GroqMind

AI chat assistant with streaming responses, conversation memory, and persona switching — powered by Groq.

Live Demo: chat.zuhdi.id

Features

• Streaming Chat — Real-time token-by-token responses via Server-Sent Events (SSE)
• Conversation Memory — Multi-turn context within sessions
• Persona Switching — General, Code Helper, Business Advisor, Bahasa Indonesia
• Model Switcher — Llama 3.1 8B (fast) or Llama 3.3 70B (powerful)
• Markdown Rendering — Syntax-highlighted code blocks with copy button
• Token Usage — Live token counter per session
• Security — CORS protection, XSS prevention, session limits, input validation

Tech Stack

Layer	Technology
Backend	Python 3.12, FastAPI
AI	Groq SDK (Llama 3.x)
Streaming	SSE (sse-starlette)
Frontend	Vanilla JS, Tailwind CSS
Deploy	uv, Cloudflare Tunnel

Quick Start

# Set your Groq API key
echo "GROQ_API_KEY=gsk_your_key_here" > .env

# Install and run
uv sync
uv run groqmind

Server starts at http://localhost:6002

Environment Variables

Variable	Description	Required
GROQ_API_KEY	Your Groq API key (console.groq.com)	Yes

API Endpoints

Method	Endpoint	Description
GET	/	Chat UI
POST	/chat	Send message (SSE streaming response)
GET	/health	Health check

How It Works

1. User selects a persona and model from the sidebar
2. Messages are sent via POST to /chat with session context
3. Groq API streams tokens back via SSE
4. Frontend renders markdown in real-time with syntax highlighting
5. Conversation history is maintained per session for multi-turn context

License

MIT