For AI agents: a documentation index is available at the root level at /llms.txt and /llms-full.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.
Ask AIPlaygroundLoginFree API Key
HomeAPI ReferenceVoice AgentSpeech-to-TextText-to-SpeechIntelligenceSelf-Hosted Deployments
HomeAPI ReferenceVoice AgentSpeech-to-TextText-to-SpeechIntelligenceSelf-Hosted Deployments
  • Get Started
    • Overview
    • Build a Voice Agent
    • Feature Overview
    • Template Apps
  • Configure
    • Overview
    • STT Models
    • LLM Models
    • TTS Models
    • Media Inputs & Outputs
    • Prompting Voice Agents
    • Multilingual Voice Agents
    • Maintaining Context
    • Reusable Agent Configurations
  • Build
    • Multi-Agent Architecture
  • Connect
      • Twilio Streaming
      • Genesys Cloud CX
      • Amazon Connect
      • AudioCodes (LiveHub)
  • Controls
  • Optimize
    • Voice Agent TTS Controls
    • Message Flow
    • Audio & Playback
    • Audio Preprocessing & Barge-In
    • Adaptive Echo Cancellation
  • Resources
    • SDKs
    • UI Components
    • API Reference
LogoLogo
Ask AIPlaygroundLoginFree API Key
On this page
  • Genesys Cloud CX and Deepgram Voice Agent
  • How it works
  • Before you begin
  • Step 1: Create an Audio Connector in Genesys
  • Step 2: Add the connector to your Architect flow
  • Step 3: Configure your AI agent
  • Example end conversation function
  • Step 4: Set up middleware server (self-hosted only)
  • Customization options
  • Additional resources
ConnectTelephony Integrations

Genesys Cloud CX and Deepgram Voice Agent

Was this page helpful?
Previous

Amazon Connect and Deepgram Voice Agent

Next
Built with

Genesys Cloud CX and Deepgram Voice Agent

The Deepgram Voice Agent API integrates with Genesys Cloud CX through the Genesys Audio Connector, establishing a bidirectional WebSocket connection for real-time voice input and output.

The Voice Agent API bundles speech-to-text, LLM reasoning, and text-to-speech into a single pipeline, delivering approximately 425 ms round-trip latency.


How it works

  1. A caller enters your Genesys call flow
  2. Genesys connects to your Voice Agent through a WebSocket
  3. Audio streams to Deepgram (STT, LLM, TTS)
  4. The agent responds with synthesized voice
  5. The caller can interrupt naturally with barge-in support

Before you begin

You will need:

  • A Deepgram API key with Voice Agent enabled
  • A Genesys Cloud CX account with Audio Connector access
  • A public server endpoint connected to Deepgram

Before you can use Deepgram, you’ll need to create a Deepgram account. Signup is free and includes $200 in free credit and access to all of Deepgram’s features!


Step 1: Create an Audio Connector in Genesys

  1. Go to Admin > Integrations
  2. Add Audio Connector
  3. Name the integration (for example, Deepgram Voice Agent)
  4. Set Connection URI to wss://integrations.deepgram.com/genesys/ac
  5. Add your Deepgram API key under Credentials
  6. Activate the integration

Step 2: Add the connector to your Architect flow

  1. Open Architect in Genesys Cloud
  2. Open or create an Inbound Call Flow
  3. Add a Call Audio Connector action
  4. Select the integration you created in Step 1

This action hands the call to the Deepgram Voice Agent.


Step 3: Configure your AI agent

In the Audio Connector’s Session Variables, add a variable:

  • Name: config
  • Value: a JSON string containing your agent configuration

For optimal audio quality, include the following in your agent configuration:

1"audio": {
2 "input": {
3 "encoding": "mulaw",
4 "sample_rate": 8000
5 },
6 "output": {
7 "encoding": "mulaw",
8 "sample_rate": 8000,
9 "container": "none"
10 }
11}

To allow the Voice Agent to end the call, add an end_conversation function to the agent configuration. The function’s properties are sent back to Genesys and can be mapped as session output variables.

Example end conversation function

1{
2 "name": "end_conversation",
3 "description": "Call this function when the user says goodbye or the conversation ends.",
4 "parameters": {
5 "type": "object",
6 "properties": {
7 "needs_escalation": {
8 "type": "boolean",
9 "description": "Set to true if the user requests a human or manager."
10 }
11 }
12 }
13}

Step 4: Set up middleware server (self-hosted only)

If you use a self-hosted setup, deploy a middleware server to connect Genesys with the Deepgram Voice Agent. Contact your Deepgram account executive for access and setup instructions.


Customization options

You can adjust your agent’s behavior using these configuration fields:

SettingField
Personalitythink.prompt
Voicespeak.provider.model
Greetinggreeting
Knowledge / abilitiesthink.functions

Additional resources

  • Deepgram Voice Agent documentation
  • Genesys and Deepgram STT integration