Dynamic Few-Shot Prompting for AI Agents

This repository demonstrates how to implement dynamic few-shot prompting for AI agents in customer support. The solution dynamically selects relevant examples based on user input, optimizing response accuracy and efficiency.

High Architecture Overview

Getting Started

Prerequisites

• Python 3.x
• HuggingFace account with an API token

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/dynamic-few-shot-prompting.git
   cd dynamic-few-shot-prompting
   

2. Install dependencies:

   pip install -r requirements.txt

3. Set up your HuggingFace API token:

   export HUGGINGFACEHUB_API_TOKEN=your_api_token

Usage

The step-by-step code implementation is explained in the blog post. Follow the instructions to set up embedding models, create vector stores, generate dynamic prompts, and get AI responses.

Contributing

Feel free to fork the repository, make improvements, and submit pull requests. If you encounter issues or have suggestions, open an issue.