This tutorial walks you through building low-latency Conversational AI agents using ElevenLabs and SambaCloud’s high-speed LLM inference engine. Low latency is crucial for smooth voice conversations, and SambaNova delivers this with specialized hardware optimized for world-class inference speeds using open-source models.

​
Pre-requisites

Before starting, ensure you have:

​
Setup

Follow these steps to set up your AI agent.

​
Access the Agent in ElevenLabs

  1. Go to the Agents page on ElevenLabs.
  2. Create a new agent, or select an existing agent to edit.

​
Configure the LLM settings

  1. Scroll to the LLM section of your agent settings.
  2. Select Custom LLM from the dropdown menu.

​
Retrieve SambaNova endpoint and model

  1. Open the SambaCloud Playground.
  2. Select View Code in the top-right to get your model endpoint URL and model name.

​
Generate your SambaNova API key

  1. Go to your SambaCloud account.
  2. Generate an API key from the portal.

​
Add API key to ElevenLabs

  1. Return to the ElevenLabs agent settings page.
  2. Under Workspace Secrets, add name and value.
  3. Name: SAMBANOVA_API_KEY.
  4. Value: Paste the API key from previous step.
This enables ElevenLabs to access your SambaNova model.

​
Set token limit

In the Limit token usage section, set maximum tokens to 1024. This helps control the response length for optimal conversational flow.

​
Save and test

  1. Select Save to apply changes.
  2. Test your setup by selecting Test AI agent followed by Call AI agent.
  3. See video walkthrough for details.

​
Video walkthrough