A branch is part of the source identity
When you connect a Neon database to Polygres, be explicit about the branch and database your retrieval project represents. A connection string can look similar while pointing to a different dataset or environment.
Name the source and project so an engineer can distinguish development, evaluation, and production data without opening credentials. This reduces confusion when reviewing query results or changing a source configuration.
Prepare a direct replication connection
Use the direct, unpooled endpoint for the intended source. Confirm the current logical replication configuration and source user permissions using the provider-specific Polygres guide.
Select eligible tables with stable identities. Begin with a narrow scope that supports a real retrieval question, then expand when you can explain the operational and capacity cost of the additional data.
Separate experiments from production evidence
A retrieval experiment may use a different branch or a curated dataset. Keep the identity of that source with your evaluation results. Otherwise, a successful test can be mistakenly attributed to data or configuration that production does not use.
Record the collection configuration and question set as well. Retrieval quality depends on the source content, the representation, and the query, so changing several of them together makes a regression harder to diagnose.
Check readiness after the snapshot
Polygres starts synchronization with an initial copy and follows source changes through logical replication. Confirm project readiness before treating its results as representative of the selected branch.
Keep writes and schema changes in the source database. If you plan a branch or schema change, review the connection and synchronization implications before switching the application’s retrieval traffic.
Explore the next step
Continue with the platform guide. For the current setup and API contract, use the Polygres documentation.