News Intelligence System

Score and analyze news articles using 5 different AI models - get the smartest insights from multiple perspectives.

---

What Does This Do?

This tool helps you evaluate news articles using multiple AI "judges":

1. Fetches articles from news APIs
2. Sends to 5 AI models (ChatGPT, Claude, Gemini, Grok, Perplexity)
3. Each AI scores the article's relevance/quality
4. Compares opinions through peer review
5. Generates consensus on what matters most

Why? Different AI models have different perspectives. By using multiple models, you get more balanced, reliable analysis than trusting any single AI.

---

Quick Start

# Clone the repo
git clone https://github.com/jjshay/news-intelligence.git
cd news-intelligence

# Install dependencies
pip install -r requirements.txt

# Run the interactive demo (no API keys needed!)
python demo.py

# Or run the visual showcase
python showcase.py

# Try with sample data
python demo.py --input examples/sample_articles.json

Sample Output

See sample_output/analysis_report.json for example results.

---

Architecture

flowchart TB
    subgraph Input
        A[News Article] --> B[Article Parser]
    end

    subgraph AI["Multi-AI Scoring Engine"]
        B --> C1[ChatGPT]
        B --> C2[Claude]
        B --> C3[Gemini]
        B --> C4[Grok]
        B --> C5[Perplexity]
    end

    subgraph Peer["Peer Review"]
        C1 & C2 & C3 & C4 & C5 --> D[Score Comparison]
        D --> E[Dispute Resolution]
        E --> F[Perplexity Fact-Check]
    end

    subgraph Output
        F --> G[Consensus Calculator]
        G --> H[6th AI Consolidator]
        H --> I[Final Report]
    end

    style A fill:#e1f5fe
    style I fill:#c8e6c9
    style C1 fill:#fff3e0
    style C2 fill:#fce4ec
    style C3 fill:#e8f5e9
    style C4 fill:#f3e5f5
    style C5 fill:#e0f7fa

Loading

How It Works

News Article
     │
     ▼
┌─────────────────────────────────────────────────────────────┐
│                    5 AI Evaluators                          │
├─────────┬─────────┬─────────┬─────────┬─────────────────────┤
│ ChatGPT │ Claude  │ Gemini  │  Grok   │    Perplexity       │
│  8/10   │  7/10   │  9/10   │  8/10   │  9/10 (Final Say)   │
└─────────┴─────────┴─────────┴─────────┴─────────────────────┘
     │
     ▼
┌─────────────────────────────────────────────────────────────┐
│  6th AI Consolidator                                        │
│  "Final score: 8.2/10                                       │
│   - Strong factual content (ChatGPT, Claude agree)          │
│   - Timely topic with high relevance (Gemini, Perplexity)   │
│   - Minor bias detected (Grok dissent)                      │
│   - Recommended for sharing"                                │
└─────────────────────────────────────────────────────────────┘

---

The Scoring Process

Step 1: Health Check

System verifies all AI APIs are working before proceeding.

Step 2: Individual Scoring

Each AI evaluates the article on:

• Factual accuracy
• Source credibility
• Relevance to your interests
• Potential bias
• Writing quality

Step 3: Peer Pairing

AIs are randomly paired to compare scores. Perplexity (with web access) always participates as a fact-checker.

Step 4: Final Verdict

Perplexity has the "final say" on disputes, using its real-time web search to verify claims.

Step 5: Consolidation

A 6th AI call summarizes all opinions into clear bullet points.

---

AI Models Used

Model	Provider	Strengths
ChatGPT	OpenAI	Strong reasoning, broad knowledge
Claude	Anthropic	Nuanced analysis, safety-focused
Gemini	Google	Multi-modal, up-to-date
Grok	xAI	Contrarian views, real-time info
Perplexity	Perplexity AI	Web search, fact verification

---

Real Usage Setup

1. Get API Keys

You'll need API keys from:

• NewsAPI - News data
• OpenAI - ChatGPT
• Anthropic - Claude
• Google AI - Gemini
• xAI - Grok
• Perplexity - Perplexity AI

2. Configure Environment

# Copy example file
cp .env.example .env

# Add your keys
nano .env

3. Run

python fetch_news.py

---

Cost Breakdown

Service	Approximate Cost
NewsAPI	Free tier: 100 requests/day
ChatGPT (GPT-4)	~$0.01-0.03 per article
Claude	~$0.01-0.02 per article
Gemini	Free tier available
Grok	~$0.01 per article
Perplexity	~$0.01 per article

Total per article: About $0.05-0.10

---

Files

File	Purpose
fetch_news.py	Main scoring pipeline
news_responder_gui.py	GUI version with visual interface
news_sheet_comment_responder.py	Google Sheets integration
process_news_in.py	News processing utilities
demo.py	Demo without API keys

---

Output Example

ARTICLE: "AI Breakthrough in Medical Diagnosis"
Source: MIT Technology Review

INDIVIDUAL SCORES:
  ChatGPT:    8/10 - "Strong technical content, well-sourced"
  Claude:     7/10 - "Good but overstates near-term impact"
  Gemini:     9/10 - "Highly relevant, confirms with recent research"
  Grok:       8/10 - "Interesting but missing regulatory context"
  Perplexity: 9/10 - "Verified claims against 3 recent papers"

PEER REVIEWS:
  ChatGPT + Perplexity: Agreement (8.5 avg)
  Claude + Grok: Minor disagreement on timeline

FINAL CONSENSUS: 8.2/10

KEY INSIGHTS:
  - Factual claims verified by multiple sources
  - Technical accuracy high but accessibility could improve
  - Relevant to healthcare AI interests
  - Recommended for professional network share

---

Features

• Multi-Model Consensus: Don't trust one AI - get 5 opinions
• Fact Verification: Perplexity cross-checks claims
• Peer Review: AIs compare and discuss scores
• Google Sheets Integration: Store results automatically
• GUI Available: Visual interface with progress tracking

---

License

MIT - Use freely for your news analysis needs!