Dynamic Conversation

Learn how to customize your AI agent for each conversation & pass additional metadata to the agent.

Learn how to customize your AI agent for each conversation by adding personalized details and passing custom parameters. For example, greet users by their name, adjust responses based on account-specific data, or include metadata to enhance interactions.

What You’ll Need

• An ElevenLabs account
• A configured ElevenLabs Conversational Agent (create one here)
• Python 3.7+ or Node.js 16+

Agent Conversation Overrides

In ElevenLabs Conversational AI, you typically define a top-level agent with default settings. This feature allows you to create tailored voice experiences, without creating new agents.

The following steps will show you how to configure the prompt & first_message on setup to greet the user by their name.

1

Enable overrides

Go to the Security tab in your agent settings and enable overrides for the first message and system prompt.

• Javascript
• Python
• Swift

Ensure you have the latest Javascript/React SDK installed.

2

Start the session with custom overrides

const conversation = await Conversation.startSession({
  ...
  overrides: {
      agent: {
          prompt: {
              prompt: `The customer's bank account balance is ${customer_balance}. They are based in ${customer_location}.`
          },
          firstMessage: `Hi ${customer_name}, how can I help you today?`,
          language: "en"
      },
      tts: {
          voiceId: "" // override the voice id.
      }
  },
  ...
})

With this agent override feature, you can now create a single agent and customize it with every customer conversation.

Troubleshooting

Configuration Issues

If the configuration override isn’t working:

• Verify the configuration structure matches the expected format
• Check that all required fields are present
• Ensure the config object is properly passed to the Conversation constructor