PandaDB: An AI-Native Graph Database for Unified Managing Structured and Unstructured Data

In many applications, data are organized as graphs (e.g., social network and smart city). There could be unstructured data on such a graph, for example, the users’ avatars and images included in a post. It is natural to think of these unstructured data as attributes of nodes or relationships. Then the users would tend to query the semantic information of unstructured data on the graph, namely hybrid queries. To meet the demand of hybrid queries, this paper introduces PandaDB, an AI-native graph database, and it has the following characteristics: (1) Unified management of unstructured data and graph data. (2) Online extracting and indexing semantic information of unstructured data. (3) Optimization of hybrid queries. The system and its concept have been verified by multiple applications based on it. Users could deploy PandaDB to support hybrid queries and data mining.