📰 The Anti-Echo Chamber (Newsroom Agent)

The Anti-Echo Chamber is an advanced AI-powered Multi-Agent Newsroom designed to combat media bias and ideological echo chambers.
It simulates a virtual newsroom where autonomous AI agents investigate the same topic from opposing political perspectives, challenge each other's conclusions, and ultimately synthesize a neutral, fact-based report.

This project is built as a portfolio-grade system to demonstrate advanced skills in: - Multi-agent orchestration - Cognitive architectures - Real-time research pipelines - Interactive AI systems

---

🚀 Key Features

🧠 Autonomous Multi-Agent Workflow (CrewAI)

• 🔵 Blue Pundit
  Researches topics using a Progressive / Social-Justice lens.
• 🔴 Red Pundit
  Researches topics using a Conservative / Economic-Freedom lens.
• 📰 The Editor
  A neutral agent that evaluates both sides and compiles the final, balanced dossier.

---

⚔️ Rebuttal Logic (Cognitive Architecture)

Agents don't just work in parallel.

They: 1. Share context 2. Read each other's drafts 3. Generate explicit rebuttals and counter-arguments 4. Refine their positions before synthesis

This simulates real editorial debate, not isolated LLM calls.

---

🎚️ Dynamic Personality Engine

A Polarization Slider (0--100%) in the UI dynamically alters system prompts at runtime:

• 0% → Polite academics
• 50% → Opinionated pundits
• 100% → Radical partisans

This allows real-time experimentation with prompt engineering and ideological intensity.

---

🎧 Multi-Modal Output

• Audio Briefings 🎙️
  Generates podcast-style summaries using gTTS
• PDF Reports 📄
  Downloadable dossiers via FPDF
• Live Web Research 🌐
  Uses Tavily API to fetch real-time sources (minimizing hallucinations)

---

💬 Interactive Interrogation Chat

Users can directly question individual agents using Session State:

"Hey Red Agent, why did you ignore the climate data?"

This enables agent-specific accountability and explainability.

---

📊 Sentiment & Bias Visualization

• TextBlob analyzes sentiment polarity
• Altair visualizes the ideological gap between agents
• Makes bias quantifiable and observable

---

⚡ Performance-Oriented Design

• Asynchronous tasks for parallel web research
• Reduced latency despite multiple agents and live data fetching

---

🧩 How It Works

1️⃣ Research Phase

Both the Blue Pundit and Red Pundit independently: - Query Tavily for live sources - Analyze data through their ideological lenses - Produce structured draft reports

---

2️⃣ Rebuttal Phase

Agents: - Read each other's drafts - Identify weak assumptions, bias, or missing data - Write targeted rebuttals and counterpoints

---

3️⃣ Synthesis Phase

The Editor Agent: - Evaluates both perspectives and rebuttals - Cross-checks factual overlap - Produces a neutral, evidence-backed final report

---

🛠️ Tech Stack

Category Technology

---

Language Python 3.12+ Agent Orchestration CrewAI LLM Llama 3.3 (Groq) Web Framework Streamlit Search Tavily API Audio gTTS PDF FPDF Visualization Altair NLP TextBlob Package Manager uv

---

⚙️ Installation & Setup

1️⃣ Clone the Repository

git clone https://github.com/kartik0905/newsroom_agent.git
cd newsroom_agent

---

2️⃣ Set Up Environment Variables

Create a .env file in the root directory:

GROQ_API_KEY=your_groq_api_key
TAVILY_API_KEY=your_tavily_api_key

---

3️⃣ Install Dependencies

Using uv (recommended):

uv sync

Or using pip:

pip install -r requirements.txt

---

4️⃣ Run the Application

uv run streamlit run app.py

The app will be available at:

http://localhost:8501

---

🎯 Why This Project Matters

This system goes beyond simple AI demos: - Demonstrates real-world multi-agent coordination - Showcases debate, critique, and synthesis - Highlights skills in AI safety, bias mitigation, and explainability - Designed to impress hiring managers and technical reviewers

---

👤 Author

Kartik (kartik0905)
AI & Full-Stack Developer

🔗 GitHub: https://github.com/kartik0905

---

📜 License

MIT License