Scheduled SQL reports
and alerts to Slack or email.
Run a query on a schedule or right after a pipeline finishes, check the result, and send it to the channels you name. Describe it in plain English; rsync.ai runs the steps you saved, exactly as written.
A workflow is a trigger followed by typed steps: query, condition, send, run a pipeline, rebuild a model, wait, or send a chart or dashboard. Ask rsync.ai drafts it, you test it without sending anything, and once it's on, the saved steps run with no model in the loop. It reports on your data and never moves it.
- Cron in your timezone, after a pipeline, or Run now
- Rows exist, no rows, or a value compared to a number
- Slack and email, with a quiet period against repeats
- No model is called when it runs
Describe it once. It runs as written.
Nothing is sent until you've seen the rows and the messages.
Describe it
Select New workflow and say what you want: "Every morning at 8, post yesterday's failed payments to #data-alerts, and flag it if there are more than 5,000." Ask rsync.ai drafts the trigger, query, conditions and messages.
Test it
Test runs the query and shows you the rows and every message as it would be sent. Nothing is sent and nothing is scheduled.
Turn it on
Turn on schedules it, and you become the user it runs as. Run now runs it once, as you, and really sends, even on a draft.
Watch the runs
Each run shows every step's outcome, the row count, which condition was true and whether each message was delivered, with links to the pipeline or model runs it started.
The report runs where the data lands
No cron box, no glue script, no second tool watching the first one.
Run on a clock, or when the data lands
Trigger a workflow on a five-field cron in your IANA timezone, after a chosen pipeline's run succeeds (or after each sync window of a CDC pipeline), or only when someone selects Run now. A failed pipeline run doesn't start it.
Typed steps, not code
A query step runs one read-only statement and returns at most 500 rows, with personal columns masked. A condition checks for rows, no rows, or a first-row value against a number; the first true condition sends to its own channels. Messages are templates with {{slot}} placeholders.
Chain the work around the report
Start a batch pipeline or rebuild a model and wait for it, pause for 1 minute to 24 hours, then report. A workflow without a query can send a saved chart, or a whole dashboard as a PDF.
Alerts that don't become noise
Each condition sends to 1–5 Slack channels or email lists you set up once in Workspace settings. A quiet period of up to 7 days stops the same alert repeating. Slack webhooks are stored encrypted and never shown again.
AI drafts it. The saved steps run as written.
An alert that asks a model what to do every morning is an alert you can't predict. In rsync.ai, Ask rsync.ai writes the first draft and you test it, then the saved steps run exactly as written, with no model called. Runs keep outcomes, not rows, and every run re-checks that the user it runs as is still an owner or admin and that the connection and SQL are still allowed; if not, it pauses and says why. Because it runs after the pipeline, the report fires when the sync finishes rather than at a time you guessed. Build the query in the Explorer, or send a whole dashboard on a schedule.
What workflows don't do yet
Workflows are built for reports and alerts on your own data. Here is where they stop.
- No Python, JavaScript or shell step. Anything smarter than a condition is written into the SQL.
- No approval step, outbound HTTP call, step that acts in another app, AI step, CSV attachment or Google Sheets delivery yet.
- A workflow never moves data; that's a pipeline. It can start a batch pipeline, but a CDC or polling pipeline runs until stopped, so it can't be a step.
- On the canvas you edit the trigger, query, conditions, messages and channels. Adding or removing a step goes through Ask rsync.ai.
- Owners and admins draft, test, turn on and edit workflows. Members and viewers see them and their runs.
- If you orchestrate custom code across many systems, keep Airflow or a similar scheduler for that, and use workflows for the SQL-and-message jobs that don't need it.
Scheduled reports and alerts — frequently asked
How is a workflow different from a pipeline?
Does AI run every time the report runs?
Where do the messages go?
Does rsync.ai keep the rows a report sends?
What happens if the person who turned it on loses access?
Can I run workflows self-hosted?
Know before someone asks why the number is wrong.
Start free on rsync.ai Cloud, sync a source, and have your first report in Slack tomorrow morning.
Live on rsync.ai Cloud today. On a self-hosted install, drafting a workflow needs Ask rsync.ai switched on with your own model key.