Checkmate 🔍

🌐 Visit the Website

AI-Powered Misinformation Detection & Fact-Checking Platform

Combating digital misinformation in Malaysia through advanced AI, NLP, and crowd-sourced verification

---

🎞️ Presentation Slides

You can view our Great Malaysia AI Hackathon 2025 pitch deck here:

👉 [Checkmate_Presentation.canva](novelty wise mainly on slides https://www.canva.com/design/DAGzfS1Aohk/l6Wi_7JUqbn8l3GkRLGpwg/edit)

🏆 Great Malaysia AI Hackathon 2025 Challenge & Our Approach

The Problem We're Solving

In today's hyperconnected world, information travels faster than ever, but so does misinformation. From misleading headlines and manipulated images to fake news and deepfakes, Malaysia, like many countries, is grappling with the real-world consequences of digital falsehoods. These include public confusion, social unrest, and a decline in trust toward media, institutions, and even one another.

As digital citizens, we all play a role in upholding the truth. But the scale and speed of today's information environment demand technological solutions that can proactively detect, counter, and educate against misinformation, while preserving freedom of speech and access to information.

Challenge Statement:

How can we harness technology to combat misinformation and promote digital integrity, truth, and accountability in Malaysia's online spaces?

Our Solution & Approach

Checkmate addresses this challenge through a comprehensive AI-powered platform that:

1. Proactively Detects misinformation using advanced NLP and content analysis
2. Verifies Claims through automated fact-checking with credible sources
3. Evaluates Creator Credibility using data-driven scoring algorithms
4. Empowers Users with accessible tools for content verification
5. Builds Community Trust through crowd-sourced verification mechanisms

🚀 Local Development Setup

Clone the repo and install dependencies:

git clone https://github.com/MohtashamMurshid/checkmate.git
cd checkmate
npm install

Start the development server:

npm run dev

Open http://localhost:3000 in your browser.

🗄️ Database: AWS RDS PostgreSQL + Drizzle ORM

This project uses AWS RDS (or Aurora) PostgreSQL with Drizzle ORM and migrations. BetterAuth stores users/sessions in the same Postgres DB via the Drizzle adapter.

1) Environment

Copy .env.local from env.example and set:

DATABASE_URL="postgres://APP_USER:APP_PASSWORD@your-instance.abcdefg123.us-east-1.rds.amazonaws.com:5432/app_db?sslmode=require"
BETTER_AUTH_SECRET="a-long-random-string"

Auth now uses a simple email/password with a signed cookie. Configure:

AUTH_SECRET="...long-random-string..."
ADMIN_EMAIL="admin@example.com"
ADMIN_PASSWORD="password"

2) Install deps

npm install
npm install drizzle-orm postgres
npm install -D drizzle-kit

3) Drizzle migrations

npx drizzle-kit generate
npx drizzle-kit migrate

4) Run

npm run dev

Notes

• DB client is defined in lib/db/index.ts with SSL required for AWS RDS.
• Schema lives in lib/db/schema.ts.
• Data-access repo functions in lib/db/repo.ts.
• BetterAuth configured in lib/better-auth.ts with Drizzle adapter.

📚 Developer Documentation

For new developers joining the project, comprehensive documentation is available:

VERCEL_URL=http://localhost:3000
OPENAI_API_KEY=sk-proj-FAKEKEYFORDEMO0987654321
CONVEX_DEPLOYMENT=dev:tidy-grouse-158 # team: mohtasham-murshid, project: checkmate-29f39
NEXT_PUBLIC_CONVEX_URL=https://tidy-grouse-158.convex.cloud
NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY=pk_test_FAKEPUBLISHABLEKEY
CLERK_SECRET_KEY=sk_test_FAKESECRETKEY
NEXT_PUBLIC_CLERK_FRONTEND_API_URL=https://magical-marmot-34.clerk.accounts.dev
NEXT_PUBLIC_CLERK_SIGN_IN_URL=/sign-in
NEXT_PUBLIC_CLERK_SIGN_IN_FALLBACK_REDIRECT_URL=/news
NEXT_PUBLIC_CLERK_SIGN_UP_FALLBACK_REDIRECT_URL=/news
FIRECRAWL_API_KEY=fc-FAKEKEYFORDEMO0987654321

