Skip to main content

ML deployment

Time to complete: ~5 minutes
What you'll accomplish: Generate an API key and get integration details for a trained model.

Before you start
  • Model training complete (not pending in Training queue)
  • Signed in; model visible under Machine Learning in sidebar or at /ml/{id}

Steps​

1. Open your model​

  1. Sidebar → Machine Learning → click your model name (or open from Training queue).
  2. You are on /ml/{model-id} with metrics and tabs.

2. Save to inventory (if prompted)​

After training completes, click Save Model if the wizard prompts you — adds the model to your inventory list.

3. Generate API key​

  1. At the top of the model page, click Generate API key.
  2. Copy the key immediately — it cannot be retrieved later.
Save your API key

If you lose it, regenerate a new key. Old keys stop working when rotated.

4. Get integration details​

  1. Open the API Integration tab on the same page.
  2. Copy the predict endpoint URL, headers, and sample payloads.

→ API integration guide

Test from the app​

Use the in-app predict/test UI on the model page. Calls appear in the Usage section with channel Heimdall UI.

Monitor after deploy​

WhatWhere
Account-wide chartsSidebar → Usage → Data Intelligence
This model's request logModel page → Usage section
Platform outages/health

→ ML monitoring · Production monitoring

Common mistakes​

MistakeFix
No API key buttonTraining may still be pending — check Training queue
401 from your appUse model-specific key, not unstructured Read/Vision key
No usage dataDeploy key first; external calls need the key in headers

Next steps​