Reflection Agent Service 🤖✨

This project is an AI microservice built around the Reflection agent design pattern to iteratively analyze, critique, and refine Large Language Model (LLM) outputs. It showcases the end-to-end engineering process of transforming a core AI script into a robust, containerized, and production-ready service.

🧠 Core Concept: The Reflection Pattern

The Reflection Pattern enhances the quality and reliability of LLM outputs through a structured, three-step self-critique process:

1. Generate: The agent produces an initial draft in response to a prompt.
2. Reflect: The agent analyzes its own draft, identifies flaws or areas for improvement, and generates a list of actionable, constructive critiques.
3. Refine: The agent re-attempts the original task, this time using its own critiques as a guide to generate a superior final output.

This cycle mimics the human process of drafting and revision, leading to responses that are more coherent, accurate, and aligned with the user's intent.

🚀 Engineering & MLOps Highlights

This project emphasizes the engineering required to serve an AI model reliably and scalably.

• Microservice API: The agent's logic is exposed via a RESTful API using FastAPI, with clear data contracts enforced by Pydantic for robust I/O validation.
• Containerization: The entire application is containerized with Docker, ensuring a consistent execution environment and simplifying deployment across any platform.
• Unit Testing: The agent's core business logic is rigorously tested using Pytest and pytest-mock, guaranteeing the reliability and integrity of each component.
• Automated CI/CD: A GitHub Actions pipeline is triggered on every push to main, automatically performing:
  • Linting with Ruff to enforce code quality and style consistency.
  • Unit Testing to prevent regressions and ensure code health.
• Secrets Management: API keys and sensitive credentials are handled securely using .env files for local development and GitHub Actions Secrets in the CI/CD pipeline, preventing any exposure in the repository.

🏗️ Service Architecture

graph TD
    A[User/Client] -->|HTTP POST Request| B(FastAPI Service);
    B -->|prompt| C{ReflectionAgent};
    C -->|1. Generate Draft| D[OpenAI API];
    D -->|Draft| C;
    C -->|2. Generate Reflections| D;
    D -->|Reflections| C;
    C -->|3. Generate Final Output| D;
    D -->|Final Output| C;
    C -->|Complete Response| B;
    B -->|JSON Response| A;

Loading

🏁 Getting Started

Prerequisites

• Git
• Python 3.9+
• Docker Desktop (running)

1. Running Locally (for Development)

1. Clone the repository:

   git clone [https://github.com/PRYSKAS/REFLECTION_PATTERN_AGENT.git](https://github.com/PRYSKAS/REFLECTION_PATTERN_AGENT.git)
   cd REFLECTION_PATTERN_AGENT

2. Set up the environment:

   • Create a .env file from the example: copy .env.example .env (on Windows) or cp .env.example .env (on Unix/macOS).
   • Add your OPENAI_API_KEY to the new .env file.

3. Install dependencies:

   pip install -r requirements.txt
   pip install -e .

4. Run tests to verify the setup:

   pytest

5. Start the API server:

   uvicorn main:app --reload --port 8001

   The API will be available at http://127.0.0.1:8001/docs.

2. Running with Docker (Production Mode)

This is the recommended way to run the service for a stable, isolated deployment.

1. Build the Docker image:

   docker build -t reflection-agent-service .

2. Run the container:

   docker run -d -p 8001:8001 --env-file .env --name reflection-agent reflection-agent-service

   The service will now be running in the background. Access the API documentation at http://127.0.0.1:8001/docs.

📡 API Endpoint

POST /run

Executes the agent's full generate-reflect-generate cycle.

Request Body:

{
  "prompt": "Write a tweet about the importance of CI/CD in AI engineering."
}

Success Response (200 OK):

{
  "initial_draft": "CI/CD is crucial in AI engineering. #AI #MLOps",
  "reflections": [
    "- The tweet is too short and generic.",
    "- It could add a specific benefit, like 'accelerating value delivery'.",
    "- An emoji would increase engagement."
  ],
  "final_output": "🚀 CI/CD in AI Engineering isn't a luxury; it's a necessity! It accelerates value delivery by automating testing and deployment, ensuring robust models reach production faster. #MLOps #AIEngineering"
}