Surprising Thing ZimaCube 2’s AI NAS Can Really Do

Introduction

As ZimaSpace, we are always exploring how intelligent hardware can make real-world workflows simpler, faster, and more secure. In this article, we highlight how ZimaCube 2, a system built for more than storage, can be transformed into an AI NAS that not only manages files but also understands them.

This blog is adapted from a hands-on video review and experiment by Zero Noichi, who explored how to attach AI to a NAS and turn it into a smart, automated data assistant. We would like to express our sincere thanks to Zero Noichi for testing ZimaCube 2 in such a creative way and for sharing practical ideas the community can learn from. Based on that video, we will walk through how to use ZimaCube 2 as an AI NAS that can classify photos, summarize documents, and help you search your “digital life” more naturally.

From Traditional NAS to AI NAS

ZimaCube 2 is designed as a personal cloud and media server, but it goes far beyond traditional NAS.

In the video, ZimaCube 2 is introduced as part of a broader Zima family that includes ZimaBoard and ZimaBlade, but here the focus is on using it as a NAS first and then upgrading it into an AI NAS. The creator explains how many people still rely on Google Drive, OneDrive, and Dropbox to store files and offload smartphone storage, but those services come with recurring subscription fees and external data processing.

A NAS, by contrast, is purchased once and then only consumes electricity as long as the hardware runs. The video explains how NAS started in enterprise environments for file sharing and internal servers, and how devices like ZimaCube 2 now make that model accessible for home users. By adding AI on top, the system evolves from simple file storage into a self-aware, AI NAS that can actually understand and organize content.

Hardware and Expansion: Why ZimaCube 2 Is Built for More Than Storage

The hardware behind ZimaCube 2 is what makes an AI NAS possible.

The creator unboxes ZimaCube 2 and highlights its server-grade design, including a sturdy aluminum chassis and a cube-shaped form factor that blends well next to devices like a Mac. Inside, ZimaCube 2 in the Pro configuration uses an Intel Core i5‑1235U processor (with support for up to 64 GB of DDR5 RAM), while the Standard configuration uses an Intel Core i3‑1215U. Dual PCIe slots (Gen4 and Gen3) and a low-profile expansion bay allow the installation of additional network cards or even GPUs.

Featured

ZimaCube 2 Personal Cloud NAS (In Stock)

ZimaCube2

$799.00 - $2,499.00

Buy Now

Key hardware points include:

• Multiple 3.5-inch drive bays for up to six hard drives
• A dedicated NVMe carrier with four M.2 slots for SSDs and potential RAID configurations
• Dual Thunderbolt 4 ports for high-speed connections, such as 10 GbE networking to a Mac
• 2.5 GbE LAN ports and standard video outputs (HDMI and DisplayPort)

This foundation enables ZimaCube 2 to function as a flexible AI NAS, where storage, compute, and expansion are designed to support both data-heavy workflows and GPU-accelerated AI tasks.

Initial Setup: Turning ZimaCube 2 into a NAS

ZimaOS Plus simplifies NAS setup so that anyone can get started quickly.

Out of the box, ZimaCube 2 boots into ZimaOS, a Linux-based operating system that has been customized for NAS use. After connecting the device to power and network, the creator opens the web-based interface from another Mac on the same LAN, selects a language, creates a local account, and logs in.

From there, the steps are straightforward:

• Use the web console to access ZimaOS and verify that the NAS is online
• Connect from macOS Finder via “Connect to Server” using the IP address (for example, 192.168.0.xxx)
• Log in with the same account to see the shared folders and demo files

Once this is done, ZimaCube 2 is already working as a basic NAS: you can browse files, stream demo videos over the network, and view PDFs in the browser. Adding additional hard drives is as simple as inserting a drive into an available bay, formatting it through the web UI, and enabling it as a new storage volume (for example, a new 2 TB disk).

This baseline NAS configuration is what the later AI NAS features will build on.

