Veritas fights misinformation by fact-checking any text against the live web. Paste an article, tweet, or claim, and Veritas uses the Google Gemini API to: (1) decompose the text into atomic, checkable claims using Gemini's structured output; (2) verify each claim with Gemini's native Google Search grounding tool, so verdicts are backed by real sources it retrieves at query time — not the model's memory; and (3) roll everything into a per-claim verdict (Supported, Disputed, Misleading, Unsupported, or Opinion) with a confidence score, real citations, and an overall 0–100 credibility score.
Unlike a typical "ask an LLM if this is true" wrapper, Veritas showcases Gemini's distinctive capabilities — grounding + typed JSON output — in a real pipeline. It degrades gracefully to model-knowledge mode when grounding isn't available, so it always runs. Backend: Python + FastAPI. Frontend: a single-file web app served by the backend, plus an optional React Native client. It targets a genuine problem: helping people verify what they read before they share it.
Log in or sign up for Devpost to join the conversation.