AI Anti-Disinformation Guardian

Multi-Agent Misinformation Detection & Fact-Checking Dashboard
Built with FastAPI, Agentic AI (A2A Protocol & MCP), and React + Chakra UI.

---

🚀 Overview

AI Anti-Disinformation Guardian is a production-ready platform that leverages multi-agent orchestration to detect, analyze, and explain misinformation in real time. Users paste any social media post, article excerpt, or news headline into the dashboard to receive risk scores, fact-check verdicts, and transparent reasoning.

---

🧠 Key Features

• Agentic Orchestration (A2A & MCP): Collaborative workflows between specialized agents
• Real-Time Risk Scoring: Color-coded indicators for misinformation likelihood
• Explainable Fact-Checking: Detailed claim analysis and evidence correlation
• Responsive Dashboard: React 19 + Chakra UI for seamless desktop and mobile experiences
• Modular & Open Source: Easily extendable with custom agents and integrations

---

📊 How It Works

1. User Input: Paste content into the dashboard.
2. MCP Orchestration: Main Control Point distributes tasks among agents.
3. A2A Communication: Agents (Content Analyzer, Fact Checker, Risk Scorer) exchange subtasks and insights.
4. Result Presentation: Dashboard displays scores, badges, and reasoning chains.

---

📦 Project Structure

``` ai-anti-disinformation-guardian/ ├── a2a_protocol/ # MCP and agent code (Python) ├── main.py # FastAPI backend entrypoint ├── frontend/ # React + Chakra UI dashboard ├── requirements.txt # Python dependencies ├── README.md # Project documentation └── docker-compose.yml # Optional container orchestration ```

---

🛠️ Tech Stack

• Backend: FastAPI, Uvicorn, Pydantic, Docker
• Frontend: React 19, Vite, Chakra UI, TypeScript
• Orchestration: A2A Protocol, Main Control Point (MCP)
• Deployment: Docker Compose, Nginx, Redis, PostgreSQL

---

▶️ Getting Started

1. Clone the repository

   git clone https://github.com/Ayush1Deshmukh/AI-Anti-Disinformation-Guardian.git
   cd AI-Anti-Disinformation-Guardian

2. Run the backend

   pip install -r requirements.txt
   uvicorn main:app --reload

3. Run the frontend

   cd frontend
   npm install
   npm run dev

4. Open http://localhost:5173 to access the dashboard.

---

⚡ Demo

![Dashboard Screenshot 1](https://github.com/user-attachments/assets/29133eea-3d4d-491

![Dashboard Screenshot 2](https://github.com/user-attachments/assets/fe0c0f87-1b16-45d

🔬 Architecture

┌───────────────────────────┐
│   Orchestration Layer     │
│    (MCP & Task Router)    │
└───────────────────────────┘
            │
┌───────────────────────────┐
│    A2A Communication      │
│ ContentAnalyzer ↔ FactChecker ↔ RiskScorer │
└───────────────────────────┘
            │
┌───────────────────────────┐
│  Data & Context Layer     │
│ Databases, APIs, Streaming│
└───────────────────────────┘

---

🧩 Agent Components

• Content Analyzer: Detects sensational language, emotional triggers, and preliminary risk factors.
• Fact Checker: Extracts and verifies claims against external sources, computes confidence scores.
• Risk Scorer: Aggregates agent outputs to compute final risk level and explanations.

---

📈 Evaluation Metrics

Metric	Content Analyzer	Fact Checker	Multi-Agent	Baseline
Accuracy (%)	87.3	91.2	94.8	85.1
F1-Score (%)	87.4	91.1	93.6	85.0
MCC	0.751	0.823	0.872	0.702
Latency (ms)	145	298	187	112

---

🌱 Future Enhancements

• Multimodal Analysis: Integrate image/video fact-checking
• Live Streaming: WebSocket support for continuous monitoring
• Advanced NLP: Incorporate cutting-edge transformer models
• Federated Learning: Distributed agent training across data silos

---

⚖️ Ethics & Privacy

• Transparency: Full agent reasoning chains for user trust
• Bias Mitigation: Diverse datasets and regular fairness audits
• Privacy: Data minimization, anonymization, and secure storage
• Human Oversight: Appeals and manual review for contested cases

---

🌟 Contributors

• Ayush Deshmukh — Lead Developer & Researcher
  ayushdeshmukh@example.com

---

🏆 License

This project is licensed under the MIT License.

---

📚 References

• Multi-Agent Architectures in AI
• FastAPI Documentation
• Chakra UI Documentation
• Agent Communication Protocols and A2A Research

For full citation details, see the project repository.# 🕊️ AI Anti-Disinformation Guardian

Multi-Agent Misinformation Detection & Fact-Checking Dashboard
Built with FastAPI, Agentic AI (A2A Protocol & MCP), and React + Chakra UI.

---

🚀 Overview

AI Anti-Disinformation Guardian is a production-ready platform that leverages multi-agent orchestration to detect, analyze, and explain misinformation in real time. Users paste any social media post, article excerpt, or news headline into the dashboard to receive risk scores, fact-check verdicts, and transparent reasoning.

---

🧠 Key Features

• Agentic Orchestration (A2A & MCP): Collaborative workflows between specialized agents
• Real-Time Risk Scoring: Color-coded indicators for misinformation likelihood
• Explainable Fact-Checking: Detailed claim analysis and evidence correlation
• Responsive Dashboard: React 19 + Chakra UI for seamless desktop and mobile experiences
• Modular & Open Source: Easily extendable with custom agents and integrations

---

📊 How It Works

1. User Input: Paste content into the dashboard.
2. MCP Orchestration: Main Control Point distributes tasks among agents.
3. A2A Communication: Agents (Content Analyzer, Fact Checker, Risk Scorer) exchange subtasks and insights.
4. Result Presentation: Dashboard displays scores, badges, and reasoning chains.

---

📦 Project Structure

``` ai-anti-disinformation-guardian/ ├── a2a_protocol/ # MCP and agent code (Python) ├── main.py # FastAPI backend entrypoint ├── frontend/ # React + Chakra UI dashboard ├── requirements.txt # Python dependencies ├── README.md # Project documentation └── docker-compose.yml # Optional container orchestration ```

---

🛠️ Tech Stack

• Backend: FastAPI, Uvicorn, Pydantic, Docker
• Frontend: React 19, Vite, Chakra UI, TypeScript
• Orchestration: A2A Protocol, Main Control Point (MCP)
• Deployment: Docker Compose, Nginx, Redis, PostgreSQL

---

▶️ Getting Started

1. Clone the repository

   git clone https://github.com/Ayush1Deshmukh/AI-Anti-Disinformation-Guardian.git
   cd AI-Anti-Disinformation-Guardian

2. Run the backend

   pip install -r requirements.txt
   uvicorn main:app --reload

3. Run the frontend

   cd frontend
   npm install
   npm run dev

4. Open http://localhost:5173 to access the dashboard.

---

⚡ Demo]

# AI Anti-Disinformation Guardian

🔬 Architecture

┌───────────────────────────┐
│   Orchestration Layer     │
│    (MCP & Task Router)    │
└───────────────────────────┘
            │
┌───────────────────────────┐
│    A2A Communication      │
│ ContentAnalyzer ↔ FactChecker ↔ RiskScorer │
└───────────────────────────┘
            │
┌───────────────────────────┐
│  Data & Context Layer     │
│ Databases, APIs, Streaming│
└───────────────────────────┘

---

🧩 Agent Components

• Content Analyzer: Detects sensational language, emotional triggers, and preliminary risk factors.
• Fact Checker: Extracts and verifies claims against external sources, computes confidence scores.
• Risk Scorer: Aggregates agent outputs to compute final risk level and explanations.

---

📈 Evaluation Metrics

Metric	Content Analyzer	Fact Checker	Multi-Agent	Baseline
Accuracy (%)	87.3	91.2	94.8	85.1
F1-Score (%)	87.4	91.1	93.6	85.0
MCC	0.751	0.823	0.872	0.702
Latency (ms)	145	298	187	112

---

🌱 Future Enhancements

• Multimodal Analysis: Integrate image/video fact-checking
• Live Streaming: WebSocket support for continuous monitoring
• Advanced NLP: Incorporate cutting-edge transformer models
• Federated Learning: Distributed agent training across data silos

---

⚖️ Ethics & Privacy

• Transparency: Full agent reasoning chains for user trust
• Bias Mitigation: Diverse datasets and regular fairness audits
• Privacy: Data minimization, anonymization, and secure storage
• Human Oversight: Appeals and manual review for contested cases

---

🌟 Contributors

• Ayush Deshmukh — Lead Developer & Researcher
  ayushdeshmukh@example.com

---

🏆 License

This project is licensed under the MIT License.

---

📚 References

• Multi-Agent Architectures in AI
• FastAPI Documentation
• Chakra UI Documentation
• Agent Communication Protocols and A2A Research

For full citation details, see the project repository.