🛡️ TruthPulse - Multimodal, Multiplatform Fact-Checker

Inspiration

Misinformation affects everything from elections to public health. Fact-checking today is slow, fragmented, and often reactive. We wanted to create a tool that makes truth accessible instantly—wherever misinformation spreads—whether that's WhatsApp, social media, or random forwarded images.

What it does

TruthPulse is a fact-checking platform that works across a web app, WhatsApp bot, and browser extension. Users can input text, images, or URLs and receive a response with a verdict, explanation, and sources. It uses live web search and AI to check claims and also caches similar ones to improve performance and reduce cost.

How we built it

  • Frontend: React, deployed on Vercel
  • Backend: FastAPI, hosted on Railway
  • Database: Supabase with PostgreSQL and vector embeddings
  • AI Services: OpenAI for NLP, Perplexity Sonar for search
  • Scraping: crawl4ai for collecting and cleaning news article content
  • Messaging: Twilio for WhatsApp integration
  • Browser Extension: Chrome Extensions

We also added caching using OpenAI embeddings to avoid repeating similar fact-checks.

Challenges we ran into

  • Coordinating multiple APIs at multiple platforms and handling failures smoothly
  • Processing multimodal inputs in a unified way
  • Deploying and syncing services across platforms
  • Making responses clear, structured, and source-based

Accomplishments that we're proud of

  • Built and deployed a working cross-platform fact-checking tool
  • Integrated caching to reduce repeated queries
  • Designed an intelligent dashboard for tracking accuracy and trends
  • Enabled image and URL verification along with text input

What we learned

  • Multimodal pipelines require tight coordination across services
  • Real-time caching with embeddings is effective for reducing cost
  • Simple messaging interfaces (like WhatsApp) can deliver huge value
  • Good UX and fast feedback matter as much as accurate AI

What's next for TruthPulse

  • Add user feedback system to improve result quality
  • Onboard partners like media orgs or educators
  • Use crawl4ai more extensively to build a real-time news article dataset

Built With

Share this project:

Updates