I have been building tons of applications. My latest one is Gitroom. It helps you to:
(it's open-sourced 🙏🏻)That's nice and all, but here is the real kicker.
I see tons of AI libraries every day, and most of them are really cool, but I ask myself, will they help me in real life?
Yes, babyagi might be the next big thing. But honestly, I don't see anything I can do with it.
So what can you do with AI, and what do people usually do?
I think in today's ChatGPT world and local LLMs, the use case is pretty simple:
Auto-complete content - perfect for Gitroom to let you generate additional information or use something like GitHub copilot.
Automation - You pass content to the LLM, which returns a JSON with operations you can perform.
Function calling - Mostly with ChatGPT, you pass different functions to the LLM, and they decide which one to run.
Miscellaneous - Classification, translation, etc.
I think "function calling" is not leveraged enough because adding it in a way that makes sense is complex.
I admit I love to build stuff, but I sometimes skip developing a good UX, which causes people to leave the system because they don't know what to do.
I think inherently, if you build systems for developers, you can get away with it - because they are more sophisticated.
But that is mostly not the case. You should have:
This is where AI can save you without writing too much code.
So, one of the things more and more companies offer is a chatbot that can interact with your data and perform actions for the user.
For example, I can tell the chatbot I want to schedule a new post for Gitroom at 7:00 p.m. US time.
And it will do it for me.
The problem with most of the solutions I found is that they are:
So, I found this incredible open-source library licensed under MIT - which is fantastic.
It's just a chatbot sitting on your right sidebar.
You feed it with:
Once you have finished doing that, it's just like magic. Does somebody want to schedule a post?
It will trigger the schedulePost
function. Of course, it uses the ChatGPT "function calling" in the background, but it simplifies everything.
I think it fits any app. You can use it with big LLMs like ChatGPT / Gemini, but also with smaller ones like LLAMA2 and Mistral.
I started a new YouTube channel mostly about Tech, open-source, and marketing :)
If that's something that interests you, feel free to subscribe to it here: https://www.youtube.com/@nevo-david
Also published here.