Multi-Agent Crew Service 🤖🤝📝

An AI microservice that implements the Multi-Agent Collaboration pattern. This system orchestrates a crew of specialized AI agents, each with a distinct role and set of tools, to collaboratively solve complex problems, such as writing a comprehensive blog post from a single topic.

This project represents the synthesis of multiple agentic design patterns (Tool Use, Reflection) into a distributed system architecture.

🧠 Core Concept: The Agentic Organization

Instead of a single, monolithic "do-it-all" agent, this architecture is built on the principle of specialization, creating a digital team:

1. Specialized Agents: Each Agent is configured with a persona (role, goal, backstory) and tools specific to its function (e.g., a Researcher with access to search tools).
2. Defined Tasks: Each Task defines a clear objective to be executed by an agent with the corresponding role.
3. Orchestration (The Crew): The Crew class acts as a project manager. It executes tasks in a defined sequence, passing the output of one agent as the context for the next, ensuring a cohesive and collaborative workflow.

This model transforms problem-solving from a monolithic task into a pipeline of specialists, mirroring how high-performance human teams operate.

🚀 Engineering & AI System Design Highlights

This project demonstrates the ability to design and build complex, distributed AI systems.

• Microservice Architecture for AI: The system is designed with a clear separation of concerns (Agent, Task, Crew), making it modular, testable, and easily extensible.
• Inter-Agent Data Pipeline: The Crew manages a data pipeline where the output of one agent becomes the input for the next, enabling the incremental construction of a complex solution.
• Synthesis of Patterns: The project combines multiple patterns: the Researcher acts as a ToolAgent, while the Critic applies principles from the ReflectionAgent.
• Infrastructure as a Pattern: The entire engineering foundation (FastAPI, Docker, Pytest, GitHub Actions) was reused, proving the effectiveness of our "agent factory" and allowing for a singular focus on the AI logic.

🏗️ The Crew's Workflow

This project implements the key phases of an MLOps pipeline for custom model creation:

graph TD
    subgraph "Início do Processo"
        A["User Input: Blog Topic"] --> B("FastAPI Endpoint /generate");
    end

    subgraph "Crew Orchestration"
        B --> C{"Orchestrator Agent (Crew)"};
        C -- "Assigns Task" --> D["Agent: Research Manager"];
        D -- "Uses Tool" --> E["Tool: Internet Search"];
        E --> D;
        D -- Output --> F["Agent: Content Creator"];
        F -- Output --> G["Agent: SEO Expert"];
        G -- Output --> H["Agent: Critic/Editor"];
        H -- "Feedback/Refines" --> F;
        H -- "Final Output" --> I["Final Blog Post"];
    end

    subgraph "Fim do Processo"
        I --> B;
        B --> J["User Output: JSON Response"];
    end

Loading

🏁 Getting Started

Prerequisites

• Git
• Python 3.9+
• Docker Desktop (running)
• An OpenAI API Key (required for LLM calls and tool usage)
• A Serper API Key (for internet search tool, required for the Researcher Agent)

1. Setup Environment and API Keys

1. Clone the repository:

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

2. Install dependencies:

   pip install -r requirements.txt

3. Configure environment variables:
   • 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 and SERPER_API_KEY to the new .env file. These are crucial for the agents to function.

2. Running the Multi-Agent Service

1. Run locally using Uvicorn (for development):

   uvicorn serving.main:app --reload --port 8001

   Access the API documentation and interact with the service at http://127.0.0.1:8001/docs.

2. Run using Docker (Recommended for stable execution):

   • Build the Docker image:

     docker build -t multi-agent-crew-service .

   • Run the container:

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

   Access the API at http://127.0.0.1:8001/docs.

---

📡 API Endpoint

POST /generate

Initiates the multi-agent crew to generate content based on a given topic.

Request Body:

{
  "topic": "The future of AI in content creation"
}

### Success Response (200 OK):

{
  "blog_post": "..." // The complete blog post generated by the crew
}
---