Hi HN, I work on FalkorDB, an open-source graph database.
We just launched FalkorDB as a Snowflake Native App. If you find yourself writing recursive CTEs and chains of self-joins to answer multi-hop relationship queries in Snowflake, this gives you Cypher instead. A variable-length path query that takes 80+ lines of SQL becomes about 3 lines of Cypher, and you run it from your Snowflake worksheet without moving data anywhere.
The way it works is you bind your tables through Snowflake's reference binding UI (scoped read access, no manual GRANT statements), and the FalkorDB engine runs on Snowflake Container Services with isolated compute, separate from your warehouse workloads. Data never leaves your Snowflake account.
We'd love to hear from anyone who has dealt with multi-hop traversals in Snowflake, whether you stuck with recursive CTEs, reached for dbt, or tried something else entirely.
Very nice project. Consider looking into a graph db with low latency for nice user experience for the question/answer portion of the project. For example: https://github.com/FalkorDB/falkordb