A connected database is a process, not a timestamp
Connecting a source starts a synchronization workflow. Existing selected rows must arrive, subsequent changes must be applied, and the retrieval structures needed by the application must be ready. Treat these as operational states to inspect.
Polygres synchronized projects use an initial snapshot followed by logical replication. Your source remains the system of record, and application writes belong there.
Use a small change to verify the path
Before relying on a large import, choose a test record whose identity and contents you know. Confirm the initial data appears as expected, then make a controlled update at the source and inspect the resulting project state.
This check helps distinguish a connection problem from a query problem. If the expected source record is not available in the project, changing the semantic query is unlikely to fix the underlying issue.
Plan for changes to the representation
A row update and an embedding update are not the same conceptual operation. Your data workflow needs to keep the representations used for retrieval aligned with the source content.
For synchronized projects, keep source writes and source embedding updates in the source system. Follow the current project documentation for readiness and supported update behavior instead of assuming that every derived representation changes automatically.
Make freshness part of evaluation
Record which source state and collection configuration were used for a retrieval test. When a result changes, compare the underlying records as well as the query settings.
For workflows where recent changes matter, define what the application should do while data is still becoming ready. A visible readiness state or a deliberate retry path is more useful than quietly presenting incomplete context as current.
Explore the next step
Continue with the platform guide. For the current setup and API contract, use the Polygres documentation.