truth

Don't live in a false world, Find the truth!

🧠 Truth – Real-Time AI-Powered Fact Verifier

🔍 Analyze claims. Understand context. Reveal the truth.
A hackathon project built for the Perplexity AI Sonar Hackathon 2025.

🚀 What is Truth?

Truth is an AI-powered fact-checking tool that lets users verify controversial or suspicious claims in real time — whether from a conversation, news article, tweet, or meme.

It’s not just another "TRUE/FALSE" classifier.

Truth explains why?

Using state-of-the-art analysis (via the Perplexity Sonar API and custom reasoning layers), the app provides:

✅ Veracity assessment (TRUE/FALSE/UNCERTAIN)
🧠 Chain of Thought (step-by-step reasoning)
📌 Key evidence summaries
📰 Trusted source linking
💬 Tone and sentiment classification
📷 OCR pipeline to extract claims from screenshots or images

Feature	Description
🔎 Natural Language Claim Input	Ask questions like "Did X happen?" or "Is it true that...?"
📜 Chain-of-Thought Reasoning	Get step-by-step explanations on how the verdict was reached
📰 Key Evidence Summary	Extracted summaries with links to credible sources
🧠 Classification Engine	Detects topic (e.g., politics, health), sentiment, and tone
🖼️ OCR Image Upload	Upload screenshots, tweets, or memes to analyze embedded claims

🛠️ How It Works

Claim Input Users type a natural language claim or upload an image.

Preprocessing If image: OCR is run → claim text extracted. If text: directly analyzed.

Sonar API Integration Claim is passed to Perplexity Sonar API for fact-verification.

Post-Processing Extracted JSON is parsed for:

Verdict

Confidence level

Explanation

Source links

Reasoning Layer (Claude) Chain-of-thought and key reasoning points are derived using LLM prompts.

UI Presentation Verdict, reasoning, sources, and classification are displayed in an intuitive interface.

🔧 Tech Stack

Layer	Tools
Frontend	React.js, TailwindCSS
Backend	Node.js, Express.js
AI & APIs	Perplexity Sonar API, Google Vision API OCR, Gnews
Data Storage	sessionStorage (for in-browser session history)

📚 Example Use Cases

Verifying political claims during elections

Debunking social media misinformation

Checking health-related rumors or statistics

Detecting scams, hoaxes, or manipulated news screenshots

🚧 Future Plans

Current Limitation	Upcoming Fix
Not fully dynamic layout	Plan to visualize source relationships
No long-term history	Will add user sessions
No real-time social media monitoring	Could integrate with Twitter/Reddit APIs
No Chrome extension yet	Planned as a post-hackathon launch

🧩 Planned Feature: Dynamic Claim-Source Map (UI Blueprint)

We are actively designing an interactive UI feature that visually maps claims to their supporting or refuting sources.

What it will do:

📌 Dynamically plot each claim at the center

🕸️ Surround it with nodes representing sources (news articles, evidence, counterclaims)

🔗 Use graph links to show relationships (support/refute/neutral)

🧠 Color-coded for quick understanding: green (support), red (refute), gray (neutral)

👆 Fully interactive — users can click any node to preview source content and influence on verdict

Inspiration: Think of it as a "mind map for truth" — allowing users to trace how a claim is built, debated, or discredited.

⚠️ Due to time constraints, this visualization was postponed to post-hackathon development — but its system design is complete and modularly integrated.