Quick Start
Automat BI turns raw data into forecasts, reports, and automated actions. This guide takes you from an empty workspace to a published dashboard in about fifteen minutes using live public weather data — no API keys, external credentials, or command-line tools required.
Before you start
You need:
- An Automat BI account — the free tier is enough for this walkthrough
- Fifteen minutes
If you are still evaluating the platform, start with the product overview. If you want the concepts before the clicks, read Core concepts.
1. Create and run a pipeline
Pipelines are YAML. You declare where data comes from and how it should be shaped; the platform fetches, transforms, and previews it in the browser.
- Click Admin (the cog icon in the upper right-hand corner) and go to Data Pipelines.
- Click New to open the pipeline editor.
- Paste the following YAML into the editor:
- name: FetchMeteo
kind: tap
connection:
kind: Rest
client:
base_url: https://api.open-meteo.com
request:
endpoint: /v1/forecast
method: get
content_type: application/json
query:
latitude: 52.52
longitude: 13.41
hourly: temperature_2m,precipitation_probability
response:
content_type: application/json
locator: hourly
- name: DataToTable
kind: transform
show: -1
query: |
SELECT
CAST(UNNEST(time) AS TIMESTAMP) AS forecast_time,
CAST(UNNEST(temperature_2m) AS DOUBLE) AS temperature_c,
CAST(UNNEST(precipitation_probability) AS INT) AS precipitation_probability,
'InterestingLocation' AS location
FROM FetchMeteo
- name: OrderTable
kind: transform
show: -1
query: SELECT * FROM DataToTable ORDER BY forecast_time ASC
- Click Execute.
FetchMeteo pulls the 7-day hourly forecast from Open-Meteo, DataToTable unnests the parallel JSON arrays into typed columns, and OrderTable sorts the rows chronologically. You will see the resulting table preview directly in the editor.
(Optional: Change latitude, longitude, and 'InterestingLocation' to match your own city, or connect your own database, SaaS app, or CSV file in Data Integrations.)
2. Store the output
Once the preview looks right, add a sink step to write the transformed data to Parquet so it is registered in your catalog and ready to join with your existing datasets:
- name: SaveForecast
kind: sink
connection:
kind: parquet
locator: weather-forecast.parquet
Click Execute again to persist the table.
3. Build a report
- Open Report Designer and select the catalog table produced by your pipeline.
- Drag
temperature_candprecipitation_probabilityinto Values, andforecast_timeinto Rows. - Switch to a line chart to inspect the hourly trend, then save and share the report.
4. Publish a dashboard
Open Dashboard Designer, add your report as a card, and place the most important metric top-left. Text blocks are useful for stating the question the dashboard answers (for example: "What is the hourly temperature and rain outlook for the next 7 days?").
5. Put it on a schedule
In Data Automation, add a daily schedule so the forecast refreshes automatically. 06:00 is a reasonable default before the workday starts.
Troubleshooting
| Symptom | Likely cause |
|---|---|
FetchMeteo times out |
Outbound internet access is restricted in your workspace network settings |
Transform fails on UNNEST |
response.locator was changed — it must point to hourly where the parallel arrays live |
| Catalog table is missing in Report Designer | - Make sure a sink step is included and executed after OrderTable - If catalog gate is enabled then you may need to add the data source to catalog first - New data source may not appear immeditely in catalog. Go to Admin>Catalog and click Rescan Inventory. |
Tip: You can try the pipeline now using the embedded demo application. Just copy the pipeline above and paste to the demo app.
Next steps
- Forecasting engine — turn the same pipeline into a demand forecast
- Dashboards and reports — design guidance and export options
- Data automation — schedules, retries, and monitoring
- Blog — practical write-ups on inventory, sell-through, and forecasting