Expanding Storage: From Terabytes to “A Life’s Worth of Data”

ZimaCube 2 is designed to scale until it can store almost everything you own digitally.

In the video, the creator inserts an external Western Digital hard drive into ZimaCube 2 and uses ZimaOS to format it as a new volume. The process takes only a few clicks:

• Insert the drive into an available bay
• Let ZimaOS detect it and show it as a “new” device
• Format the disk and enable it for use
• Confirm that it appears in both the NAS management console and the network share

With all drive bays filled and NVMe slots utilized, ZimaCube 2 can reach up to 100 TB of total capacity, depending on the drives used. The creator notes that filling all slots with large drives could cost around 500,000 yen, but that such capacity could realistically store an entire lifetime of photos, videos, and documents.

This massive storage capacity is essential for an AI NAS because the system can only analyze and index what it can store. The more data ZimaCube 2 holds, the more powerful its AI-based search and classification become.

Why Add AI to a NAS?

The goal is not just to chat with an AI, but to let AI manage your files.

The creator’s concept is clear: instead of using an AI like ChatGPT only as a conversational tool, use a local AI model to analyze the files stored on the NAS. This is where ZimaCube 2 becomes a true AI NAS:

• When photos are uploaded, AI analyzes them and tags them with descriptive labels
• When documents are stored, AI reads them and extracts key information such as amounts, names, and topics
• When you search using natural language, AI and vector search work together to surface the most relevant files

The creator emphasizes that AI should:

• Automatically categorize photos (e.g., “family trip,” “Yokohama,” “amusement park,” “night view”)
• Add tags in the background so that users can search later using fuzzy queries
• Periodically review unused files and suggest archiving or reorganizing them

This approach makes the AI NAS a quiet but powerful assistant that cleans, labels, and organizes data with minimal manual effort.

How Vector Search Powers Natural Language Queries

Vector search allows the AI NAS to understand meaning, not just filenames.

Instead of relying on exact text matching, the system uses vector embeddings (numeric representations of meaning) to store how the AI understands each file. The creator explains this with a simple mental model:

• Each concept (for example, “cat,” “dog,” “animal,” “cute”) is mapped to a position in a numerical space
• When a photo is analyzed, the AI gives it a set of vectors that reflect its content (e.g., cat, outside, cute)
• When a user searches for “animals,” the system finds vectors near that concept, even if the word “animal” was never explicitly tagged

This means that:

• A search for “orange photo” or “red clothes” can still surface relevant images
• A query like “Japanese astronaut” can find a folder of portraits labeled only with names
• A vague request such as “that café shot I took recently” can be interpreted as a combination of environment, color, and objects

By storing vector metadata for every file, the AI NAS can deliver results that feel closer to how humans remember and describe content.

Building the Software Layer: NAS AI on ZimaOS

The AI engine runs on top of ZimaOS, without replacing the NAS foundation.

Instead of rewriting ZimaOS, the creator builds a separate software module called “NAS AI” and runs it inside the system using standard tools like SSH and terminal access. The key steps include:

• Enabling developer mode and SSH access in the ZimaOS interface
• Opening a web-based console to log into the system shell
• Uploading the “NAS AI” software to ZimaCube 2
• Executing the program so that it can start monitoring designated folders

The AI engine then begins to:

• Watch specific directories where new files will be uploaded
• Extract text and metadata from documents (PDFs, reports, configuration files, code repositories)
• Analyze images and generate descriptions, tags, and vector representations

This keeps the NAS and the AI logic loosely coupled: ZimaOS continues to handle storage, permissions, and network sharing, while the AI NAS layer focuses entirely on understanding and indexing the data.

Practical Demo: Documents, Code, and Photos

Real files show how an AI NAS behaves in daily use.

