Quickstart — Real-time Ambient Scribe

Spin up a real-time ambient scribe with the AI Services clinical endpoints — diarized capture, in-encounter SOAP drafting, ICD-10 / CPT code suggestions, and EHR sync.

Architecture

Prerequisites

Medera Console account and a Developer API Key with write:phi and execute:agents scopes.

A microphone — see Microphone Devices.
An EHR target configured under Console → Integrations.

Authenticate

Set your API key as an env variable.

export MEDERA_API_KEY="<your-key>"

Start the clinical session

Open a clinical session — this anchors the transcript, note, and code suggestions.

import { MederaClient } from '@medera/sdk'; const client = new MederaClient({ apiKey: process.env.MEDERA_API_KEY! }); const { session_id } = await client.clinical.start({ session_type: 'clinical_visit', patient_id: 'pat_abc', }); 

from medera import MederaClient client = MederaClient(api_key=os.environ["MEDERA_API_KEY"]) session = client.clinical.start(session_type="clinical_visit", patient_id="pat_abc",) 

cURL curl -X POST https://api.medera.info/api/clinical/start \ -H "X-API-Key: $MEDERA_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "session_type": "clinical_visit", "patient_id": "pat_abc" }' 

Stream chunks

Send PCM audio chunks (16 kHz 16-bit mono recommended) as the clinician speaks.

// Encode PCM as base64 const audio_data = btoa(String.fromCharCode(...pcmBuffer)); const { transcript } = await client.clinical.chunk({ session_id, audio_data, }); 

import base64 audio_b64 = base64.b64encode(pcm_buffer).decode res = client.clinical.chunk(session_id=session.id, audio_data=audio_b64,) 

cURL curl -X POST https://api.medera.info/api/clinical/chunk \ -H "X-API-Key: $MEDERA_API_KEY" \ -d '{ "session_id": "ses_abc", "audio_data": "<base64 PCM>" }' 

Local Medera STT is the default with Medera STT as fallback. The response returns the accumulated transcript so far.

Stop the session

Finalize the transcript and trigger the initial clinical analysis.

const result = await client.clinical.stop({ session_id }); // result.transcription, result.clinical_data 

result = client.clinical.stop(session_id=session.id)

cURL curl -X POST https://api.medera.info/api/clinical/stop \ -H "X-API-Key: $MEDERA_API_KEY" \ -d '{ "session_id": "ses_abc" }' 

Generate the SOAP note

Use the enhanced generator for specialty templates and inline code suggestions.

const note = await client.therapy.generateSoapNoteEnhanced({ session_data: { session_id }, specialty: 'psychiatry', template_config: { format_version: 'v3', include_assessments: true, include_billing_codes: true, }, }); 

note = client.therapy.generate_soap_note_enhanced(session_data={"session_id": session.id}, specialty="psychiatry", template_config={ "format_version": "v3", "include_assessments": True, "include_billing_codes": True, },) 

cURL curl -X POST https://api.medera.info/api/therapy/generate-soap-note-enhanced \ -H "X-API-Key: $MEDERA_API_KEY" \ -d '{ "session_data": { "session_id": "ses_abc" }, "specialty": "psychiatry", "template_config": { "include_billing_codes": true } }' 

The response includes the structured SOAP plus a list of suggested ICD-10 / CPT codes anchored to transcript spans.

Sign and post to the EHR

After clinician sign-off, finalize the note. Medera posts the signed note and code set to the configured EHR target via the integration adapter.

curl -X POST https://api.medera.info/api/therapy/notes/{note_id}/finalize \ -H "X-API-Key: $MEDERA_API_KEY" 

Best practices -

Capture audio at 16 kHz 16-bit PCM mono. Higher rates work but waste bandwidth.

Use a dedicated lavalier or beamforming mic. See Microphone Devices.
Always require clinician sign-off before posting. Never auto-post to the EHR.
Use the transcript span IDs to render the documentation anchor next to every code suggestion.

What’s next

Visit Scribe Agent

Agent-level documentation.

Encounter Coding

How code suggestions work.

EHR Integrations

Connect Epic, Cerner, athenahealth.

STT — best practices

Audio quality recommendations.

​Architecture

​Prerequisites

​Best practices -

​What’s next