Getting started with ambient documentation

This tutorial walks you through the minimum steps to run your first ambient documentation session, from creating a subscription to retrieving the generated clinical note. Each step links to a topic with full details — use this page to understand the overall flow, and the linked topics as reference when you implement each step.

Prerequisites

Before you start, make sure you have:

Step 1: Create a subscription

Associate a provider with the ambient documentation agent by calling the CreateSubscription API operation. This gives you a subscriptionId that you use to start streaming sessions. For details, see Subscription management.

Step 2: Choose a clinical note template

Decide how the generated note should be structured. For most physical health encounters, the default managed template (HISTORY_AND_PHYSICAL) is a good starting point — you can switch to a different managed template or define a custom template later as your requirements become clearer. For details, see Clinical note templates.

Step 3: Prepare encounter context (optional)

If you have background information about the patient that won’t be spoken aloud during the visit — active problems, current medications, the reason for the visit — prepare it as encounter context. This step is optional, but it improves the accuracy of the generated note. For details, see Encounter context.

Step 4: Start a session and stream audio

Send a configuration event with your subscription, note template, and any encounter context, then stream audio over HTTP/2 or WebSocket for the duration of the conversation. Send audio in small chunks at close to real time. For details, see Streaming audio.

Step 5: End the session and retrieve outputs

Send an END_OF_SESSION control event when the conversation ends. The service writes the transcript, clinical documentation, and after-visit summary to your configured Amazon S3 bucket. For details, see Storage and outputs.

Next steps

Once your first session succeeds end-to-end: