---
title: "NVIDIA brings agents to life with DGX Spark and Reachy Mini"
description: "At CES 2026, NVIDIA presents a desk-based personal assistant that combines DGX Spark processing with a Reachy Mini robot, allowing an agent to converse, perceive camera input, speak, and perform actions. The tutorial assembles NVIDIA Nemotron 3 Nano for reasoning, NVIDIA Nemotron Nano 2 VL for vision, ElevenLabs for text-to-speech, and Reachy Mini hardware or simulation through the NeMo Agent Toolkit and Pipecat. Its configuration routes text, visual, and action requests to different models or a ReAct tool-calling agent, while the robot daemon exposes movement and state capabilities. The guide covers local, cloud, and serverless deployment, then starts the services needed for voice and vision interaction; the completed setup greets users and supports example text and camera prompts."
---

# NVIDIA brings agents to life with DGX Spark and Reachy Mini

[Hugging Face](https://yomu.fyi/company/hugging-face) · Jeff Boudier, Nader Khalil, Alec Fong · Jan 5, 2026

**Type:** Tutorial

## Summary

At CES 2026, NVIDIA presents a desk-based personal assistant that combines DGX Spark processing with a Reachy Mini robot, allowing an agent to converse, perceive camera input, speak, and perform actions. The tutorial assembles NVIDIA Nemotron 3 Nano for reasoning, NVIDIA Nemotron Nano 2 VL for vision, ElevenLabs for text-to-speech, and Reachy Mini hardware or simulation through the NeMo Agent Toolkit and Pipecat. Its configuration routes text, visual, and action requests to different models or a ReAct tool-calling agent, while the robot daemon exposes movement and state capabilities. The guide covers local, cloud, and serverless deployment, then starts the services needed for voice and vision interaction; the completed setup greets users and supports example text and camera prompts.

## Context

The work is motivated by turning a conventional chat interface into a more natural, interactive agent that can see through a camera, speak aloud, perform actions, and process personal data privately. It also aims to provide an open, inspectable alternative in which users control the models, prompts, tools, and robot actions.

## Approach / What changed

The implementation composes existing components: Nemotron reasoning and vision models, ElevenLabs text-to-speech, the NeMo Agent Toolkit for orchestration and routing, a ReAct agent for tool calls, Pipecat for low-latency audio and video, and Reachy Mini hardware or simulation. The guide configures services, launches a robot daemon, runs the bot and agent servers, and connects through the Pipecat Playground.

## Takeaways

- The demo uses approximately 65GB of disk space for the reasoning model and 28GB for the vision model when deployed locally.
- The router directs simple text queries to a fast text model, environment questions with camera input to a VLM, and real-time information or action requests to a ReAct agent with tools.
- For physical actions, the guide recommends a confirm-before-actuation pattern and setting max\_tool\_calls to prevent uncontrolled tool-call loops.

**Tags:** [LLMs](https://yomu.fyi/topic/llm), [Open Source](https://yomu.fyi/topic/open-source), [Performance](https://yomu.fyi/topic/performance), [Privacy](https://yomu.fyi/topic/privacy), [Python](https://yomu.fyi/topic/python)

- Source: [Hugging Face](https://huggingface.co/blog/nvidia-reachy-mini)
- Source URL: https://huggingface.co/blog/nvidia-reachy-mini
- Ingested by Yomu: 2026-08-27T16:01:01.576Z

[Read original post](https://huggingface.co/blog/nvidia-reachy-mini)