title	emoji	colorFrom	colorTo	sdk	sdk_version	app_file	pinned	license	header	fullWidth	short_description	models
Chat with DuckDuckGo Agent	🦆	yellow	red	gradio	4.36.1	app.py	true	apache-2.0	mini	true	Chat llama-cpp-agent that can search the web.	mistralai/Mistral-7B-Instruct-v0.3 meta-llama/Meta-Llama-3-8B-Instruct

Chat with DuckDuckGo Agent 🦆

A Gradio-based web interface that allows users to chat with an AI agent capable of searching the web using DuckDuckGo. The agent uses llama.cpp to run various open-source language models locally.

Features

• Web search capabilities using DuckDuckGo
• Support for multiple LLM models:
  • Mistral 7B Instruct v0.3
  • Mixtral 8x7B Instruct v0.1
  • Meta Llama 3 8B Instruct
• Real-time chat interface
• Source citation for responses
• Customizable model parameters
• Dark mode support

Technical Details

• Framework: Gradio 4.36.1
• Models: Uses GGUF quantized models
• Agent: Powered by llama-cpp-agent
• Web Scraping: Uses trafilatura for content extraction
• Context Window:
  • 32k tokens for Mistral and Mixtral
  • 8k tokens for Meta Llama 3

Usage

The chat interface provides several customizable parameters:

• Model selection
• System message
• Max tokens (1-4096)
• Temperature (0.1-1.0)
• Top-p (0.1-1.0)
• Top-k (0-100)
• Repetition penalty (0.0-2.0)

Example Queries

- "latest news about Yann LeCun"
- "Latest news site:github.blog"
- "Where I can find best hotel in Galapagos, Ecuador intitle:hotel"
- "filetype:pdf intitle:python"

Installation

1. Clone the repository:

git clone https://github.com/pabl-o-ce/ddg-web-search-chat
cd ddg-web-search-chat

2. Install dependencies:

pip install -r requirements.txt

3. Download the required model files: The application will automatically download the following models:

• Mistral-7B-Instruct-v0.3-Q6_K.gguf
• Meta-Llama-3-8B-Instruct-Q6_K.gguf
• mixtral-8x7b-instruct-v0.1.Q5_K_M.gguf

4. Run the application:

python app.py

Configuration

The application can be configured through various parameters in settings.py:

• Model context limits
• Message formatter types for different models
• System prompts and templates

UI Customization

The interface uses a custom theme with:

• Orange primary color
• Amber secondary color
• Dark mode support
• Custom CSS for message bubbles and layout

License

Apache 2.0

Links

• Discord Community
• GitHub Repository

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

---

Note: This project is powered by llama-cpp-agent and uses Hugging Face Spaces for deployment.