Xentree Live

Xentree Live brings real-time voice and avatar interaction to the Xentree platform.

It is designed for immersive AI sessions with usage controls, transcript persistence, and personalization.

Core capabilities

start and end voice or avatar sessions

monitor remaining monthly live quota

update a user nickname for more natural interactions

store full transcript history

connect to realtime audio and avatar infrastructure

Authentication and scopes

Most Live REST endpoints support both bearer and API key auth.

Recommended API key scopes:

live:read for quota and transcript reads

live:write for session start/end/heartbeat and nickname updates

REST endpoints

Quota and profile

GET /api/live/quota

PATCH /api/live/nickname

Session lifecycle

POST /api/live/session/start

POST /api/live/session/heartbeat

POST /api/live/session/end

Transcript storage

POST /api/live/transcripts

GET /api/live/transcripts

GET /api/live/transcripts/{transcript_id}

POST /api/live/transcripts is intended for bearer (and internal service-key flow). For general server-to-server access, prefer documented hybrid endpoints first.

Realtime architecture

For browser-based realtime interactions, Xentree Live uses a dedicated socket endpoint:

Socket endpoint: wss://orchstr.xentree.ai

Representative event types include:

start-session

end-session

audio-input

audio-response

session-started

session-ended

avatar-expression

Secure integration token lifecycle

For embeds, use subject-scoped integration grants via Integration Tokens (for example POST /api/integration/grant). Grants default to issuing user; organization scope requires explicit organizationId from an organization administrator.

Recommended flow:

Call POST /api/integration/grant with short TTL (for example 90-120 seconds).

Attach the returned token as integration_token in embed bootstrapping.

Validate using POST /api/integration/verify in backend middleware/gateway.

Use POST /api/integration/rotate to issue a new token and revoke old token in one call.

Use POST /api/integration/revoke for immediate emergency invalidation.

Operational notes:

Revocation is persisted by jti in database for immediate invalidation.

Public verification material is available at GET /api/integration/jwks and GET /api/integration/fingerprint (query params: subjectType, subjectId).

Endpoint-level rate limit headers are returned on grant/verify/rotate responses.

Example start request

{
  "session_type": "voice"
}

Example transcript payload

{
  "session_started_at": "2026-04-16T10:00:00Z",
  "session_ended_at": "2026-04-16T10:05:00Z",
  "transcripts": [
    {
      "role": "user",
      "text": "Hello Xentree",
      "timestamp": "2026-04-16T10:00:03Z"
    },
    {
      "role": "ai",
      "text": "Hi, how can I help you today?",
      "timestamp": "2026-04-16T10:00:05Z"
    }
  ]
}

Quota behavior

Live usage is tracked in seconds and is evaluated against the user’s plan entitlements. If the quota is exceeded, the API can return a limit-related error and the client should gracefully end the session.

Core capabilities#

Authentication and scopes#

REST endpoints#

Quota and profile#

Session lifecycle#

Transcript storage#

Realtime architecture#

Secure integration token lifecycle#

Example start request#

Example transcript payload#

Quota behavior#