SlickGPT is a light-weight "use-your-own-API-key" ChatGPT client written in Svelte. It offers GPT-4 integration, a userless share feature and other superpowers.
"Auf den Entwickler kam allerdings direkt eine weitere Hürde hinzu: MS-DOS integriert nativ keinen Netzwerk-Stack oder eine entsprechende API. Die ist allerdings für den Zugriff auf ChatGPT zwingend notwendig."
DeepL:
"However, the developer was immediately faced with another hurdle: MS-DOS does not natively integrate a network stack or a corresponding API. However, this is absolutely necessary for accessing ChatGPT."
Golem is a very highly frequented German technology website.
As with other OpenAI based applications, I don't feel comfortable sharing my private API key with an application I don't know. Especially when the webpage isn't telling me what this app is about. I suggest adding a small demo video or at least adding some bullets on what this app is trying to solve.
I get that fear and it's fair. SlickGPT is OS, self-hostable and just runs locally, so you can check the code and run your own instance if you don't trust that ours running on Vercel is on-par with the one in the repo. You can also put a soft and hard cap on your API quota on the OpenAI page and monitor it to verify that only you are using it.
It's a self-hostable chat interface for the OpenAI GPT API.
From their Github Readme:
SlickGPT allows you to run your own local ChatGPT instance, host it yourself or just use our instance if you like. Users bring their own OpenAI API keys. SlickGPT offers them a very fancy user interface with a rich feature set like managing a local chat history (in the localStorage), a userless "Share" function for chats, a prominent context editor, and token cost calculation and distribution.
> SlickGPT allows you to run your own local ChatGPT instance
I find it interesting that so many projects claim to be "run your own ChatGPT", when they're in fact "run your own web UI". Are people really so ignorant of how this works that they're equating these concepts, or believing that the web UI is anything more than a razor thin veneer?
I guess maybe people use ChatGPT to refer to the interface, and GPT-N to refer to the models, but that's not very accurate given the amount of tuning, it's much more accurate to say that ChatGPT is a productised GPT instance, with a web UI.
If you're using a third-party UI for ChatGPT, it might as well be a browser extension [0] or a native app [1]. It really shines when you break it out of its website interface and start using it like a command palette/context menu.
I'm beginning to think there's one too many of these kinds of apps, but if anyone is interested in a cross-platform (Mac, Linux, Windows) alternative, I've built a command-palette type interface [1] to GPT that supports code blocks/LaTeX too.
Agreed! I've found that staying in my workflow leads to fewer inadvertent context switches (like ChatGPT in-browser -> HN -> ... -> rabbit hole). MacGPT looks great. I created a similar plugin for my Windows keystroke launcher (think Alfred/Spotlight) for the same reason. If anyone's interested, you can check it out here: https://github.com/dkgv/pinpoint
It's a userfriendly web client for the latest OpenAI chatCompletion API with additional features compared to the official chat.openai.com client (mainly: "Share chats" & better param customization), it's Open Source and can be run locally or self-hosted easily. It's built in one of the most popular frontend tech stacks of the last years. Not more and not less than that.
This. There is very noticeable latency when typing. This is on a very powerful desktop PC using Chrome. If you hold down a character on your keyboard it shows the issue more clearly, the input will sporadically appear every few seconds.
Background was that we wanted to explore providing access to ChatGPT to employees at $company, but restrict access with Google OAuth using company email, and use the company OpenAI API key on the back. So I hacked together something quick that matches our requirements and we can self-host.
It's very slow when typing characters in the input field. It feels like a react application that's doing a bunch of component re-rendering on every keypress.
Readit News
No, it doesn't. If it did, it would be a 700GB+ download and not need OpenAI API keys. Wish folks would stop saying it.
https://www.golem.de/news/ki-auf-dem-intel-8088-entwickler-b...
"Entwickler bringt ChatGPT auf MS-DOS zum Laufen"
in English:
"Developer gets ChatGPT to run on MS-DOS".
"Auf den Entwickler kam allerdings direkt eine weitere Hürde hinzu: MS-DOS integriert nativ keinen Netzwerk-Stack oder eine entsprechende API. Die ist allerdings für den Zugriff auf ChatGPT zwingend notwendig."
DeepL:
"However, the developer was immediately faced with another hurdle: MS-DOS does not natively integrate a network stack or a corresponding API. However, this is absolutely necessary for accessing ChatGPT."
Golem is a very highly frequented German technology website.
what they are doing: adding a prompt
> ChatGPT is a sibling model to InstructGPT
OpenAI's own words
Deleted Comment
The key is passed (see https://github.com/ShipBit/slickgpt/blob/main/src/lib/ChatIn...) to a Vercel edge function (see https://github.com/ShipBit/slickgpt/blob/main/src/routes/api...) that does the actual call to the OpenAI API. This function is stateless and doesn't store anything. I also put it in your localStorage using a store (see https://github.com/ShipBit/slickgpt/blob/main/src/misc/store...) so that you don't have to enter it over and over again.
As for the "what is this trying to solve" - read our documentation in the GitHub repo over at https://github.com/ShipBit/slickgpt.
From their Github Readme:
SlickGPT allows you to run your own local ChatGPT instance, host it yourself or just use our instance if you like. Users bring their own OpenAI API keys. SlickGPT offers them a very fancy user interface with a rich feature set like managing a local chat history (in the localStorage), a userless "Share" function for chats, a prominent context editor, and token cost calculation and distribution.
https://github.com/ShipBit/slickgpt
I find it interesting that so many projects claim to be "run your own ChatGPT", when they're in fact "run your own web UI". Are people really so ignorant of how this works that they're equating these concepts, or believing that the web UI is anything more than a razor thin veneer?
I guess maybe people use ChatGPT to refer to the interface, and GPT-N to refer to the models, but that's not very accurate given the amount of tuning, it's much more accurate to say that ChatGPT is a productised GPT instance, with a web UI.
So this seems a bit paranoid.
[0] https://sublimegpt.com
[1] https://macgpt.com
[1] https://github.com/JinayJain/gpt-anywhere/releases/latest
Deleted Comment
https://t.me/god_in_a_bot
Background was that we wanted to explore providing access to ChatGPT to employees at $company, but restrict access with Google OAuth using company email, and use the company OpenAI API key on the back. So I hacked together something quick that matches our requirements and we can self-host.
Deleted Comment