Quick start: sample model

Time to complete: ~10 minutes
You need: A free Heimdall account, signed in at heimdallapp.org

This is Path B in User journeys — explore the product without uploading data to Lake.

Before you start

• No CSV or Lake setup required
• Samples are for learning — use your own data for production

ML with Housing or Iris

1. Sidebar → Data Intelligence → Machine Learning (/ml).
2. Click New model (green training wizard opens).
3. Scroll past the Gold dataset picker → click Sample datasets.
4. Choose one:
   • Housing — regression; target column MEDV (median home value)
   • Iris — classification; target column Species
5. Confirm target column → enter a model name → Train.
6. Switch to the Training queue tab; open the model when training completes.
7. On the model page → Generate API key → copy the API Integration details.

Housing vs Iris

Housing predicts a number (regression). Iris predicts a category (classification). Both are classic ML teaching datasets.

Forecast with AirPassengers

1. Sidebar → Data Intelligence → Forecast (/forecast).
2. Start a new forecast build (opens the wizard).
3. Choose Sample dataset → AirPassengers.
4. Set Interval → Monthly → review the analysis preview.
5. Name the forecast → start training.
6. When complete, deploy with Generate API key on the forecast detail page.

What you just skipped

Samples bypass Lake and Gold publish. Your own CSV must follow Path A: Data → Bronze → Gold → ML.

Common mistakes

Mistake	Fix
Looking for a file upload in the wizard	Only Gold (your data) or samples — no arbitrary CSV upload
Gold list is empty	Expected for this quick start — use Sample datasets instead
Training stuck in queue	Refresh after a few minutes; check Platform status at /health

Next steps

• Deploy ML or deploy forecast
• Ready for real data? Quick start: your data
• User journeys — all three paths
• Monitoring — track calls after deploy

FAQ

No file upload in the wizard?
Correct for your own files. Production data goes through Lake → Gold. Samples are the exception.

Samples are for learning

Do not use built-in samples in production. Publish your own Gold for real projects.