WackyWidgets is a dark-themed, beautifully designed AI widget that delivers one thing exceptionally well: absolutely nothing useful.

Instead of answering questions seriously like a typical AI assistant, it goes completely off the rails — giving users random facts, absurd suggestions, and unfiltered sarcasm. It’s chaotic, savage, and kind of hilarious. Think of it as ChatGPT’s evil twin with a sharp tongue and zero desire to help.

It’s powered by the OpenRouter API, using the nvidia/llama-3.1-nemotron-ultra-253b-v1:free model, which has been carefully prompted to behave like a smart-ass with a flair for trivia and insults. Whether you ask it something technical, personal, or just say “hi,” you’re guaranteed to get roasted while being distracted by completely unrelated nonsense.

💡 What It Can (and Can’t) Do 🔥 RoastBot 9000 — The Main Widget Built on nvidia/llama-3.1-nemotron-ultra-253b-v1:free via OpenRouter

Replies with a roast every time — even if you’re just asking about the weather

Intentionally unhelpful: gives weird, incorrect answers on purpose

Drops bizarre trivia out of nowhere (like: “Did you know octopuses have three hearts? Unlike you — who has none.”)

Supports markdown and code-style formatting for fun delivery

🧑‍🎨 Clean & Simple UI Fully responsive and built with a slick black-and-white theme

Styled using TailwindCSS, with smooth animations and a super minimal layout

Has a fixed input bar at the bottom and a scrollable chat area — just like ChatGPT

Runs in dark mode by default, to give the chaos some class

🧠 Smart API Handling (Because the Bot Isn’t) Includes built-in handling for OpenRouter rate limits

Retries requests with exponential backoff to avoid overloading the API

Debounces input and queues calls to prevent user spam (because people will try)

🏆 Why It Fits the Challenge ✅ Humor & Wackiness WackyWidgets was made to be completely ridiculous — and it delivers. It’s a parody of AI assistants and a reminder that not everything smart needs to be serious.

✅ Best Use of OpenRouter API We use the free-tier nvidia/llama-3.1-nemotron-ultra-253b-v1 model in a very custom way — pushing its personality to the edge and implementing fallback logic to deal with rate limits smoothly.

✅ UI/UX Polish Despite being a chaotic little bot, the interface is clean, responsive, and fun to use. It feels modern, intentional, and easy on the eyes — even if the replies aren’t.

✅ Original & Unexpected It flips the concept of “useful AI” on its head. Instead of solving problems, it creates them — but in the funniest way possible.

🚀 Tech Stack Frontend: React + TypeScript + TailwindCSS

AI Backend: OpenRouter API

LLM: nvidia/llama-3.1-nemotron-ultra-253b-v1:free

Extras: Retry logic, API throttling, markdown support

If you’ve ever wanted to chat with an AI that feels like a mix of Deadpool, Wikipedia, and a snarky barista, WackyWidgets is it.

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for WackyWidgets

Built With

Share this project:

Updates

Submission history