JSON is a very popular data format for storing information. So far, retrieval-augmented pipelines have primarily focused on parsing/storing unstructured text data, and querying using embedding-based retrieval. But there is a need for designing a query interface that can better leverage the semi-structured “tree” nature of JSON data.
In this tutorial, we show you our JSONQueryEngine, a query engine designed to natively query JSON files. We can do this by compiling to JSONPath, a DSL for querying JSON documents
Add comment