ProofPulse

• 

  Main Dashboard

Inspiration In an era of viral misinformation and deepfakes, we saw the urgent need for real-time fact-checking tools that go beyond simple text analysis. Traditional fact-checkers are slow and manual, while existing automated tools are limited to basic claims. ProofPulse was born from the vision of empowering everyday users to instantly verify factual claims from any source—whether it's a video, PDF document, web article, or raw text—with AI-powered precision and transparency. We wanted to build something that could analyze a YouTube video mid-conversation, extract claims from a research paper during study sessions, or fact-check a news article before sharing it. The goal: make truth accessible in seconds, not hours.

What it does ProofPulse is a real-time AI claim verification platform that: Extracts text from multiple input types: videos (with transcription), URLs (web scraping), PDFs, and plain text Identifies factual claims using advanced NLP to distinguish verifiable assertions from opinions or questions Scores each claim using a multi-factor rubric: Evidence strength from web sources Agreement between multiple sources Contextual accuracy Model confidence levels Provides detailed explanations (600-800 characters) with source citations and reasoning Delivers verdicts (SUPPORTED, CONTRADICTED, or UNCLEAR) with transparency into the scoring methodology All processing happens asynchronously with real-time status updates, and results are cached for instant access to previously analyzed content.

Built With

• asyncio-ai/apis:-twelvelabs
• axios-backend:-python
• backboard-sdk
• beautifulsoup4
• docker-compose
• fastapi
• git
• google-gemini-data-&-caching:-valkey
• httpx
• javascript
• pydantic
• pypdf2-devops:-docker
• react
• redis
• tailwind-css
• typescript
• uvicorn