE AGENTIC SYSTEMS

Enterprise knowledge bases are inherently dynamic, with continuous updates, additions, and modifications that challenge traditional Retrieval-Augmented Generation (RAG) systems. This paper presents a novel adaptive RAG framework specifically designed for cloud-native agentic systems that can dynamically update their knowledge representations in real-time. Our approach integrates continuous data monitoring, incremental indexing, adaptive embedding refinement, and update-aware retrieval logic within a scalable cloud architecture. The proposed system demonstrates significant improvements in knowledge freshness (94.2% vs 67.8% for static systems), retrieval accuracy (87.3% vs 72.1%), and response relevance (91.5% vs 78.4%) when evaluated on dynamic enterprise datasets. The framework's cloud-native design ensures horizontal scalability while maintaining sub-second response times even with frequent knowledge base updates. This research contributes to the advancement of enterprise AI systems by addressing the critical challenge of maintaining knowledge currency in rapidly evolving organizational environments.