读取页面🐠meeeyo.com

Here's the thing you need to know before you touch ReadWebNode: it is not a generic web scraper. The name and the Chinese display name (读取页面, "read page") suggest you hand it any URL and get the content back. You don't. You hand it an instruction, and it builds a URL pointing at the author's own server - https://www.meeeyo.com/u/getnode/<instruction>.php - fetches that, and returns the response body. If the page you wanted isn't hosted by the author, this node will not get it for you.

I want to be plain about this because it's the kind of node people install in bulk from a pack without reading the source. The mechanism is: your input is lowercased, jammed into a fixed URL on meeeyo.com, and requests.get runs against it with a randomly generated Authorization header. It's a remote content fetch, not a reader. Two implications follow.

First, it requires internet and the author's server to be up. The failure mode is a returned string that says (in Chinese) "Error! Failed to parse, please check and retry" - which tells you nothing about what went wrong.

Second, and more importantly: every run of this node sends a request to the author's server. Given the ComfyUI ecosystem's history with custom nodes that silently phone home (the LLMVISION malware incident of 2024 lives on in every security discussion), you should be aware of what a node is doing before you wire it into a graph you run unattended. This one is transparent if you read the source - the URL is hardcoded but not hidden - and it's plausibly just how the author serves workflow content to their own users. But "node fetches from the author's domain with a fake bearer token" is not something to install blindly.

What you'd actually use it for

The two inputs are Instruction (the key that becomes part of the URL) and prefix_suffix (a prefix|suffix wrapper - split on the first |; the prefix is prepended and the suffix appended to whatever comes back). Output is a single STRING.

Honestly? Unless you're part of the author's ecosystem - the pack targets Chinese users via bilibili and WeChat, and this node looks like a delivery mechanism for the author's own prompt/content library - there's no compelling reason to reach for it. If you want to read an arbitrary web page into ComfyUI, a general HTTP node from a more established pack will serve you better and won't route your traffic through a third party.

Installing it

The usual: ComfyUI Manager → ComfyUI_StringOps, or

cd ComfyUI/custom_nodes
git clone https://github.com/MeeeyoAI/ComfyUI_StringOps.git

Restart. The requests dependency it needs is already in the pack's requirements.txt and is almost certainly already in your ComfyUI environment.

The bottom line

I'll give it this: it's honest about its failure modes (no crash, just an error string) and the dependency footprint is trivial. But the gap between what the name promises ("read page") and what the code does ("fetch from the author's domain") is big enough that I'd only install it if you already run the rest of this pack and know exactly what meeeyo.com/u/getnode/ is supposed to return. For everyone else, treat it as a red flag worth a glance at the source before you trust it with anything.