Skip to main content
Skip table of contents

Agent Assist Operations Websocket API

Overview

This API enables real-time interaction between client applications and the Agent Assist backend over a WebSocket connection. It provides information to support live agent interactions during a conversation, such as sound amplitude, dialog state, user sentiment, task suggestions, and more.

  • Protocol: WebSocket (wss)

  • Authentication: Token-based (via session token)

  • Content-Type: application/json

  • Schema Format: AsyncAPI 3.0

Access Requirements

  • WebSocket client must connect to a valid wss:// endpoint provided by the deployment.

  • Connection must be authenticated using a valid token.

  • All messages are exchanged in JSON format.


Real-Time Server Messages

The following messages are sent from the server to the connected client:

Sound Info

Field

Type

Description

level

Integer

Current amplitude level of the speaker (0–2)

AI Info

Field

Type

Description

user_sentiment

String

POSITIVE, NEGATIVE, or NEUTRAL

user_emotions

String

Detected emotion label

user_anger_detected

Boolean

Boolean flag for anger detection

user_intents

String

Current user intents

agent_sentiment

String

Agent emotional state

agent_tasks

Array

List of key-value task objects

agent_remaining_tasks

Array

Remaining tasks

agent_next_task

String

Next task to perform

agent_suggested_prompt

String

Suggested response prompt

Dialog Info

Field

Type

Description

latestSeqId

Integer

Latest dialog sequence ID

utterances

Array

List of utterance steps

agentPartialUtterance

String

Current agent utterance (if not final)

userPartialUtterance

String

Current user utterance (if not final)

Dialog Diff

Field

Type

Description

seqId

Integer

Dialog step ID

participantType

Enum

USER or AGENT

utterance

String

Text spoken

startTime

Integer

Epoch start of speech

endTime

Integer

Epoch end of speech

isFinal

Boolean

Boolean flag for utterance finality

Call Info

Field

Type

Description

status

Enum

Call state (ACTIVE, CLOSED, etc.)

error_description

String

Optional error details

Verification Result

Field

Type

Description

label

String

Result label

status

String

Status value

risk_level

Enum

LOW_RISK, MEDIUM_RISK, HIGH_RISK

score

Number

Overall score

fraudster_id

String

ID of matched fraudster (if any)

blocklist_id

String

Blocklist entry ID

custom_agent_message

String

Agent message

custom_ui_message

String

UI message

score_llr

Number

Log-likelihood ratio score

score_nllr

Number

Normalized log-likelihood ratio score

speech_frames

Integer

Number of speech frames

list_size

Integer

Size of blocklist

match_count

Integer

Number of matches

Voice Info

Field

Type

Description

liveness_score.label

Enum

GENUINE, SPOOFED, SYNTHETIC, etc.

liveness_score.global_post

Number

Posterior score

satResult.label

Enum

OK_DATA, LOW_DATA, etc.

satResult.valid

Boolean

SAT validity

satResult.gender

String

Predicted gender

satResult.ageEstimate

Integer

Predicted age

Error Info

Field

Type

Description

message

String

Error summary

messageDetails

String

Detailed error explanation

Call Summary

Field

Type

Description

data

String

Call summary text

TTS Detection

Field

Type

Description

label

Enum

GENUINE or SPOOFED

global_post

Number

Detection confidence score

ANI Validation Results

Field

Type

Description

calledNumber

String

Phone number used in call

trustIndicator

Integer

Trust score (0–100)

Business Flow Info

Field

Type

Description

flow_instance_id

String

Unique instance ID

flow_instance_status

Enum

NOT_STARTED, RUNNING, COMPLETED, etc.

intent

String

Detected intent triggering the flow

entity_instances

Array

List of entity values and matched prompts

Business Flow Diff

Field

Type

Description

flow_instance_id

String

Flow ID

flow_instance_status

Enum

Updated flow status

entityId

String

Changed entity

entity_value

String

New value

Attached Data

Field

Type

Description

data

Array

Array of key-value maps of attached metadata

Client-Initiated Commands

The client may send commands via the following channel:

Field

Type

Description

command

Enum

One of GET_DIALOG, GET_BUSINESSFLOW_INSTANCES, etc.

parameters

Object

Parameters such as { "flow_instance_id": "..." } if applicable

Example Topics and Channels

Channel

Direction

Purpose

command_request

receive

Send client command

sound_info

send

Sound level updates

ai_info

send

AI analysis and task suggestions

dialog_info

send

Full dialog information

dialog_info_diff

send

Dialog deltas

call_info

send

Call status

verification_result

send

Voice verification result

voice_info

send

Voice analysis (liveness, age, etc.)

error_info

send

Error messages

summary_info

send

Post-call summary

tts_detection

send

TTS spoofing detection

ani_validation_results

send

Caller number validation

business_flow

send

Business flow state snapshot

business_flow_diff

send

Business flow change notification

attached_data

send

Custom metadata injection


JavaScript errors detected

Please note, these errors can depend on your browser setup.

If this problem persists, please contact our support.