Realtime

guided

This PowerLink runs in guided mode — it launches a real-time session with a human or AI agent.

The Realtime PowerLink launches a live audio, video or chat session between the customer and an agent. It supports AI agents (via LiveKit) or human agents from your workforce. You can customize language, input identity and session behavior before starting the interaction.

Key Features

🎙️ Live Audio or Video

Launch real-time calls powered by LiveKit (AI or human).

🤖 Supports AI Agents

Use AI agents to handle live sessions with integrated STT/LLM.

📞 Human Agent Routing

Connect the session to a waiting human agent (optional).

🌍 Language & Identity Control

Configure language, identity and session metadata dynamically.

How to Install

Go to the Store tab in Agentix cPanel, find Realtime and click Install to add it to your PowerLinks list.

How it Works

When this PowerLink is triggered, it prepares session metadata, configures connection identity and launches a live session with either an AI agent (default) or a routed human agent. The session can include audio, video or chat and runs on LiveKit infrastructure.

You can control the launch behavior using input parameters like LANGUAGE, CALL_TYPE or TELEPHONY_WEB_PANEL_IDENTITY. The session is launched immediately after the task starts.

This PowerLink is meant to be triggered from LaunchPad or via API. It’s ideal for workflows where customer input must be handled live in real time.

How it Looks for Customers

This is how the real-time session interface appears to customers when triggered by a LaunchPad or invitation link. The layout works across desktop and mobile, providing an intuitive experience for video or audio interactions.

Key Interface Components

Left Panel (Customer View):
Displays the customer’s avatar or video feed (if camera is enabled). The name and language are shown at the bottom left.
Right Panel (Agent View):
Shows the AI or human agent. When using AI, the agent displays a branding visual along with transcribed responses as subtitles.
Bottom Toolbar:
Customers can mute/unmute their microphone, turn camera on/off, share screen, open chat, adjust settings or leave the session.

The AI or human agent handles the entire conversation. No manual submission or rejection is required from the customer side.

Sessions may be voice-only or video-enabled depending on your workflow configuration and assigned agent capabilities.

This interface also supports multiparty conferences. When additional participants join, the layout automatically adjusts to accommodate them.

Configuration Parameters

These parameters control the behavior of the Realtime PowerLink at the PowerLink level, such as call recording, pre-join prompts and outbound dialer settings.

Show Configuration Parameters

USER_PROMPT

TEXT

Disclaimer text to display prior to joining the call. Supports basic HTML.

GUEST_INVITE

LIST

default:"ON"

Allow users to generate invitation links for guest users.

Value	Description
`ON`	Enable guest invitation link
`OFF`	Disable guest invitation link

RECORDER

LIST

default:"OFF"

Enable or disable recording of the call session.

Value	Description
`ON`	Enable call recording
`OFF`	Disable call recording

RECORDER_LAYOUT

LIST

default:"grid"

Specify the layout for recorded output.

Value	Description
`grid`	All participants shown equally
`speaker`	Focuses on the active speaker
`single`	Only one participant shown at once

ALLOWED_TASK_ACTIONS

MULTI

Select which task actions to show or hide for the user (not the customer). The selected actions will appear in the user’s task actions menu.

Task Action	Description
`getUserLocation`	Requests the user’s geolocation
`mediaUpload:Personal ID`	Upload of a personal identification document
`mediaUpload:Passport`	Upload of passport image
`mediaUpload:Selfie`	Capture and upload of a selfie
`mediaUpload:Utiltiy Bill`	Upload of a utility bill
`mediaUpload`	General media upload (image or document)

SOURCE_LANGUAGE

MULTI

default:"en-US"

List of languages supported in the session. Displayed to the user during pre-join. Default language is always the first one in the list order.

You can add any language locale later and select it in the agent assignement.

START_MODE

LIST

default:"audio-video"

Specifies the initial mode for the session.

Mode	Description
`audio-video`	Start with audio/video
`chat`	Start with chat only

HIDE_LOCAL_TITLE

LIST

default:"OFF"

Hide or show the local participant’s title in the call interface.

