Copying source tables into Neuralift
Browse the tables a connection exposes and copy the ones you want into Neuralift as source tables.
Once a connection is Active, browse the tables it exposes and copy the ones you want to work from. Copied tables become source tables: point-in-time snapshots inside Neuralift that data preparation turns into run-ready datasets. A source table does not stay in sync with your warehouse; copy again for a fresh snapshot. Nothing is read into Neuralift until you copy it.
Before you begin
- You need the admin role in your organization; see Roles & permissions.
- You need at least one Active data connection.
Browse a connection’s tables
- Go to Settings → Data Connections.
- On an Active connection card, click Browse tables. The Source tables sheet opens for that connection.
- Click Sync tables to discover the tables currently available upstream. The button shows Syncing… while discovery runs.
Each discovered table appears as a row with its schema.table name, an Upstream status, a Copy status, the Destination it was copied to, and when it was Last copied. You can sort the columns and filter by schema.table name.
| Upstream status | What it means |
|---|---|
| Available | The table exists upstream and can be copied. |
| Missing | A previously discovered table no longer appears upstream. |
| Sync failed | Neuralift couldn’t check this table on the last sync. |
Copy a table into Neuralift
- In the Source tables sheet, find the table you want.
- Click Copy. The row shows Copying…, then Copied when the snapshot is complete, along with its destination and the copy time. If a copy fails, the row shows Failed and a Retry copy action.
The note at the top of the sheet shows where copied tables are written inside your organization’s workspace. Once copied, a source table can be selected during data preparation when Neuralift prepares a use case’s dataset.
Keep tables up to date
A copied table is a snapshot; it doesn’t update on its own.
- Click Copy again on a copied table to refresh it with the latest upstream data.
- Click Sync tables again whenever tables have been added or removed upstream.
- A table marked Missing can be removed with Remove; it reappears if a future sync finds it again.
Tip. Refresh a source table with Copy again before Neuralift prepares or re-runs a use case, so results reflect your latest data.
Copy files delivered over sFTP
Organizations that use Neuralift’s sFTP gateway can also bring data in as files, without a warehouse connection. The gateway is enabled per organization. If you don’t see an sFTP delivery section under Settings → Data Connections, contact your Neuralift representative.
- Drop each dataset as Parquet files in its own folder under
inbound/in your gateway home directory. - In Settings → Data Connections, under sFTP delivery (the gateway must show Active), find Inbound datasets and click Browse inbound files. Each dropped folder is listed with its Parquet file count. Click Refresh to re-list after new drops.
- Click Copy to source table next to a dataset. It shows Copied when done; use Re-copy to refresh it later, or Retry if the copy fails.
Troubleshooting
- A table you expect isn’t listed after Sync tables. The connection can only see what it was granted: check the share contents (Databricks) or the connecting user’s read access (Snowflake, BigQuery, Redshift), then sync again.
- A copy shows Failed. Click Retry copy. If it keeps failing, use Test on the connection card to confirm the connection is still healthy, and check that the table still exists upstream.
- Rows show Missing after a sync. The tables were renamed, moved, or dropped upstream. Restore or re-share them, or Remove the stale rows.
Next steps
- See how a source table becomes a run-ready dataset in the data preparation overview.
- Ready to frame the work? Start creating a use case.