VerAi is an intelligent, multi-layered news verification system designed to combat misinformation. It cross-references claims against real-time web data and established fact-checking databases using a high-performance, parallelized AI pipeline.
In an era of deepfakes and AI-generated disinformation, manual fact-checking is too slow and prone to human bias. VerAi cuts through the noise. It transforms a simple news snippet into a comprehensive investigation, providing a verdict, confidence score, and verifiable source citations in seconds. It’s the "truth layer" for modern digital media.
VerAi employs a three-stage asynchronous pipeline to ensure accuracy and speed:
- Web Intel (Stage 1): Uses Groq’s
compound-betamodel to perform targeted web searches and extract relevant news sources. - Deep Reasoning (Stage 2): Passes the search evidence and the claim to Llama 3.3 70B, which performs a comparative analysis to determine credibility.
- Fact-Database Sync (Stage 3): Concurrently queries the Google Fact Check API to see if the claim has been debunked by major international fact-checking organizations.
Built for scale and reliability:
- Backend: Flask API with
ThreadPoolExecutorfor high-performance concurrent analysis. - Database/Auth: Integrated with Supabase for secure, stateless user management.
- Intelligence: Orchestrated through the Groq Cloud API for ultra-fast, low-latency AI inference.
- Flexibility: Supports multi-language analysis (UZ, EN, RU, JA, ZH) with contextual awareness.
1. Prerequisites
- Python 3.10+
- API Keys: Groq, Supabase, Google Fact Check.
2. Installation
pip install flask python-dotenv supabase groq requests