Value	Description
`ON`	Hide local participant title
`OFF`	Show local participant title

OUTBOUND_TRUNK_CONFIG

TEXT

Trunk configuration for outbound dialer agent. Use this format:
{ TRUNK_NAME=telnyx-outbound-trunk, SIP_ADDRESS=sip.telnyx.com, AUTH_USERNAME=username, AUTH_PASSWORD=password }

OUTGOING_CALLING_NUMBER

TEXT

Phone number to use for outbound calls.

ACTION_PANEL

LIST

default:"ON"

Enable/Disable action panel for the user.

Value	Description
`ON`	Enable action panel
`OFF`	Disable action panel

VISUALIZER_BACKGROUND_IMAGE

MEDIA

Background image for AI agent during the conversation.

VISUALIZER_TYPE

LIST

default:"3D-SPHERE"

Select the visualizer type for AI agent during the conversation.

Value	Description	Preview
`BAR`	Bar visualizer with audio wave patterns
`3D-SPHERE`	3D sphere visualizer with dynamic animations

VISUALIZER_CONFIG

TEXT

Configuration options for the AI agent visualizer. This JSON object controls the appearance and behavior of the visualizer during different conversation states.Configuration Structure:

{
  "width": 450,              // Width of the visualizer in pixels
  "height": 450,             // Height of the visualizer in pixels
  "state_label": "ON",       // Display state labels: "ON" or "OFF"
  "state_colors": {
    "disconnected": "#333333", // Color when agent is disconnected (dark gray)
    "connecting": "#FFAA00",   // Color when establishing connection (amber)
    "initializing": "#00AAFF", // Color during initialization (bright blue)
    "listening": "#00FF88",    // Color when agent is listening (green)
    "thinking": "#AA00FF",     // Color when agent is processing (purple)
    "speaking": "#88DDFF"      // Color when agent is speaking (light blue)
  }
}

Parameter Details:

width & height: Control the size of the visualizer canvas (recommended: 450x450)
state_label: Shows current state text below visualizer when set to “ON”
state_colors: Hex color codes for different conversation states
- disconnected: No active connection
- connecting: Establishing connection to agent
- initializing: Loading agent and preparing session
- listening: Agent is actively listening to user input
- thinking: Agent is processing user request
- speaking: Agent is providing response

Use standard hex color codes (e.g., #FF0000 for red). Colors should provide good contrast and be accessible for all users.

QUEUE_WAITING_MESSAGE

TEXT

Message to display while waiting in the queue.

Agents

Here is list of available agents that can be used with this PowerLink. Each agent has its own set of parameters that can be configured to suit your use-case.

Multimodal Realtime Agent
AI Agent for video/audio calls using multimodal realtime LLMs (OpenAI GPT or Google Gemini) that support audio input and output in real time.

Show Agent Parameters

LLM

LIST

default:"openai:gpt-4o-mini-realtime-preview"

Select the LLM provider and model for the assistant.

Provider	Model
OpenAI	`gpt-4o-mini-realtime-preview-2024-12-17`
OpenAI	`gpt-4o-realtime-preview-2024-12-17`
OpenAI	`gpt-4o-realtime-preview`
OpenAI	`gpt-4o-mini-realtime-preview`
Google	`gemini-2.0-flash-exp`
Google	`gemini-2.0-flash-live-001`
Google	`gemini-2.5-flash-preview-native-audio-dialog`
Google	`gemini-2.5-flash-exp-native-audio-thinking-dialog`

TEMPERATURE

TEXT

default:"0.8"

Temperature control from 0-1 (Default: 0.8). Minimum allowed temperature is 0.6.

VOICE

LIST

default:"openai:alloy"

Select the voice for the assistant.

Provider	Voice Options
OpenAI	`alloy`, `ash`, `ballad`, `coral`, `echo`, `sage`, `shimmer`, `verse`
Google	`Puck`, `Charon`, `Kore`, `Fenrir`, `Aoede`, `Leda`, `Orus`, `Zephyr`

PRONUNCIATIONS

TEXT

Add custom pronunciation rules for names and terms. Use multi-line format with trailing semicolon.Format: language_code|name=pronunciation;Examples:

en|1=One;
ar|2=اثنان;
fr|3=trois;

VERBALIZE_NUMBERS

LIST

default:"OFF"

Control whether numbers in RAG content are verbalized during conversation.

Value	Description
`ON`	Verbalize numbers in conversation
`OFF`	Do not verbalize numbers

USER_RESPONSE_TIMEOUT

TEXT

default:"30"

Timeout for user response in seconds.

