AI Therapy Sessions

Chat, voice, video — how sessions flow, conversation tokens, safety layer.

AI Therapy Sessions are the core of the therappai experience. Through a small set of API calls, your application can let users send messages and receive supportive, therapeutic responses in real time. Sessions can be delivered through chat, voice, or video depending on your integration needs.

How AI therapy works

At a high level, every session follows the same pattern:

  1. User sends a message

    • text, audio, or a mixture depending on the mode

  2. therappai processes the message

    • evaluates tone, intent, context, and wellbeing signals

  3. An AI therapist response is generated

    • grounded in CBT/DBT principles and supportive conversation design

  4. Response is returned to your application

    • as text, audio, or a rendered video/avatar clip

Your app chooses how to display it: chat bubbles, voice playback, video frame, etc.

Session Modes

1. Chat Therapy

The simplest and most flexible option. Your app sends a text message, and the API responds with a therapist reply.

Best for:

  • chat interfaces

  • onboarding conversations

  • lightweight support flows

  • low-bandwidth environments

2. Voice Therapy

Voice sessions accept an audio file and return a generated response based on the speech content.

Best for:

  • users who prefer to speak more naturally

  • hands-free interactions

  • accessibility use cases

Note: Your app handles recording and sending the audio file.

3. Video Therapy (AI Avatar)

For a more immersive experience, you can call the AI avatar endpoints to generate video responses. You supply:

  • a message

  • avatar ID

  • context/persona

  • session token

The API returns a video segment of the avatar delivering the therapist’s response.

Best for:

  • higher-engagement experiences

  • wellness apps

  • structured lesson or coaching flows

Session Context

Sessions are stateless per request — meaning you do not need to hold open a “session object” on the server.

Instead:

  • You can store conversation history on your end if your UI requires it.

  • The API can accept short context snippets to improve continuity.

  • For video therapy, you may need to request a “session token” before sending/receiving clips.

This keeps integration simple, predictable, and free of long-running connections.

Safety Layer (Important)

Every message processed by the API goes through a safety and wellbeing filter. This includes:

  • distress signals

  • crisis language

  • self-harm indicators

  • abusive or harmful content

If risk-related content appears, your app will receive a safe, supportive response rather than raw model output.

This layer is designed to protect your users and your platform and requires no setup on your side.

Important: therappai does not automatically contact emergency services. You decide what to do with risk-related signals or Crisis Buddy information.

Typical Integration Flow

Here’s how developers usually integrate AI therapy:

  1. Create/login user

  2. Get access token

  3. Display chat/voice/video UI

  4. When the user sends a message:

    • Send it to /chatting/ (text/voice)

    • Or use /liveavatar/... (video)

  5. Display the AI response

  6. Loop from step 4

You can combine this with:

  • content recommendations

  • mood check-ins

  • daily routines to build a complete wellbeing experience.

PreviousOverview: How therappai worksNextContent Library

Last updated