🤖 CrewAI Learning Project — Automated Research & Article Generator

A simple multi-agent AI system built using CrewAI that automatically:

• Researches a topic from the web 🌐
• Extracts key insights 🧠
• Generates a structured article ✍️

This project demonstrates how agent-based AI workflows can replace single-prompt LLM systems for better structure, modularity, and scalability.

---

📁 Project Structure

crewai-learning/
│
├── config/
│   ├── agents.yaml        # Agent definitions (role, goal, backstory)
│   └── tasks.yaml         # Task definitions (description, output)
│
├── src/
│   ├── agents.py          # Loads agents from YAML
│   ├── tasks.py           # Loads tasks from YAML
│   ├── tools.py           # Serper search tool setup
│   └── crew.py            # Crew orchestration
│
├── outputs/
│   └── output.txt         # Final generated article
│
├── screenshots/
│   └── example_run.png
│
└── requirements.txt

---

🚀 What This Project Does

Given a topic like:

AI in Healthcare

The system automatically:

1. Searches the internet for relevant information
2. Extracts important trends and insights
3. Structures the information
4. Writes a clean, readable article

---

🔄 Workflow

User Topic
   ↓
Research Agent (Web Search via Serper)
   ↓
Structured Research Output
   ↓
Writer Agent (Gemini LLM)
   ↓
Final Article (Markdown/Text)

---

🤖 Agents

🔍 Research Agent

• Searches the web for relevant information
• Identifies trends, risks, and opportunities
• Collects structured insights

Tool used: Serper API (Google Search)

---

✍️ Writer Agent

• Uses research output
• Converts findings into a readable article
• Formats output in markdown/text

Tool used: Gemini 2.5 Flash LLM

---

⚙️ Configuration (YAML-Based)

Instead of hardcoding agents and tasks in Python, this project uses YAML configuration files.

Why YAML?

• Cleaner code
• Easier to update prompts
• No need to modify Python logic
• Better scalability for new agents

---

🧠 Example: Agents (config/agents.yaml)

researcher:
  role: Senior Researcher
  goal: Research and analyze {topic}
  backstory: Expert in identifying tech trends and insights

writer:
  role: Tech Writer
  goal: Write engaging articles about {topic}
  backstory: Skilled in simplifying complex technical ideas

---

📄 Example: Tasks (config/tasks.yaml)

research_task:
  description: Research latest trends in {topic}
  expected_output: Structured research report

write_task:
  description: Write article using research findings
  expected_output: Markdown article with clear sections

---

🛠 Tech Stack

• 🧠 CrewAI — Multi-agent orchestration
• 🔎 Serper API — Google Search integration
• 🤖 Google Gemini — LLM for generation
• 🐍 Python — Core language
• 🔧 PyYAML — Configuration handling
• 📦 python-dotenv — Environment variable management

---

⚙️ Setup Instructions

1. Clone Repository

git clone <your-repo-url>
cd crewai-learning

---

2. Install Dependencies

pip install -r requirements.txt

---

3. Setup Environment Variables

Create a .env file:

GEMINI_API_KEY=your_gemini_api_key
SERPER_API_KEY=your_serper_api_key

---

4. Run the Project

python src/crew.py

---

📤 Output

Final generated article is saved in:

outputs/output.txt

It includes:

• Summary of topic
• Key trends
• Market insights
• Future outlook

---

💡 Why This Project Matters

Traditional LLM apps:

• Single prompt → single response
• Limited reasoning structure
• Hard to scale workflows

This project demonstrates:

✔ Multi-step reasoning ✔ Agent-based decomposition ✔ Tool usage (search + LLM) ✔ Real workflow orchestration

---

🧭 Future Improvements

• Add Fact Checker Agent
• Add SEO Optimization Agent
• Switch fully to @CrewBase architecture
• Add vector database memory
• Deploy using FastAPI
• Add UI using Streamlit or Next.js

---

👨‍💻 Author

Built as a learning project for understanding CrewAI and agentic AI systems.