MAX_REMINDERS

TEXT

default:"3"

Maximum number of reminders to send to the user before considering hangup.

HANDOFF

LIST

default:"OFF"

Select the handoff provider for the assistant. Disable by selecting OFF.

Value	Description
`OFF`	No handoff capability
`NOVOMIND`	Integration with Novomind platform

HANDOFF_PLUGIN_URL

TEXT

Custom handoff plugin URL for the assistant.

HANDOFF_LIVE_ASSIST_AGENT

LIST

default:"OFF"

Live assist agent to dispatch call upon successful handoff.

Value	Description
`OFF`	No live assist integration
`agent-live-assist`	Standard live assist agent
`live-assist-agent`	Alternative live assist agent

WHATSAPP_NOTIFICATION

LIST

default:"OFF"

Select WhatsApp notification provider for the assistant.

Value	Description
`OFF`	No WhatsApp notifications
`NOVOMIND`	WhatsApp integration via Novomind

EMAIL_NOTIFICATION

LIST

default:"OFF"

Select Email notification provider for the assistant.

Value	Description
`OFF`	No email notifications

BACKGROUND_AUDIO

MULTI

Select background audio options for the conversation environment.

Option	Description
`OFFICE_AMBIENCE`	Office ambient sounds
`music01.ogg`	Background music track

BACKGROUND_AUDIO_VOLUME

TEXT

default:"0.5"

Volume level for background audio (0-1 range, e.g., 0.5 for 50%).

THINKING_SOUND

MULTI

default:"KEYBOARD_TYPING,KEYBOARD_TYPING2"

Select sound effects to play while the agent is in thinking state.

Option	Description
`KEYBOARD_TYPING`	Keyboard typing sound effect
`KEYBOARD_TYPING2`	Alternative keyboard typing sound
`bell.ogg`	Bell notification sound
`screenshot.ogg`	Screenshot capture sound

THINKING_SOUND_VOLUME

TEXT

default:"0.8"

Volume level for thinking sounds (0-1 range, e.g., 0.8 for 80%).

AVATAR

LIST

default:"OFF"

Select the avatar for the assistant. Choose OFF to disable or select a provider.

Avatar	Configuration
`OFF`	No avatar
`Sandra`	`tavus:name=Sandra\|replica=r9fa0878977a\|persona=pc074769bad9`
`Steph`	`tavus:name=Steph\|replica=r9c55f9312fb\|persona=pc074769bad9`
`James`	`tavus:name=James\|replica=r70c81a0519b\|persona=pc074769bad9`

TELEPHONY_WEB_PANEL

LIST

default:"OFF"

Enable telephony web panel integration. When enabled, the AI agent will send notifications to users to open the web panel.

Value	Description
`ON`	Enable telephony web panel
`OFF`	Disable telephony web panel

Voice Pipeline Realtime Agent
AI Agent for video/audio calls using a voice pipeline supporting different providers — voice activity detection, speech-to-text (STT), LLM processing and text-to-speech (TTS).

Show Agent Parameters

STT

LIST

default:"openai:whisper-1"

Select the Speech-to-Text provider and model.

Provider	Model
OpenAI	`whisper-1`
Google	`latest_long`
Google	`nova-2`
Groq	`whisper-large-v3`
Deepgram	`nova-2`
Azure	`whisper`
Assembly AI	`best`
AWS	`transcribe`

LLM

LIST

default:"openai:gpt-4o"

Select the LLM provider and model for the assistant.

Provider	Model
OpenAI	`gpt-4o`
OpenAI	`gpt-4o-mini`
OpenAI	`o1`
OpenAI	`o1-mini`
OpenAI	`o1-preview`
Google	`gemini-1.5-flash`
Google	`gemini-1.5-pro`
Google	`gemini-exp-1206`
Groq	`llama-3.3-70b-versatile`
Groq	`llama-3.2-90b-text-preview`
Claude	`claude-3-5-sonnet-20241022`
Mistral	`mistral-large-2407`
xAI	`grok-beta`

TTS

LIST

default:"openai:tts-1"

Select the Text-to-Speech provider and model.

