Best way to move more ephemeral data for specific queries into KuzuDB queries?

[SevorisDoe]

Hey community,

This question comes from me researching how to leverage KuzuDB for specific kinds of graph analysis in a personal knowledge graph application. The catch here is that some of the context for the query has to come from outside of the KuzuDB database itself, because I can't store everything in it (and resolve every bit of query logic I am thinking about in it). So I need a way to include a "selection table" into the query, that contains the nodes that I have filtered out as relevant outside of Kuzu - and I am unclear on how to do this best in particular in languages outside of Python.

In Python I have before used the ability to load and scan in-memory Arrow tables to just include the relevant data into a table, run the query(or queries), then remove the data from the table again. From DuckDB have learned a similar trick where I used a substitution scan and a join to filter down relevant data with in-memory context.

From Neo4Js cypher implementation I am familiar with the use of UNWIND to introduce exterior context or previous query results into the next steps of a query, but I'm unsure if/how this would work in KuzuDB at all, and particular outside of the Python library. if I want to temporarily create nodes, can I pass a vec-of-hashmap to a prepared statement and that can be unwound as creating nodes and setting properties of them? And even if so, how well does that scale?

Somewhat related - does copying from Arrow tables through prepared statements also work with pointers to Rust Arrow tables? Or is this purely a capability of the Python library? (because if it works for everything that would add significant power to loading data into and out of KuzuDB from other parts of the application and integrating it)