Howdy Comrades! Just wanted to show off an app I just started using. This LLM app uses an array of open source AI models to respond to your questions. Then its super secure through the assist app found on F-Droid. In settings you can set it to access the web if you want.
My question is how safe it to use with the web search enabled and have you tried this app before?
Side note*** Its unlikely I’ll end up using this as my LLM. I prefer locally run. But locally run has harder setup and response times. I’m studying this app because I teach. There’s people who heavily use LLM’s and those people are unlikely to move to the FLOSS movement and make the radical changes needed if they lose their LLM and other key apps that they like. Once they get to where they like open source LLM’s over proprietary then you can start introducing then to locally run LLM’s.
Do you know a better one aside from locally run?
You misunderstood what the app is, it’s only a webview (a stripped down webbrowser). Its setting for webaccess is whether it will load websites outside of hugging chat in the webview.
I went back and checked. You are correct. The app operates as a “secure” web wrapped blocking nonessential stuff. Though it was much easier, not as easy as it should, to setup security and privacy on this LLM. I think I made too much of the simple security additions of the web wrapper. I thought it applied default settings in it’s own or something. I don’t know there’s a lot of these things now and I thought this one looked promising.
Sides stepping the web problem, you could set it up on a local server with a vpn and effectively have a local LLM.
You state that the app is super secure and then ask how safe it is?
The app is very secure. By default web access is turned off. My question was how safe is it with web access turned on. Read my question again. My question, in case you didn’t understand the first time, is how secure is web access and what risks am I running with that feature enabled.
