📰 News Reader AI Agent 🚀

A powerful multi-agent system built with CrewAI that automates the process of discovering, summarizing, and curating daily news reports on specific topics.

🔄 Workflow

The system employs a sequential multi-agent orchestration to transform a simple topic into a professional news briefing.

graph TD
    %% Styling Definitions
    classDef agent node font-weight:bold
    classDef hunter fill:#f9f,stroke:#333,stroke-width:2px,color:#000
    classDef summarizer fill:#bbf,stroke:#333,stroke-width:2px,color:#000
    classDef curator fill:#dfd,stroke:#333,stroke-width:2px,color:#000
    classDef file fill:#fff,stroke:#333,stroke-dasharray: 5 5,color:#000

    Start((Input: Topic)) --> Hunter

    subgraph "Phase 1: Discovery"
        Hunter[news_hunter_agent]:::hunter -->|SerperDev & Scrape| ContentHarvest[(content_harvest.md)]:::file
    end

    ContentHarvest --> Summarizer

    subgraph "Phase 2: Analysis"
        Summarizer[summarizer_agent]:::summarizer -->|Multi-tier Summarization| Summary[(summary.md)]:::file
    end

    Summary --> Curator

    subgraph "Phase 3: Editorial"
        Curator[curator_agent]:::curator -->|Final Assembly| FinalReport[(final_report.md)]:::file
    end

    FinalReport --> End((Final News Briefing))

    %% Additional styling
    style Start fill:#f5f5f5,stroke:#333
    style End fill:#f5f5f5,stroke:#333

🌟 Key Features

• Multi-Agent Orchestration: Specialized agents for high-quality news delivery.
• Automated Web Search: Uses SerperDev to find the most relevant and credible news articles.
• Intelligent Web Scraping: High-performance scraping using Playwright and BeautifulSoup to extract clean content from article pages.
• Hierarchical Summarization: Generates multi-tier summaries (Headline, Executive, and Comprehensive) for different reader needs.
• Publication-Ready Reports: Assembles a professional final report in Markdown format.

🏗️ Architecture

🤖 Agents

1. Senior News Intelligence Specialist (news_hunter_agent):
   • Goal: Discover and collect relevant, credible news articles.
   • Tools: SerperDevTool, scrape_tool.
2. Expert News Analyst (summarizer_agent):
   • Goal: Transform raw articles into clear, multi-level summaries.
   • Tools: scrape_tool.
3. Senior News Editor (curator_agent):
   • Goal: Curate summaries into a cohesive, professional narrative.

📋 Tasks

• Content Harvesting: Searching for articles, filtering by credibility/relevance, and extracting full text.
• Multi-Tier Summarization: Creating tweet-style headlines, executive summaries, and detailed analysis.
• Final Report Assembly: Combining all work into a premium daily briefing.

📁 Project Structure

.
├── config/
│   ├── agents.yaml      # Agent roles and backstories
│   └── tasks.yaml       # Task definitions and process flow
├── output/              # Generated markdown reports
├── .env                 # Environment variables (API Keys)
├── main.py              # Entry point & CrewAI orchestration
├── tools.py             # Custom scraping & search tools
└── pyproject.toml       # Project dependencies & metadata

🛠️ Setup & Installation

Prerequisites

• Python: 3.13 or higher.
• uv: A fast Python package manager (recommended).
• Playwright: For web scraping (will be installed automatically via dependencies).

1. Clone the repository

git clone <repository-url>
cd AI-Agents.News_Reader_Agent

2. Install dependencies

Using uv:

uv sync

3. Environment variables

Copy the example environment file and fill in your API keys:

cp .env.example .env

Fill in the following:

• OPENAI_API_KEY: Your OpenAI API key (for models like o4-mini-2025-04-16).
• SERPER_API_KEY: Your Serper.dev API key for web search.

4. Install Playwright browsers

uv run playwright install chromium

🚀 Usage

Run the agent simply by executing:

uv run python main.py

By default, the agent searches for "Trends of Korea-Japan couples". You can modify the input topic in main.py:

result = NewsReaderAgent().crew().kickoff(inputs={"topic": "Your Topic Here"})

📈 Output

Generated reports are saved in the output/ directory:

• content_harvest.md: Raw collected articles and metadata.
• summary.md: Multi-tier summaries for each article.
• final_report.md: The curated, publication-ready news briefing.

🧰 Tech Stack

• Framework: CrewAI
• LLM: OpenAI (o4-mini-2025-04-16)
• Search: Serper.dev
• Scraping: Playwright, BeautifulSoup
• Project Management: uv