Mizuna - AI Companion Pet Robot

Mizuna: Your AI Companion on Wheels

The Story

I got tired of forgetting tasks, missing deadlines, and feeling like I needed three different apps just to stay organized. During exam season, I'd have random conversations with friends where I'd mention "oh yeah, that assignment is due Friday" and then... completely forget about it two hours later. Sound familiar?

So I built Mizuna—a little robot companion that actually listens to your life happening around it, remembers the important stuff, and gently keeps you on track. No wake words, no complicated commands, just a friendly presence that rolls around, watches your space, and makes sure nothing falls through the cracks.

The best part? It has personality. Mizuna can be your cheerful morning motivator, your focused study buddy, or your calming evening companion. It adapts to you, learns your routines, and feels less like a gadget and more like a helpful friend who's always excited to see you.

What Mizuna Actually Does

Captures Tasks from ConversationsThrough the Omi wearable device (a small recorder you can wear), Mizuna listens to your day and picks out tasks, deadlines, and commitments automatically. Mention "I need to submit that report by Wednesday" while chatting with a friend? Done—it's on your list. No apps to open, no manual entry.

Becomes Your Daily CompanionThat OLED screen on the front? Those are Mizuna's eyes, and they change based on mood and context. Happy in the morning, focused during work hours, sleepy at night. You can customize the personality profile to match your vibe—energetic, chill, professional, whatever works.

How It Works (The Fun Part)

The hardware is surprisingly simple. A Raspberry Pi serves as the brain, handling camera streaming, speech output, and coordination. Two NodeMCU microcontrollers manage the motors and face animations—one for movement, one for expressions. Four BO motors let it roll around smoothly, and a 12V battery keeps everything powered for hours.

For intelligence, Mizuna uses OpenAI's GPT-OSS-20B model through Groq's blazing-fast API. This open-weight model understands natural conversation, extracts tasks, and generates responses that feel genuinely helpful. Groq's Language Processing Units deliver over 1000 tokens per second, so interactions feel instant—no awkward waiting for the cloud to respond.

Everything gets stored locally in MongoDB—your preferences, routines, tasks, conversation history. Over time, Mizuna learns when you study, when you take breaks, what kind of reminders work best for you. It's memory that builds context instead of forgetting everything every session.

The mobile app (built with React Native) gives you manual control, live camera access, and task management. Check what Mizuna captured today, adjust its personality, or drive it around manually when needed.

Building Your Own

The entire build uses affordable, accessible parts—total cost under $150. You'll need basic tools (screwdriver, wire strippers, zip ties) and about an afternoon to assemble everything.

What goes into it:

• Raspberry Pi 3/4/5 (the brain)
• Two NodeMCU boards (motor control + display)
• Motor shield and four BO motors with wheels
• Raspberry Pi Camera module
• 1.3" OLED screen for the face
• Bluetooth speaker for voice
• 12V battery pack and voltage regulators
• MPU6050 for orientation sensing

Assembly follows a logical flow: mount motors to chassis, wire the motor shield, secure the Pi with good airflow, connect the camera and OLED display, then manage cables neatly with zip ties. The GitHub repository includes detailed circuit diagrams and step-by-step Arduino code for both NodeMCU boards.

Software setup involves flashing Raspberry Pi OS, enabling camera and I2C interfaces, installing Python dependencies, and configuring API keys for Groq and Azure Speech Service. Upload the Arduino sketches to each NodeMCU, launch the Flask backend and app server, then test each component individually before running everything together.

The modular design means you can start basic—just motors and expressions—then add intelligence gradually as you gain confidence.

Why This Matters

Most smart devices feel like products—designed to extract data, push subscriptions, lock you into ecosystems. Mizuna is different. The GPT-OSS-20B model is open-weight with Apache 2.0 licensing, meaning full transparency and no proprietary restrictions. The hardware is modular and repairable. The code is yours to modify, fork, and improve.

It's also genuinely affordable. Premium personal robots cost thousands; Mizuna costs roughly what you'd spend on a nice pair of headphones. Students, families, hobbyists, tinkerers—anyone can build one without breaking the bank.

And unlike voice assistants that forget everything the moment you stop talking, Mizuna remembers. It learns your patterns, adapts to your schedule, and becomes more helpful over time. That's the difference between a tool and a companion.

What's Next

Future versions could add SLAM mapping for multi-room navigation, wake word detection for voice commands, expanded sensors (temperature, air quality, PIR motion), and even offline operation with locally-hosted models. Multiple Mizuna robots could coordinate patrol routes or share context across household members.

But honestly? The current version already handles the daily grind beautifully. It's functional, affordable, and genuinely helpful.

Build One

All code, schematics, and documentation are available on GitHub at https://github.com/cyrixninja/MizunaRobot—fully open source and ready to fork. Start with basic motor control, add the expressive face, then layer on intelligence at your own pace. Whether you're learning robotics, exploring AI integration, or just want a friendly helper around the house, Mizuna grows with you.

Detailed Build Guides:

• Project Overview & Features
• Complete Materials List
• Step-by-Step Assembly Instructions

Build one, break it, improve it, share what you create. That's how great projects evolve.

Components: Raspberry Pi, NodeMCU (x2), Motor Shield, BO Motors (x4), Camera, OLED Display, Bluetooth Speaker, Battery Pack, LED MatrixSoftware: Python, Flask, GPT-OSS-20B via Groq, MongoDB, Azure Speech, React Native, ArduinoLicense: Open source—hardware, software, documentation all freely available

Ready to get started? Check out the repository and build guides above. Questions? Drop them in the comments—I'd love to see what you build!