[ESP-Claw + UNIHIKER K10] Build a Desktop AI Air Butler with Zero Code

Introduction

Have you ever felt dizzy, drowsy, or struggled to focus during long hours of brainstorms in a small, closed

meeting room, or while deeply immersed in R&D technical troubleshooting at your desk? Often, the hidden

culprit behind this "brain fog" is the silent accumulation of carbon dioxide (CO2 ) in micro-environments.

Modifying the entire building's centralized ventilation or HVAC system is highly impractical, whereas typical off

the-shelf air monitors do nothing more than statically output cold numbers on a screen, lacking any engaging,

closed-loop interaction.

By leveraging the ESP-Claw Framework, the UNIHIKER K10 controller, and a high-precision Gravity:

STCC4 CO₂ Sensor, you can construct a personalized, Desktop-grade AI Air Butler at a minimal hardware

cost—entirely without writing a single line of traditional embedded code.

In this project, the hardware ceases to be a passive gauge and transforms into an intelligent agent capable of

"understanding the air". The ESP-Claw framework seamlessly tracks local environmental dynamics. All you

need to do is give natural language instructions to your AI Agent: "When the room's CO₂ levels spike and start

making people sleepy, turn on my desktop fan to exhaust the air and show a friendly alert on the screen." The

AI handles the entire perception, analytical reasoning, and hardware execution stack natively, ensuring your

workspace remains fresh and high-performing.

Hardware Bill of Materials (BOM)

1x UNIHIKER K10 Main Controller

1x Gravity: STCC4 CO₂ Sensor Module

1x Fan Module (for active localized air circulation)

Hardware Connection

The physical setup is completely plug-and-play, avoiding messy breadboards or discrete components:

Connect the Gravity: STCC4 CO₂ Sensor directly to the I2C interface located on the right-hand side of the UNIHIKER K10 using a 4-pin wire sequence (VCC, GND, SCL, SDA).

Connect the Fan Module directly to the P0 interface on the left-hand side of the UNIHIKER K10 using a 3-pin control wire.

Step-by-Step Implementation Guide

Before proceeding, ensure your UNIHIKER K10 has been successfully flashed with the latest ESP-Claw firmware and connected to your local Wi-Fi network with your LLM API configuration ready. We will convert this hardware combination into an autonomous assistant across four structured steps.

Step 1: Establishing Indoor Air Perception

Instead of rushing into automated controls, we first guide the AI to interpret data from the STCC4 sensor.

Since ESP-Claw can digest knowledge-base references on the fly, we can point it directly to the component's specification.

Send the following prompt to your K10 via the chat interface:

"I have connected the STCC4 carbon dioxide sensor to the right-hand I2C port of my UNIHIKER K10.

Please look up this skill documentation to learn how to communicate with the hardware: https://skills-lab.esp-claw.com/zh-cn/skill/dfrobot_stcc4_i2c . Read the current CO₂ concentration right now, evaluate it against human health standards, and tell me if the current room air quality is suitable for long hours of intense mental work."

Following this instruction, ESP-Claw autonomously invokes the driver primitives detailed in the documentation

to poll the sensor. Crucially, the system does not just print raw parts-per-million (ppm) integers; it reasons

about the physical data in relation to cognitive performance thresholds and human health context.

Step 2: Defining Intelligent Ventilation Policies

Once the AI establishes cognitive environmental awareness, we integrate the executive actuator (the fan) to close the loop on autonomous decision-making.

Send this message in the chat:

"My fan module is hooked up to the left-hand P0 interface (GPIO1) of the K10 board. Please configure an intelligent ventilation policy for me: Poll the CO₂ concentration once every 10 minutes. If the reading climbs above 1000 ppm (the threshold where air begins to induce drowsiness), turn on the fan to bring in fresh air. Once the concentration returns below 800 ppm, turn the fan off. Print the execution status and live environment metrics clearly on the screen after each evaluation cycle."

The ESP-Claw framework breaks down this intent into conditional runtime logic. Unlike hard-coded scripts,

you can seamlessly add nuanced context using conversational caveats, such as: "If it's past 10:00 PM, run the

fan at half speed or skip the cycle if the room is quiet to prevent noise distractions." This highlights the

flexibility of an AI Agent over traditional IoT rule engines.

Step 3: Continuous Data Logging and Cloud Analytics (Optional)

To analyze long-term trends and identify peak stuffiness periods in your workspace throughout the week, you can request your butler to stream telemetry to an IoT data broker like ThingSpeak.

Once you set up a dashboard containing fields for "CO₂ Concentration" and "Fan State" on ThingSpeak and grab your Write API Key, instruct your agent directly:

"Please stream the polled CO₂ concentration values and the digital state of the fan to my ThingSpeak

channel using Write API Key: [YOUR_API_KEY] automatically at every check interval."

The system will then run the remote API hooks in the background, mapping out clear historic visualization curves for your work zone.

Step 4: Compiling and Packaging the "Air Butler" Skill Set

At this stage, your UNIHIKER K10 has successfully demonstrated real-time perception, localized actuator control, and cloud logging. To finalize the deployment, we lock these behaviors into a permanent, production ready system skill.

Issue the final stabilizing command:

"Please package all the behaviors we just configured into a persistent, autonomous 'Smart Air Butler'

skill and install it as a default daemon. The system must run quietly in the background, continuously

safeguarding the space, handling the ventilation cycles, and generating a concise workplace health

report at 5:00 PM every evening."

Once compiled, the system loops autonomously on your desk—a dedicated, smart companion keeping your mind clear and your workspace refreshed.

Summary

By blending the plug-and-play capability of the UNIHIKER K10 and the STCC4 sensor with the conversational

intelligence of the ESP-Claw framework, we have bypassed the tedious paradigm of low-level C++ firmware

debugging. The sensor bridges the physical environment, while the LLM-driven agent provides the cognitive

engine. In just a few lines of natural dialogue, we transformed raw hardware elements into an articulate,

adaptive agent—proving that the future of hardware engineering lies in natural intent orchestration.