Koa-Chatbot

Designed and implemented Koa, a machine-learning chat assistant featuring a FastAPI inference backend, configurable JSON-based system prompts, and a responsive light/dark-mode web UI with profile controls and voice input, delivering fast, streaming-style conversational responses and a clean path to web deployment.

Features

• FastAPI inference server with a simple chat API
• Streaming-friendly chat flow (designed for responsive UX)
• Responsive frontend with light/dark mode
• Voice input support in the web UI
• JSONL-based sample data for quick testing

Repo Structure

• frontend/ — Web UI (HTML/CSS/JS)
• inference/ — FastAPI server + chat logic
• data/ — Small sample data (JSONL)
• requirements.txt — Python dependencies

Demo

• Watch the demo video

UI Screenshot

Colab Demo

Prerequisites

• Python 3.10+ (recommended: 3.11)
• (Optional) NVIDIA GPU + CUDA for faster inference

Quick Start (Local)

1) Backend (FastAPI)

pip install -r requirements.txt
python -m uvicorn inference.api_server:app --host 127.0.0.1 --port 9010 --reload