Good-Looking Chat-UIs to locally run on your computer

Posted by Crazyscientist1024@reddit | LocalLLaMA | View on Reddit | 17 comments

I want to ask about what good UIs people have to run locally on your computer. I personally use [Jan UI](https://github.com/janhq/jan), but I would be curious about good alternatives. I know about sites such as [BetterGPT](https://bettergpt.chat/), but being honest, their UI is just even painful for me to even look at. Any good alternatives?

Reply to Post

17 Comments

[-]

Tanvir1337@reddit

Open WebUI is the best

[-]

askgl@reddit

May I suggest Msty [1] I am one of the developers working on it and curious to see what you think of it. We have some fresh ideas that you might like and don't find in other apps such as split chats, no-restriction branching, RAG with Obsidian, etc. Since you have used other apps, if you have any suggestions on how to improve it, I would appreciate them. 1: https://msty.app

[-]

randomfoo2@reddit

Giving Msty a spin and it's pretty nice. I was curious, it looks like it's a free app, but not open source - what's the catch? One suggestion - I see that it tries to load ollama when it starts but you can setup an OpenAI compatible endpoint as well. Since I run a local AI machine on the network, that's great but I noticed that when you switch it doesn't automatically shut down the ollama service (so it' just in memory), but also, what would be convenient for OpenAI endpoint is if it tried to query /v1/models and simply gave you the list of available models when it loads instead of having to set it up manually - I swap around models so this is actually a big paint point.

[-]

askgl@reddit

Why should there be a catch? ;) We love working on it at this time and down the road we are thinking of maybe introducing some paid features (Obsidian style) and have us work full time on it. Thanks for the suggestion. Using Ollama via OpenAI API compatibility is, dare I say, a side effect of just having Ollama supporting Open AI compatibility API. We are working towards just allowing users to connect to any Ollama instance running locally or remote servers and then allowing to manage models from within Msty (pulling, cloning, updating, deleting etc). So this for now is just a stop gap solution. The onboarding page will give you an option to either use embedded service or select a remote one. > I swap around models so this is actually a big paint point I hear ya! Please give us a couple of weeks when we fix this once and for all. Thank you for giving Msty a try and for your valuable feedback.

[-]

dylanger_@reddit

Open WebUI is nice, also supports being run remotely.

[-]

fingerthief@reddit

I have my own project I've been using for awhile, I won't say the UI is state of the art but the mobile PWA support etc...is nice imo. Also with all of the nice to haves in some form or another. * Local LLM support * Browser loaded LLM support * Any OpenAI formatted API support * All storage is local to the browser with basically no setup. Enter a port and go in docker. * experimental conversational mode * Basic file upload support (browser stored) * Vision support with maintained context The idea being it's basically dummy proof to deploy or setup while being generally full feature and nice looking. https://github.com/fingerthief/minimal-chat Demo - https://youtu.be/bO78W8MPWgE?si=GW0vtbNq-JZs1eKG Public link if anyone wants to just try it out: https://minimalchat.app Mobile layout ![](https://preview.redd.it/7oxotwp8zs4d1.jpeg?width=1080&format=pjpg&auto=webp&s=df7771ccfc0ce7ff7695e5656f44a03cb9b332d2)

[-]

Hammer_AI@reddit

You might like my tool, https://www.hammerai.com! It's free and works on Mac, PC, and Ubuntu.

[-]

condition_oakland@reddit

I'm super happy with msty.app

[-]

remghoost7@reddit

I'm a fan of SillyTavern. It's simple, clean, and has everything you could want to edit at your fingertips. Looks pretty decent on mobile as well. You can hide both panels on the sides if you want. And the panel on the right can show all of your different "characters" as well, if you'd like. Plus, it has quick editing buttons on each message. The menu to the left of the textbox also has various features like regenerate/continue/etc. It allows for custom backgrounds and custom icons for characters as well, if you're into that sort of thing. https://preview.redd.it/7th7o219tv5d1.png?width=1189&format=png&auto=webp&s=23c2276e576fdfd4cfdc350ddfb19ddbaa22bbbc

[-]

staberas@reddit

Using SillyTavern cause im working on a standalone waifu assistant

[-]

MrVodnik@reddit

For me it's Open-WebUI. You can point it to any OpenAI API compatible back-end, and it works. It lacks character cards if you're into RP, but other than that user experience is superb.

[-]

AmericanKamikaze@reddit

Jan, AnythingLlm, LM Studio, WebUi, Lobehub the list grows

[-]

AdHominemMeansULost@reddit

I'm trying to make one https://github.com/DefamationStation/Retrochat It's extremely bare-bones for now, but it does support markdown and it's a single file. Trying to keep it as neat as possible. No chat history saved after you close the app yet!

[-]

Ok-Excuse-4371@reddit

please list a few key features a chat ui needs to win you over.

[-]

redoubt515@reddit

In terms of UI, **LM-studio** and **open-webui + ollama** are my preferences. Jan would be on that list, I like the aesthetics, but they need to clarify/tone down their language/messaging. They over-promise and gloss over technical details in ways that mislead less knowledgeable users.

[-]

mrskeptical00@reddit

[https://github.com/AugustDev/enchanted](https://github.com/AugustDev/enchanted) for Apple devices.

[-]

Healthy-Nebula-3603@reddit

I am using terminal with llamacpp ;)