Provider	Model	Voice Options
OpenAI	`tts-1`	`alloy`, `echo`, `fable`, `onyx`, `nova`, `shimmer`
OpenAI	`tts-1-hd`	`alloy`, `echo`, `fable`, `onyx`, `nova`, `shimmer`
ElevenLabs	`eleven_multilingual_v2`	Various premium voices
Google	`standard`	Multiple language voices
Azure	`neural`	Multiple language voices
Deepgram	`aura`	`asteria`, `luna`, `stella`, `athena`, `hera`, `orion`, `arcas`, `perseus`, `angus`, `orpheus`, `helios`, `zeus`

TEMPERATURE

TEXT

default:"0.8"

Temperature control from 0-1 (Default: 0.8). Minimum allowed temperature is 0.6.

VOICE

TEXT

default:"alloy"

Select the voice for the assistant based on the TTS provider chosen above.

PRONUNCIATIONS

TEXT

Add custom pronunciation rules for names and terms. Use multi-line format with trailing semicolon.Format: language_code|name=pronunciation;Examples:

en|1=One;
ar|2=اثنان;
fr|3=trois;

VERBALIZE_NUMBERS

LIST

default:"OFF"

Control whether numbers in RAG content are verbalized during conversation.

Value	Description
`ON`	Verbalize numbers in conversation
`OFF`	Do not verbalize numbers

USER_RESPONSE_TIMEOUT

TEXT

default:"30"

Timeout for user response in seconds.

MAX_REMINDERS

TEXT

default:"3"

Maximum number of reminders to send to the user before considering hangup.

HANDOFF

LIST

default:"OFF"

Select the handoff provider for the assistant. Disable by selecting OFF.

Value	Description
`OFF`	No handoff capability
`NOVOMIND`	Integration with Novomind platform

HANDOFF_PLUGIN_URL

TEXT

Custom handoff plugin URL for the assistant.

HANDOFF_LIVE_ASSIST_AGENT

LIST

default:"OFF"

Live assist agent to dispatch call upon successful handoff.

Value	Description
`OFF`	No live assist integration
`agent-live-assist`	Standard live assist agent
`live-assist-agent`	Alternative live assist agent

WHATSAPP_NOTIFICATION

LIST

default:"OFF"

Select WhatsApp notification provider for the assistant.

Value	Description
`OFF`	No WhatsApp notifications
`NOVOMIND`	WhatsApp integration via Novomind

EMAIL_NOTIFICATION

LIST

default:"OFF"

Select Email notification provider for the assistant.

Value	Description
`OFF`	No email notifications

BACKGROUND_AUDIO

MULTI

Select background audio options for the conversation environment.

Option	Description
`OFFICE_AMBIENCE`	Office ambient sounds
`music01.ogg`	Background music track

BACKGROUND_AUDIO_VOLUME

TEXT

default:"0.5"

Volume level for background audio (0-1 range, e.g., 0.5 for 50%).

THINKING_SOUND

MULTI

default:"KEYBOARD_TYPING,KEYBOARD_TYPING2"

Select sound effects to play while the agent is in thinking state.

Option	Description
`KEYBOARD_TYPING`	Keyboard typing sound effect
`KEYBOARD_TYPING2`	Alternative keyboard typing sound
`bell.ogg`	Bell notification sound
`screenshot.ogg`	Screenshot capture sound

THINKING_SOUND_VOLUME

TEXT

default:"0.8"

Volume level for thinking sounds (0-1 range, e.g., 0.8 for 80%).

AVATAR

LIST

default:"OFF"

Select the avatar for the assistant. Choose OFF to disable or select a provider.

Avatar	Configuration
`OFF`	No avatar
`Sandra`	`tavus:name=Sandra\|replica=r9fa0878977a\|persona=pc074769bad9`
`Steph`	`tavus:name=Steph\|replica=r9c55f9312fb\|persona=pc074769bad9`
`James`	`tavus:name=James\|replica=r70c81a0519b\|persona=pc074769bad9`

TELEPHONY_WEB_PANEL

LIST

default:"OFF"

Enable telephony web panel integration. When enabled, the AI agent will send notifications to users to open the web panel.

Value	Description
`ON`	Enable telephony web panel
`OFF`	Disable telephony web panel

Chat Realtime Agent
AI Realtime Chat Assistant for text-based conversations, supporting different LLM text-to-text providers.

