iranee/fnos-hermes-agent — explained in plain English
Analysis updated 2026-05-18
Run a personal AI chat assistant directly on your fnOS NAS with no external server needed.
Connect the assistant to messaging apps like WeChat, Telegram, Discord, or Slack so you can chat with it from your phone.
Send images and files to the assistant for analysis without leaving the NAS.
| iranee/fnos-hermes-agent | ddekadt/my580_agentic_ai | drobiazkin/ai-agent-architecture | |
|---|---|---|---|
| Stars | 24 | 24 | 24 |
| Language | HTML | HTML | HTML |
| Setup difficulty | moderate | easy | easy |
| Complexity | 3/5 | 1/5 | 3/5 |
| Audience | ops devops | researcher | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires an fnOS NAS device and an API key from a supported model provider.
Hermes Agent for fnOS is an AI assistant app built specifically to run on fnOS, a NAS (network storage) operating system made by a company called Feiniu. It installs as a native app through fnOS's own app center, using a package format called fpk, so you do not need to set up a separate server or open extra network ports yourself. Once installed, you manage and chat with the assistant through a web based control panel. In that panel you pick which AI model provider to use, options include OpenRouter, OpenAI, Anthropic, Kimi, and MiniMax, and enter an API key for that provider. The chat interface supports full Markdown formatting, lets you upload images and files for the assistant to look at, and remembers context across separate conversation sessions. It can also turn a web link into a QR code so you can open it quickly on a phone. Beyond the web chat, the agent can connect to outside messaging apps such as WeChat, Telegram, Discord, Slack, QQ, DingTalk, and Feishu, so you can talk to it from those platforms too. Behind the scenes, a small program called Monitor, built with the Bun JavaScript runtime, manages the whole system. It talks to the web control panel over a local Unix socket and forwards chat requests to a separate Hermes Gateway process, which is the part that actually calls the AI model providers. The assistant itself runs in a Python environment managed by a tool called uv, with its packages typically downloaded from an Alibaba Cloud mirror and falling back to GitHub if needed. A randomly generated token is created each time the app starts, and any action that changes settings or restarts a process must include that token, while simply viewing status or logs does not require it. The app needs about 1 gigabyte of free storage, and its data, including configuration files, chat history, and downloaded skills, is stored on the NAS itself so it persists across restarts and updates. Updates are checked from the control panel and can be installed with one click. The project is released under the MIT license.
An AI assistant app built specifically for fnOS NAS devices, installed as a native app with a web control panel for chatting with multiple AI providers.
Mainly HTML. The stack also includes Bun, Python, uv.
MIT license: free to use, modify, and distribute for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly ops devops.
This repo across BitVibe Labs
Verify against the repo before relying on details.