#reflection #demo #agents #llms #ai

Reflection Pattern Demo with OpenAI Agents

This repository demonstrates the use of two agents to showcase the Reflection Pattern: one agent that generates code, and another agent that reflects on the generated code and suggests improvements.

Project Overview

• Agent 1 (Generation Agent): This agent generates code based on prompts or specifications.
• Agent 2 (Reflection Agent): This agent reviews the code produced by the Generation Agent and provides suggestions for improvements.

This pattern mimics a cyclical process where one agent produces something and the other agent iterates upon it, improving the result over several steps.

---

Installation

Before running the project, you'll need to install the required packages. You can install the dependencies by using the requirements.txt file.

1. Clone the repository:

   git clone https://github.com/your-username/your-repo-name.git
   cd your-repo-name

2. Install dependencies using pip:

   pip install -r requirements.txt

The necessary packages to install are:

• openai: For interacting with the OpenAI API.
• pydantic: For data validation and settings management.
• python-dotenv: For loading environment variables.

---

Usage

To run the demo, execute the main.py file with the following command:

python main.py --model <MODEL_NAME> --temperature <TEMPERATURE_VALUE> --num_steps <NUM_STEPS>

Arguments:

• --model (<MODEL_NAME>): The name of the OpenAI model you wish to use for the generation and reflection agents (e.g., gpt-4o).
• --temperature (<TEMPERATURE_VALUE>): Controls the randomness of the responses. A value between 0 and 1. Lower values make the output more deterministic, while higher values introduce more creativity and variability. Example: 0.7.
• --num_steps (<NUM_STEPS>): The maximum number of reflection iterations the reflection agent will perform. Example: 3.

---

Example

Here's an example of how to run the demo with the gpt-4o model:

python main.py --model gpt-4o --temperature 0.7 --num_steps 3

---

How It Works

1. Generation Agent: The first agent generates code based on a given prompt or problem specification.

2. Reflection Agent: The second agent reviews the generated code, identifies potential areas for improvement (such as optimization, readability, etc.), and suggests improvements.

3. Iteration: The agents interact for a predefined number of steps (num_steps), where the reflection agent iterates on the output of the generation agent, refining the code further.

This interaction creates a more refined and well-optimized solution over time.

---

Requirements

• Python 3.7 or higher
• The following Python packages:
  • openai
  • pydantic
  • python-dotenv

These packages are specified in the requirements.txt file.

---

Environment Variables

Make sure to set up the necessary environment variables for the OpenAI API key. You can create a .env file in the root directory with the following content:

OPENAI_API_KEY=your-openai-api-key

Replace your-openai-api-key with your actual OpenAI API key.

---

License

This project is licensed under the MIT License - see the LICENSE file for details.