Show Agent Parameters

LLM

LIST

default:"openai:gpt-4o-mini"

Select the LLM provider and model for the assistant.

Provider	Model
OpenAI	`gpt-4o-mini`, `gpt-4o`
Groq	`llama-3.3-70b-versatile`, `deepseek-r1-distill-llama-70b`

USER_RESPONSE_TIMEOUT

TEXT

default:"30"

Timeout for user response in seconds.

MAX_REMINDERS

TEXT

default:"3"

Maximum number of reminders to send to the user before considering hangup.

HANDOFF

LIST

default:"OFF"

Select the handoff provider for the assistant.

THINKING_SOUND

MULTI

default:"KEYBOARD_TYPING,KEYBOARD_TYPING2"

Select sound to play while in thinking state.

Sound Option	Description
KEYBOARD_TYPING	Standard typing sound
KEYBOARD_TYPING2	Alternative typing sound
bell.ogg	Bell notification
screenshot.ogg	Screenshot sound

Live Assist Realtime Agent
AI Agent for video/audio calls. Silently listens to the customer and sends real-time text suggestions to the human agent on how to respond or answer user questions.

Show Agent Parameters

HANDOFF_PROVIDER

LIST

default:"OFF"

Select the handoff provider for the assistant.

STT

LIST

default:"OFF"

Select the STT provider for the assistant. Disable by selecting OFF.

Provider	Options
OFF	Disabled
OpenAI	`whisper-1`, `gpt-4o-mini-transcribe`
Azure	`Azure`

LLM

LIST

default:"OFF"

Select the LLM provider and model for the assistant.

SYSTEM_INSTRUCTIONS

TEXT

System instructions for the assistant.

TURN_DETECTOR_CONFIG

TEXT

Turn Detector Configuration settings.

Speech Live Translation Agent
AI Agent for video/audio calls that transcribes and translates speech in real time using Azure Cognitive Speech Services.

Show Agent Parameters

TARGET_LANGUAGE

MULTI

default:"en"

Multi selection of target language. If empty, speech recognizer will work in Auto Mode. Must be specified for translation.

Language	Code
Arabic	`ar`
English	`en`
French	`fr`
Bengali	`bn`
German	`de`

It is recommended to use source language to have faster results for Transcription and translation.

Use Cases

Real-time sales or support assistance
Interactive document review and guidance
Video KYC, onboarding or ID verification
On-demand call handling from LaunchPad or web forms
AI agent-driven real-time surveys and lead capture

Getting Started

Create Action from Realtime PowerLink

After installing the PowerLink, go to PowerLinks in your cPanel, select Actions and click Create New Action using this PowerLink.

Assign Agent to Action

After creating the action, assign the appropriate Agent to handle the real-time task associated with this PowerLink.

Configure PowerLink Action Parameters

Set up the action’s configuration parameters such as SOURCE_LANGUAGE or USER_PROMPT, depending on your expected usage scenario.

Configure Agent Instructions, Sources and Tools

Define the instructions, knowledge sources and function tools required by the assigned agent to handle the conversation effectively.

Attach to Workflow

Add the action to a workflow using workflow studio.

Test the Live Session

From LaunchPad or Create New Job or via API trigger, initiate a live session and confirm the agent joins, responds correctly and the task flow behaves as expected.

🚀 Getting Started

⚡ PowerLinks

🔌 Agent Assignment Utilities

📊 Monitoring

⚙️ Settings

Key Features

🎙️ Live Audio or Video

🤖 Supports AI Agents

📞 Human Agent Routing

🌍 Language & Identity Control

How to Install

How it Works

How it Looks for Customers

Key Interface Components

Configuration Parameters

Agents

Use Cases

Getting Started

🚀 Getting Started

⚡ PowerLinks

🔌 Agent Assignment Utilities

📊 Monitoring

⚙️ Settings

​Key Features

🎙️ Live Audio or Video

🤖 Supports AI Agents

📞 Human Agent Routing

🌍 Language & Identity Control

​How to Install

​How it Works

​How it Looks for Customers

​Key Interface Components

​Configuration Parameters

​Agents

​Use Cases

​Getting Started

Key Features

How to Install

How it Works

How it Looks for Customers

Key Interface Components

Configuration Parameters

Agents

Use Cases

Getting Started