Groq API Chatv1

A blazing-fast, context-aware chatbot built with Python, Streamlit, and Groq-hosted LLMs. Designed for rapid, multi-model experimentation and professional chat UX.

Features

• Multi-model support:
  • meta-llama/llama-4-maverick-17b-128e-instruct
  • meta-llama/llama-4-scout-17b-16e-instruct
  • gemma2-9b-it
  • llama-3.1-8b-instant
• Streaming responses for real-time, token-by-token output
• Per-model chat history (switch models, keep your context)
• Temperature adjuster (slider, 0.0–1.5) for creative vs. focused responses
• Download chat as JSON (user + assistant turns)
• Average response time tracker
• Modern, dark-themed UI with avatars, chat bubbles, and sidebar controls
• Clear all chat histories button
• Persistent context (history is saved and restored per model)
• Error handling for API, rate limits, and timeouts

---

Installation

1. Clone the repo:

   git clone https://github.com/0xnomy/groq_chatbot
   cd groq_chatbot

2. Install dependencies:

   pip install -r requirements.txt

3. Add your Groq API key:
   • Create a .env file in the project root:

     GROQ_API_KEY=your-groq-api-key-here

4. Run the app:

   streamlit run app.py

---

Usage

• Select a model from the sidebar (each model has its own chat history)
• Adjust temperature (creativity) with the slider
• Type your message and press Enter
• See responses stream in real time
• Download your conversation as JSON with the download button
• Clear all chat histories with the sidebar button
• Switch models to compare answers or continue previous chats

---

Supported Models

• meta-llama/llama-4-maverick-17b-128e-instruct – Balanced, strong generalization (Meta)
• meta-llama/llama-4-scout-17b-16e-instruct – Fast, low-latency (Meta)
• gemma2-9b-it – Safe, helpful, instruction-tuned (Google DeepMind)
• llama-3.1-8b-instant – Compact, high-speed (Meta)

---

How Context & History Work

• All chat context is managed by the app (not the model)
• Each model has its own chat history (saved as chat_history_<model>.json)
• When you send a message, the full history is sent to the model for context-aware replies
• Switching models loads the last conversation for that model

---

Controls

• Temperature:
  • Lower = more focused, deterministic
  • Higher = more creative, varied
• Export:
  • Download your current conversation as a JSON file (groqchatv1_history.json)

---

Links

• [Groq API] (https://console.groq.com/docs/)