CEPi
A compact, always-on personal AI server
Ready to use out of the box, CEPi keeps your data local and gives everyone access to a truly private, always-on AI experience.
AI NPC
Your dedicated, locally hosted AI agent
Powered by the Hermes system-level AI engine, it runs around the clock, autonomously handling coding and long-running tasks to put AI to work for you.
Naohua AI NPC
A local AI system built to act, not just chat.
Naohua AI NPC is a next-generation local AI appliance powered by Hermes, an OS-integrated agent engine, and built around Lucy AI OS and the Cephalon skill ecosystem. Based on Ubuntu, Lucy AI OS extends the conventional graphical interface with a Language User Interface (LUI), making natural language a first-class way to interact with the system.
Unlike a session-based chatbot, Naohua runs continuously as a persistent system service. Within the permissions granted by the user, it interprets intent, plans and breaks down complex tasks, generates and executes code, and coordinates applications, tools, and system services. It turns natural-language instructions into concrete, traceable actions on the device.
By default, model inference, user content, and task data are stored and processed entirely on the device, with no automatic fallback to the cloud. External services are used only when explicitly configured by the user. Software updates may require network access, but they do not upload user content.
Through the Cephalon skill ecosystem, Naohua can expand beyond the capabilities it ships with. Skills can be created, shared, and reused, allowing the system to gain new capabilities over time.
Naohua AI NPC
Product introduction
CPU: AMD Ryzen™ AI Max+ 395
16 cores, 32 threads, Zen 5 architecture, 5.1 GHz max boost clock, excellent performance.
GPU: AMD Radeon™ 8060S Graphics 40
compute units, RDNA 3.5 architecture, outstanding performance.
NPU: AMD XDNA™ 2
50 TOPS of dedicated computing power, up to 126 TOPS of total AI performance, optimized specifically for on-device AI workloads.
RAM: 128 GB LPDDR5X-8000 unified memory
Extremely high bandwidth of 256 GB/s, supports dynamic sharing, and can allocate up to 96 GB of video memory.
System disk: PCIe 4.0 x, 1 × M.2 slot
Supports up to 8 TB capacity.
NAS: 5 SATA hard drive bays
Supports ZFS stripe, mirror, RAIDZ1 and RAIDZ2, providing more than 100 TB of raw capacity of local private cloud storage.
Wired network connection: 1000 Mbps network port
Enabling you to build a high-speed and secure home gateway.
WI-FI: WI-FI 7 high-speed wireless network
The MT7925 WiFi module supports 802.11be protocol and Bluetooth 5.4.
Physical size: 222.5 mm (W) × 222.5 mm (D) × 224 mm (H)
Power supply : Built-in 300 W industrial-grade power supply
No bulky adapter required; deeply optimized for AI scenarios and peak power consumption during startup.
A personal AI agent for work, life, and play
CEPi
A compact, plug-and-play local AI server
The AI assistant runs locally on the device. CEPi also provides private file storage, cross-device synchronization, and LAN file sharing. By default, user files and synchronization data remain on the device or local network, without relying on vendor-hosted cloud storage.
A compact mini PC with Hermes built in
Local AI for home and small offices
Local AI, ready out of the box
Hermes comes preinstalled, configured, and tested. From first boot, you can manage files, automate routine tasks, and work with information through natural
language. Core AI features run locally and remain available without an internet connection.
Sync and share across devices
CEPi provides local storage for documents, photos, media, and project files. Automatic synchronization and file versioning keep data consistent across your computers, while LAN sharing makes it available to authorized devices on the same network.
Efficient, always-on operation
CEPi’s compact, energy-efficient hardware is designed for continuous use at home or in the office. It keeps local storage, synchronization, and AI services available across your network whenever you need them.
Hermes built in
It comes pre-installed with the Hermes Smart Agent Platform, so it's ready to use right out of the box, requiring no installation or configuration.
Private local storage
Photos, documents, media, and project files are stored on CEPi by default, keeping your data under your control.
File sync and versioning
Automatic synchronization keeps selected folders up to date across your devices, while file versioning helps you recover earlier copies when needed.
Direct LAN sharing
Authorized devices on the same local network can access shared files directly, without routing them through a cloud storage service.
Automated file workflows
Powered by Hermes, CEPi can classify files, organize data on a schedule, and run rule-based workflows automatically.
Unified control with the Naohua app
Set up the device, manage files, configure synchronization, and create automation tasks from a single app.
OTHER
Cephalon machine display area
CEPHALON
Company and Leadership
distributed compute resources through a DePIN-based network.
Cephalon’s software portfolio includes the EndBrain Cloud Platform for professional generative AI workflows and Lucy for everyday AI interaction. Its proprietary hardware portfolio supports widely used foundation models across training and inference workloads, with
an emphasis on performance, cost efficiency, and flexible deployment. These products are now in commercial use.
Cephalon’s chief executive also serves as CTO and holds a PhD from the Hong Kong University of Science and Technology (HKUST). He has extensive experience in supercomputing and artificial intelligence, holds multiple patents, and has published at leading
international conferences. He has received industry technology awards, led major national research projects, helped build and operate supercomputing centers, and served as chief scientist at high-growth technology companies.