0 parts selected
Pick a smart local brain, add hearing and voice, and connect motors without headaches.
Each stage holds 1 active component. Click Swap to replace it cleanly.
Click any blueprint to see its architecture story and auto-load its components.
Select one drop-in replacement to take this stage in your build:
Quickly check if an AI model will run smoothly on your home PC or Mac.
⚡ Fits easily on a standard 16GB Mini PC or Mac
Price details
Suggest an open-source brain, motor, power module, or 3D chassis.
Everything you need to know about the software files on your desktop.
Put them in a folder: Create a new folder on your computer (e.g. MyDeskbot) and move both downloaded files into it.
Rename the settings file: Rename env.example to .env (put a dot at the start). Open it in Notepad only if you need to paste private API keys.
Start your robot: Make sure Docker Desktop is open, open a terminal in your folder, and run:
docker compose up -d
Think of Docker like digital shipping containers.
Normally, installing AI models requires setting up dozens of confusing background tools, code libraries, and Python versions that can mess up your regular computer files.
Docker keeps everything neatly sealed inside its own clean sandbox. It downloads the exact tools you picked, connects them together automatically, and can be deleted cleanly in one click without leaving junk on your hard drive.
This is a plain-text instruction recipe for your computer.
When Docker reads this file, it sees: "Fetch the brain tool, open port 8283 for memory, give it a private storage folder, and hook up the voice engine."
Instead of typing 20 different setup commands into a terminal, this single recipe card tells your computer everything it needs to know in one breath.
"ENV" stands for Environment. It’s where you safely store private account keys or passwords.
If you ever use cloud services like ElevenLabs (for realistic voices) or RunPod (for rented cloud graphics cards), you paste your secret keys into this file.
The blueprint reads your keys from here quietly so you never have to paste private passwords into your main software code!
The complete, no-jargon blueprint to bringing an AI brain into the physical world.
Every companion robot on this directory breaks down into five distinct sensory layers:
Option A: Pure Cloud (Fastest & Lightest Hardware)
You use an inexpensive $7 ESP32-S3 microcontroller on your desk. When you speak, it streams audio over your home Wi-Fi to Groq and ElevenLabs. Total hardware: under $25 AUD.
Option B: 100% Sovereign Local (Private & Offline)
You run a dedicated Mini PC (like an AMD Ryzen with 32GB RAM or an Apple Silicon Mac) on your desk. Ollama, Whisper, and Piper run completely offline. Zero subscriptions, zero internet required.
Standard USB webcams send raw, uncompressed 1080p video streams into your computer. This forces your CPU or graphics card to decode millions of pixels every second, which burns battery and creates sluggish response times.
The Smart Solution: Dedicated vision sensors (like OpenMV or ESP32-CAM) do the face-math directly on the sensor itself. Instead of video, they only send a lightweight text message down the wire: X: 120, Y: 85. Your robot tracks your face instantly with zero computer lag.
The #1 mistake beginners make is using cheap 3.5mm analog microphone jacks plugged into hobby boards. Microcontrollers emit electrical noise that sounds like a loud frying pan to AI speech recognizers.
Always choose I2S (Inter-IC Sound) digital modules like the INMP441 or MS4030. They convert sound directly into 24-bit binary data right on the silicon chip, giving the speech model pristine, noise-free audio every single time.
To give your companion lifelike body language, you only need two motors mounted in a Pan/Tilt Gimbal Bracket (~$5):
Use metal gear servos like the MG90S rather than plastic nylon gears so your companion's neck never strips when moving heavy screens or camera brackets.
If your speaker makes an annoying electrical whine every time the robot tilts its head, you have a ground loop or a floating ground reference.
The Golden Rule of Robotics: All GND (Ground) pins across your battery, your motor driver, and your microcontroller must connect to a single shared wire rail on your breadboard. When all parts share the exact same 0-volt baseline, electrical motor buzzing disappears completely.
When you click on an open-source project link on this site (like Ollama, faster-whisper, or Piper), you will usually land on a website called GitHub.
For beginners, GitHub looks like a confusing wall of code files and developer jargon. Do not panic! You do not need an account, and you do not need to know how to code. Here is your 3-step survival guide:
Ignore the intimidating list of files at the top. Scroll down past them until you see a formatted document called the README.md. This is the official project manual. It is written in plain English and tells you what the tool does, system requirements, and step-by-step setup guides.
If you want to download a project's files without typing any terminal commands, look near the top right of the page for a bright green button that says <> Code. Click it, then click Download ZIP. Unzip that folder on your computer and you have the complete software ready to explore.
Many tools offer pre-built installer files (like an everyday .exe for Windows or .dmg for Mac). On the right-hand sidebar of any GitHub page, look for the word Releases. Clicking the latest version gives you direct download links for regular point-and-click installers!