Question 1

What kinds of data end up in Supabase?

Accepted Answer

Generally anything the application needs but should not compute while a request is waiting. Scores, aggregates, entitlements, recommendation sets and precomputed views all fit that description, and there are plenty of others. What they have in common is being expensive or awkward to derive inside an application database and straightforward to model over lakehouse data. The result arrives as an ordinary table the app reads through the client library it already uses.

Question 2

Is Supabase different from a plain Postgres target?

Accepted Answer

Only at the connection. Supabase uses the same PostgreSQL implementation rather than a separate writer, so write modes, identity matching, batching and schema creation behave identically. What differs is SSL and the postgres database by default, and simple protocol when connecting on the transaction pooler port. Both are defaults rather than settings you manage, and row level security applies to the synced table as to any other.

Question 3

Can I use the pooled connection string?

Accepted Answer

For writing, yes. The transaction pooler port is handled with simple protocol so write syncs work through it. The restriction people run into applies to logical replication rather than writes, since a pooled connection cannot serve the long lived replication protocol, so reading change data out of Supabase later would need the direct connection. For pushing modelled data in from a lakehouse, the pooler is fine.

Question 4

How should nested values be handled?

Accepted Answer

Decide in the model rather than at the mapping. STRUCT and ARRAY retain their nested schema where it can be parsed and MAP is treated as a generic JSON shape, but an application table is usually easiest to work with when those have been resolved into scalar columns first. Where the application genuinely wants a nested value, a JSON column read with Postgres JSON operators works well. One thing to watch is that TIMESTAMP is timezone aware while TIMESTAMP_NTZ is wall clock time without one.

Question 5

How do I sync Databricks to Supabase?

Accepted Answer

Connect Databricks with a SQL warehouse hostname and a token, or an OAuth service principal, and connect Supabase with your project's database credentials. Either the direct connection or the transaction pooler works for writing. Then build a model over the data your application needs, map it onto a Supabase table, and choose a write mode and identity field. The table can be created for you if it does not exist.

Connect Databricks with Supabase

Connect Databricks and Supabase with Polytomic

Straight into the app database

Postgres write behaviour

No primary key required

Trusted by businesses worldwide

Step-by-step integration guide

Connect Databricks in Polytomic

Connect Supabase in Polytomic

Create a sync between Databricks and Supabase

Map fields and enable the sync

Frequently asked questions

The sync and ETL platform that does it all

Powerful data sync

Advanced analytics

Enterprise security

Easy setup

Ready to connect Databricks with Supabase?