To show how the system works in practice, the creator prepares a mixed sample dataset:

• Source code repositories (for example, Go, Python, React, and shell scripts)
• Configuration files like NGINX configs
• Various PDF documents such as contracts, licenses, and reports
• Notes and lists (like a reading list or memo)
• A set of photos including food, stadiums, landscapes, cafés, and portraits of NASA astronauts

These files are then copied into the NAS share that the AI is monitoring. As soon as the upload completes, the AI NAS starts parsing and describing each item. In the management UI, you can see:

• Folders recognized as “personal development and learning code directories”
• Config folders described as “NGINX configuration files”
• Photo libraries summarized as “portrait photos,” “tourist spots,” or “stadium scenes”

This automated annotation makes it much easier to understand what is inside each folder without manually opening every file.

Smart Search Examples: From “Café Photo” to “Total Invoice Amount”

The AI NAS enables complex searches that would be impossible with filenames alone.

The video walks through several concrete search scenarios:

1. Café photo search

   • The user types “café” as a keyword.
   • The system returns images that contain café-like environments: drinks, tables, indoor scenes with a café atmosphere.
   • Some results are more accurate than others, but higher-scoring images appear at the top thanks to vector similarity.

2. Landscape and color-based search

   • Queries like “landscape” or “green” surface photos of mountains, nature scenes, and images dominated by green tones.
   • Tags such as “landscape,” “travel,” or “green” are automatically generated by the AI.

3. “Man” or “male person” search

   • A query like “man” returns portraits and images where the subject is likely a male.
   • Even when faces are in shadow or partially visible, related images appear with lower similarity scores, showing the flexibility of vector-based matching.

4. Invoice total calculation

   • The user asks the system to “summarize the total amount from all invoices.”
   • AI scans all relevant invoice documents, reads the amounts, and calculates a combined total (for example, 2,835,360 yen).
   • The result includes references to which PDF files contributed to the total, making the process auditable.

These examples demonstrate how an AI NAS can replace manual bookkeeping, complex folder naming, and rigid file search with a more human-friendly interface.

Relationship Graphs and Entity-Centric Views

The AI NAS can also visualize relationships between files and entities.

Beyond simple search results, the system creates relationship graphs that show how files, tags, and people connect. For example:

• A folder of astronaut portraits is recognized as containing specific Japanese astronauts
• An entity-centric view for one astronaut shows all images in which that person appears, along with related tags and descriptions
• For technical files, tags like “NGINX,” “configuration,” “server,” or specific programming languages appear in a graph that clusters relevant resources together

This structure makes it easier to:

• See all files related to a person, topic, or project
• Understand how code, configs, and documentation relate to each other
• Navigate large datasets visually rather than through static folder trees

It is another way that the AI NAS turns raw storage into an intelligent, navigable knowledge base.

Privacy and Local Control: Why Local AI Matters

Local AI keeps sensitive data on your own hardware, not on external servers.

The creator contrasts cloud AI services with a local AI NAS approach:

• Services like Google Photos or Gemini can already interpret images, but they do so by sending data to external servers
• For family photos, private documents, or internal company files, some users are uncomfortable with external analysis and storage
• A locally hosted AI on ZimaCube 2 keeps all processing inside the device

By combining:

• Large storage capacity (up to around 100 TB)
• A NAS-optimized operating system (ZimaOS Plus)
• An AI engine running either locally on a GPU or via carefully controlled online models

users can build a privacy-preserving AI NAS that does not depend on third-party platforms for search, classification, or automation.

Future Potential: From Home Lab to Business Workflow

An AI NAS on ZimaCube 2 can scale from personal use to professional environments.

The video concludes with scenarios where this system could be deployed beyond a single user:

• Small businesses that want AI-powered document management without sending files to the cloud
• Teams that need to search years of project archives, code, and configs using natural language
• Home users who want to search “family trip to Yokohama with the Ferris wheel” and instantly find the right photos

Because ZimaCube 2 offers expandable storage, PCIe slots for GPUs, and a stable NAS-focused OS, it can grow with these needs over time. As AI models become faster and more efficient, ZimaCube 2 is positioned to host them locally, turning a simple NAS into a long-term AI NAS platform for both enthusiasts and professionals.