Towards personal data store-based retrieval augmented generation in mobile computing

<p>Conversational AI agents have the potential to dramatically improve human ability to interact with the digital world. However, this is currently limited by data sharing between disparate application domains. A decentralized data exchange based on personal data stores can exist for applications to enable knowledge integration across these domains. Knowledge graphs and vector embeddings are critical to scale this behavior. In this work, we create and evaluate a prototype mobile-based retrieval augmented generation architecture based on personal data stores.</p>