Corti Home
Corti Home

Theme switcher

Getting Started

The Corti Co-Pilot API offers a comprehensive solution for integrating advanced speech-to-text, live note taking, and summary document generation into voice-based interactions. The overall flow of the API involves several key steps, from session initiation to real-time communication and session management.

Overall Flow

  1. Initialization: The integration process begins with a POST request to /interview, which sets up a new session with unique authentication credentials and allows for the customization of AI features to suit specific medical consultation needs.
  2. WebSocket Communication: A WebSocket connection is then established using the session ID and token, enabling the streaming of audio data from the consultation. This step is essential for facilitating real-time interaction between the client and the server.
  3. Session Management: After the consultation, the session can be securely closed, or if necessary, it can be reopened within a 4-hour window to resume the interaction, ensuring flexibility and reliability in the documentation process.

Accessing the API

Access to the Corti Co-Pilot API, particularly the development environment, is granted exclusively upon request to ensure that each integration meets the high standards of security and functionality required for medical applications.

Your initial access will be upon request approval granted to you with access to an environment suitable to develop and test your integration, mimicking the production setup.

Access to the API on all environments is governed by OAuth 2.0. This authentication protocol offers enhanced security measures, ensuring that access to patient data and medical documentation is securely managed and compliant with healthcare regulations.

oAuth Method

While initially provided with a username and password, as your PoC advances, we can ensure you switch to the relevant authentication flow, whether a standard client credentials flow or authenticating individual users via SSO.

Moving to Production

Once your integration is ready to be deployed live and to handle real health care data, you will be set up with your production environment. At this point, you will have to change your API and Access Token baseURLs and switch to

Interviews

A POST request to /copilot/v1/interviews initiates a new session, generating a session ID and allowing configuration of AI settings, such as language preferences for transcription and documentation, as well as which template to use for the documentation.

Participants are used for multi-channel streams. If you're only sending a single channel stream, you can leave the participants array empty, otherwise you'd need to send a list of participants mapping each channel to the role of the participant in question. These participants will be referenced in incoming lines of transcript for multi-channel streams.

Real-Time Interview Stream

Languages

Language

Code (BCP-47)

English

en

French

fr

German

de

Spanish

es

Danish

da

Swedish

sv

Norwegian

no

Templates

Transcription Stream

For cases where there is a need for real time transcription, for example voice commands or dictation, the system provides this by using a websocket connection

Connecting to the WebSocket

Select...
GET baseUrl/audio-bridge/v1/transcribe?token=Bearer%20$token

Sending data to the server

The only message type the client needs to send is an audioBuffer. The API currently supports webM and m4a audio encoding, and performs best when audio chunks are between 150-250ms (and no more than 1000ms) in length.

TYPESCRIPT

typescript
Select...
// Setup the websocket

const websocket = new WebSocket("https://api.beta-eu.corti.app/audio-bridge/v1/transcribe?token=Bearer%20TOKEN_VALUE");
// Get the recording object

navigator.mediaDevices
  .getUserMedia({ audio: true })
  .then((stream) => {
    // Create a MediaRecorder from the stream
    const recorder = new MediaRecorder(stream);

    
    // Set the recorder to capture 250ms chunks of audio
    recorder.start(250);

    // When there is data available, send the data over the websocket
    recorder.addEventListener("dataavailable", (event) => {
      if (event.data.size > 0 && websocket.readyState == 1) {
        websocket.send(event.data);
      }
    });
});

Receiving Transcriptions

Transcription Response

typestring
eventstring
dataobject

Show child attributes

1 2 3 4 5 6 7 8 9 10 { "type": "string", "event": "string", "data": { "text": "string", "start": "timestamp", "end": "timestamp", "final": "boolean" } }