Choosing a Model or Sign
Decide between the Model API and the Sign API, and pick a name.
A sign is one analysis target. A model is a named bundle of signs. These docs use pulse (General Wellness) as the primary model. The other bundles stay available.
Model API or Sign API
| Interface | Endpoint | When done, result contains |
Use when |
|---|---|---|---|
| Model API | POST /v2/models/{model_name}/analyze |
summary + signals[] + audio_quality + extended_metrics |
You want a named bundle and one aggregate routing decision (summary.recommended_action). |
| Sign API | POST /v2/signs/{sign_name}/analyze |
signal + audio_quality |
Your workflow acts on exactly one sign. |
Prefer a named bundle when one matches your population. Use the Sign API when you need exactly one sign.
How to choose a model
Start with pulse for general wellness or employee wellness. Specialized contexts (clinical behavioral health, substance use, cognition, sports, cardiometabolic, respiratory) have their own names.
If you serve more than one population, send a separate request with the most specific model for each group.
Eight bundles: pulse, haven, clarity, aria, apex, harbor, tide, breath. Domain, signs, and scenarios: Model Catalog. Standalone vs model-internal signs: Sign Catalog. Live composition: GET /v2/models on the Model API.
Evidence tiers
Each sign is Established, Emerging, or Investigational (published research depth, not a product tier). Bundles mix tiers — put clinical oversight in place when you route on Emerging or Investigational signs. Definitions: Sign Catalog — evidence tiers. As-is statement: Introduction.
After you choose
- Audio Requirements — formats, size, quality codes.
- Quickstart — submit your first recording.
- Interpreting Results — read
levelandrecommended_action. - Billing and Cost — the Model API and Sign API bill at different rates.
If no existing model fits your deployment context, contact support@amplifierhealth.com to discuss your requirements.