A wild, interactive universe deployed on Vercel. Engineered with advanced state control, custom routes, and the Groq API because standard LLMs are too slow, and waiting is for humans.
- Client
- Furina CourtSelf-initiated project
- Year
- 2024Live on Vercel
- Role
- DeveloperFull-stack & AI
- Stack
- Node · PythonGroq API · Vercel
- Status
- LivePublicly accessible
Furina Court started as an experiment to bypass the sluggish response times of standard Large Language Models. I wanted an AI companion that could respond almost instantly, wrapped in a wild, custom-animated universe.
The project required engineering a highly responsive middle-layer. By integrating the Groq API, I achieved sub-second inference times, shifting the user experience from 'waiting' to 'interacting'.
I designed the interface with fluid layout animations, making the virtual playground feel alive. From route handling to state management, the architecture is tuned for immediate, lag-free updates.
An AI universe
that feels instant.
Standard LLMs take seconds to compile tokens. By utilizing Groq's high-speed LPU inference, Furina Court delivers text chunks back to the client at extreme speed. Latency is treated as a bug.
The custom routes handle user contexts and fast state caching. No loading indicators or spinning wheels — just immediate response.
Speed first,
style always.
The frontend design embraces fluid motion. User queries spawn instant animations that slide and snap into view, keeping the user locked into the interface.
Behind it, a clean Node and Python service routes inputs, manages prompt injection, and stream-pipes tokens directly to the client view.
Interactive 3 theatre chat.
Under the
hood.
Behind the UI, Furina Court utilizes a fast Node.js gateway and Groq API stream pipelines to feed real-time conversational states to the frontend.
- React
- Vite
- Tailwind
Responsive web client featuring custom spring animations and real-time state listeners.
- Node.js
- Express
REST API managing prompt templates, rate-limiting, and streaming response chunks.
- Groq LPU
- Llama 3
Ultra-fast inference engine delivering conversational outputs under 300ms.
Metrics of
speed and flow.
Created by a lone wolf.
- Engineering & Design
- Zenith Soumya
- Full-stack developer · AI architect
Zenith Soumya is an IoT & AI Developer. He designs and builds next-gen web applications, Java/DSA optimized tools, and micro-controller IoT systems.
Furina Court is live on Vercel, representing a core research project into high-speed LLM gateway architectures.