I was tired of getting rate limited when using Anthropic models like Claude, so I decided to build my own custom chat UI. Not being a developer, what looked like a simple task quickly turned into a sisyphean endeavor, especially trying to fix CSS. But with the help of Claude and GPT-4, (almost) everything is possible now.
In my day-to-day use, I often use snippets of prompts that I have to copy-paste here and there, so I added them as variables (just clicking on a variable will insert it into the text field and it will expand during runtime).
Yesterday, I saw that Groq added llama3 support, and it is not only very good but also blazingly fast. You need to try it to believe it:
(it uses my API key on a free tier, so the worst that can happen is the rate limit, which is generous on Groq).