Skip to main content
This guide shows how to build low-latency Conversational AI agents using ElevenLabs and SambaCloud’s high-speed LLM inference engine. Low latency is critical for smooth voice conversations — SambaNova’s specialized hardware delivers the inference speeds required for real-time audio.

​
Prerequisites

​
Setup

1

Create a new agent in ElevenLabs

Log in to ElevenLabs, open Agents, and click + New agent.
ElevenLabs Agents page with the New agent button
2

Set the LLM to Custom

Open the agent’s LLM section and switch the configuration to Custom.
ElevenLabs Custom LLM configuration
3

Point the agent at SambaCloud

In the Custom LLM panel, fill in the SambaNova details:
  • Server URL: select Chat Completions and enter https://api.sambanova.ai/v1/ (ElevenLabs appends /chat/completions automatically).
  • Model ID: enter a SambaCloud model (for example, DeepSeek-V3.1).
  • API Key: select Secret, then add your SambaCloud API key as a workspace secret and select it from the dropdown.
  • Max tokens (optional): in the Limit token usage section, set maximum tokens to 1024 for optimal conversational response length.
Wait for the Connection successful confirmation, then click Save.
ElevenLabs Custom LLM config with SambaCloud details
4

Preview and chat with your agent

Click Test AI agent → Call AI agent to launch a live preview. Talk to the agent and confirm it responds in real time.
ElevenLabs agent preview panel
5

Verify the calls route through SambaCloud

To confirm your agent is hitting SambaNova:
  1. Open the SambaCloud Console and go to Usage (or Dashboard → Usage).
  2. Filter by the API key you added to ElevenLabs and narrow the time window to the last hour.
  3. Note the current Requests count.
  4. Run a 1–2 turn test call in ElevenLabs.
  5. Refresh the Usage page — the counter should increment by the number of model turns.
If the counter stays flat, recheck the Server URL trailing slash, the secret name, and that the API key is active.

​
Troubleshooting

  • Verify your SambaCloud API key is correct and active in the SambaCloud portal
  • Confirm the Server URL is https://api.sambanova.ai/v1/ (trailing slash required)
  • Re-enter the secret in ElevenLabs and click Save
  • Check the model ID matches a valid model from the SambaCloud models page
  • Confirm your SambaCloud account is active
  • Confirm the correct API key is selected in ElevenLabs
  • Wait 1–2 minutes and refresh the Usage page
  • Check the ElevenLabs agent logs for errors