Veritas: Autonomous Multi-Agent Truth Engine

Veritas: Autonomous Multi-Agent Truth Engine "Nox est norma, Lux per rationem reconstruitur." (Darkness is the norm, Light is reconstructed through reason.) 🚨 THE INTERNET IS BROKEN. FIX YOUR FEED! You are drowning in noise. Every day, you are bombarded with clickbait, logical fallacies, emotional manipulation, and half-truths. Algorithms don't care about what is true; they care about what keeps you clicking. The result? A fragmented reality where facts are optional and outrage is the currency. Veritas is not just a browser extension. It is a Military-Grade Cognitive Defense System. It is a "Digital Brutalist" tool designed for one purpose: to reduce information entropy. It doesn't tell you what to think. It shows you what to question. By overlaying a "Truth Lens" on top of any website, Veritas reveals the hidden structure of arguments, exposes manipulation, and verifies facts in real-time using a swarm of specialized AI agents. 🧠 MEET THE AUTONOMOUS SWARM Veritas is not a simple chatbot wrapper. It is a coordinated Multi-Agent System (MAS). Four specialized AI agents work in parallel, communicating via a high-speed internal protocol to analyze what you read in milliseconds. 1. 🛡️ VELOX (The Sentry) Model: Gemini 2.5 Flash-Lite (Optimized for Speed) Role: Rapid Defense & Noise Filtering. The Problem: Most content is fluff, ads, or emotional triggers designed to bypass your logic. The Solution: Velox acts as a "World-Class Logic Professor". It scans the page instantly, dimming low-value content (ads, sidebars) and highlighting Logical Fallacies in high-contrast blue. Capabilities: Fallacy Detection: Identifies 12+ types of rhetorical tricks including Ad Hominem, Strawman, Gaslighting, Slippery Slope, and False Dichotomy. Emotional Severity Matrix: Grades manipulation attempts from "Low" (Sarcasm) to "High" (Dehumanization/Fear-mongering). Noise Suppression: Visually fades out ads and boilerplate so you can focus on the signal. 2. ⚖️ RATIO (The Analyst) Model: Gemini 2.5 Flash (Optimized for Precision) Role: Entropy Reduction & Fact Extraction. The Problem: Facts are often buried in paragraphs of opinion. The Solution: Ratio follows a strict "Extract, Do Not Judge" protocol. It acts as a surgical scalpel, parsing text to isolate atomic Factual Claims and Entities. Capabilities: Atomic Extraction: Separates objective claims ("GDP grew by 3%") from subjective spin ("The economy is booming"). Multilingual Processing: Preserves original language (Chinese/English) to ensure accurate cross-border verification. Entity Linking: Identifies and tags People, Organizations, and Locations for the Knowledge Graph. 3. 🔍 VERITAS (The Investigator) Model: Gemini 2.5 Flash + Google Search Grounding Role: Real-Time Web Verification. The Problem: Checking facts takes time. You have to open new tabs, search, and compare sources. The Solution: Veritas is the "Arbiter of Truth". It autonomously performs deep web searches to cross-reference claims against trusted sources (Gov, Edu, Major Media). Capabilities: Anti-Recitation Protocol: Synthesizes answers instead of copying them, preventing hallucinations. Source Credibility Scoring: Ranks evidence based on domain authority. The Verdict: ✅ VERIFIED (Green): Confirmed by multiple independent sources. ❌ FALSE (Red): Explicitly debunked or contradicted by evidence. ⚠️ DISPUTED (Yellow): Sources disagree, or context is missing. 💬 COMMANDER: THE HIVE MIND (Autonomous Orchestrator) Commander is not a chatbot. It is a Cybernetic Orchestrator that wields the other agents as tools. It operates on a Reasoning + Acting (ReAct) loop, allowing it to perform complex, multi-step research missions autonomously. ⌨️ How to Command Trigger: Press Alt + C (or Option + C on Mac) to open the Command Terminal. Context: Commander automatically ingests the page context. You can refer to "this article", "that claim", or "the second paragraph". Natural Language: No code required. Just tell it what you need (e.g., "Verify the statistics in this section"). 🤖 The Cognitive Protocol Commander intelligently switches modes based on your intent: Direct Execution Mode: "Highlight all mentions of 'Elon Musk' in red." -> Commander immediately calls DOM tools to paint the page. Analysis Mode: "Is this article biased?" -> Commander reads the text, triggers Velox for fallacy detection, and summarizes the findings. Creative Mode: "Rewrite this paragraph to be neutral." -> Commander uses extracted facts to reconstruct the text without the spin. ⚡ Autonomous Research Chains (The Agent Loop) You can give Commander complex instructions that require planning and execution: Command: "Find all claims about 'nuclear energy' on this page, verify them, and then summarize the three biggest misconceptions." The Chain: Plan: Commander breaks this down: Read Page -> Extract Claims -> Verify Claims -> Synthesize Report. Action: It deploys Ratio to extract claims and Veritas to verify them in parallel. Output: It presents a final report in the chat window, highlighting the misconceptions. 🎮 Agent Retasking & Correction Even the Swarm makes mistakes. You can correct them in real-time: Command: "You missed the claim in the second paragraph. Look again." Mechanism: Commander receives this as a "System Error", re-plans, and re-executes the extraction with a narrower focus. ✨ KEY FEATURES 👁️ The Truth Scan (Alt + V) Press the hotkey, and watch the Neon Ritual begin. A breathing cyan and gold halo surrounds your screen. Scanlines flicker. The system comes alive. In seconds, the page is transformed: ✅Fallacies are boxed. ✅Facts are underlined. ✅Lies are struck through. It’s like putting on X-Ray glasses for information. ⚛️ Enhanced Holographic Cards Hover over any highlighted element to reveal a floating, interactive dashboard. This isn't a simple tooltip; it's a deep-dive analysis station with 6 tabs: Risk Gauge: A real-time, animated meter showing the "Misinformation Risk Score" (0-100). Fallacies: A br…