Your first workflow

Tutorial

Walk through building a real end-to-end workflow: read rows from PostgreSQL, transform them, and load into Snowflake. Same shape as most production integrations.

Prerequisites

This walkthrough assumes you've finished Quickstart (account + first connection + first run). It also assumes you have a Snowflake account; if you don't, the Library Project labs let you exercise the same flow against shipped sample data without external systems.

Step 1: Create the project

From Projects, click New Project, choose Starter Project (creates a Start, Source, Target, Decision, two Target template), and name it CRM Lead Sync (or similar). The starter graph is a good base; we'll customise it next.

Step 2: Register the source connection (PostgreSQL)

1. 1
   Connections page

   Click New Connection. Pick PostgreSQL.
2. 2
   Fill the basics

   Host, port (5432), database, user, password. Mark the connection scope (Project or Global).
3. 3
   Add an environment override

   Add an Environment Setting so DEV/QA/Prod can point at different databases without changing the connection.
4. 4
   Test Connection

   Verifies the credentials and lets the wizard save.
5. 5
   Save

   The connection appears in the list under the project's scope.

Step 3: Register the target connection (Snowflake)

Same wizard, pick Snowflake. Snowflake supports two auth modes:

• Password -- username + password + account + warehouse + database + schema + role.
• Keypair -- username + account + warehouse + database + schema + role + private key (+ optional passphrase). Recommended for service-account use.

Test Connection. Save.

Step 4: Wire the flow

On the canvas:

1. 1
   Source

   Configure with the Postgres connection. Pick the table or write a SELECT in script mode.
2. 2
   Transform

   Open the Map Fields modal. Drag source columns to target columns. Use expressions for derived fields, lookups, or coalesce.
3. 3
   Load

   Configure with the Snowflake connection. Pick the destination table. Choose mode: Append, Replace, or Merge by key.
4. 4
   Wire them

   Click and drag from each node's output port to the next node's input port. Source -> Transform -> Load.

Step 5: Quick Debug Run

Click Quick Debug Run (10 rows) in the canvas header. The runtime pulls 10 rows from Postgres, walks them through Transform, and dry-runs the Load. The right-rail Inspector shows per-node row counts and payload samples. Iterate on the mapping until the Load preview looks right.

Step 6: Deploy as a Job

Open Jobs, click New Job, pick the project + workflow + environment, set a cron schedule (e.g. hourly), and save. The Runtime Executions page picks up the next scheduled run.

Step 7: Watch it run

Open Monitor -> Runtime Executions. The latest run shows status, duration, per-node row counts, and logs. Click into a run to expand per-node detail; the payload sample shows the actual rows that moved.

Where to go next

• Configure Action for the full Source, Target, Script, and Notification reference.
• Transform (Map Fields) for advanced mapping (expressions, lookups, type coercion).
• Execution Hooks for pre/post-run database hooks (e.g. truncate before load).
• Runtime Executions for troubleshooting failed runs.