🛠️ Technologies Used

Frontend Technologies

• Next.js 14 - React framework with App Router
• TypeScript - Type-safe development
• Tailwind CSS - Utility-first styling
• Shadcn/UI - Modern component library
• React Hooks - State management and side effects
• Flutter - Cross-platform mobile wrapper for Checkmate web app
  _Codebase: see @/checkmate_wrapper_checkmate_/_
• Browser Extension - Chrome/Edge extension for real-time fact-checking overlay
  _Codebase: see @/checkmate_browser_extension/

🚀 Key Features & Capabilities

🔍 Multi-Platform Content Analysis

• TikTok Videos: Extract metadata, transcribe audio, analyze claims
• Twitter/X Posts: Process tweets, images, and embedded media
• Web Articles: Scrape and analyze blog posts, news articles
• Direct Media: Upload videos, images, or audio files

🤖 AI-Powered Detection

• Sentiment Analysis: Detect manipulative language and bias
• Claim Extraction: Identify factual statements requiring verification
• Pattern Recognition: Spot common misinformation tactics
• Language Processing: Support for multiple languages (Malaysian context)

✅ Automated Fact-Checking

• Real-time Verification: Cross-reference claims with credible sources
• Source Credibility: Evaluate reliability of information sources
• Confidence Scoring: Provide certainty levels for fact-check results
• Evidence Compilation: Generate comprehensive verification reports

👤 Creator Credibility System

• Historical Analysis: Track creator's accuracy over time
• Community Feedback: Incorporate user ratings and comments
• Cross-Platform Tracking: Unified credibility across social platforms
• Transparency: Clear methodology for credibility calculations

🌐 Accessibility & Localization

• Multilingual Support: English, Bahasa Malaysia, and regional languages
• Mobile-First Design: Responsive across all devices
• Dark/Light Mode: User preference customization
• Screen Reader Compatible: WCAG accessibility standards

Mobile App Wrapper

• Mobile App Wrapper: Access Checkmate as a native-like app on iOS/Android via Flutter

Browser Extension

• Browser Extension: Instantly fact-check content while browsing TikTok, Twitter, or news sites

📱 Usage Instructions & Demo

Demo Screenshots

Note: Screenshots are from a live demo of the platform.

1. Landing Page & Hero Section

• Clean, modern landing page with a clear value proposition.
• Explains the platform's mission to combat misinformation.
• Clear call-to-action buttons to get started.

2. Browser Extension

3. Fact-Check Results Dashboard

• Comprehensive results display with an overall credibility score.
• Detailed fact-check breakdown with sources and explanations.
• Links to creator credibility profiles.

4. Detailed Analysis View

• In-depth view with full transcription, sentiment analysis, and identified claims.
• Allows users to scrutinize the evidence and analysis process.

5. Creator Credibility Profile

• Historical credibility trends for content creators.
• Analysis of past content and community feedback.

6. Saved Analyses & History

🏗️ Technical Architecture

System Overview

Checkmate follows a modern full-stack architecture with the following components:

Database Schema (Convex)

Our data model consists of four main entities:

users

• Synchronized from Clerk authentication
• Stores user profile information and preferences

contentCreators

• Tracks credibility metrics for content creators across platforms
• Maintains credibility ratings (0-10 scale) based on analysis history
• Supports multi-platform creator identification

tiktokAnalyses

• Stores comprehensive analysis results for each processed content
• Links to users and content creators
• Contains transcription, metadata, news detection, and fact-check results

creatorComments

• Enables community feedback on creator credibility
• Supports crowd-sourced verification efforts

AI/ML Tools Architecture (@/tools)

The @/tools directory contains the core AI-powered functionality, organized into modular components:

helpers.ts - Core Utilities

// Video transcription using OpenAI Whisper
export async function transcribeVideoDirectly(videoUrl: string);

// Web content scraping using Firecrawl
export async function scrapeWebContent(url: string);

tiktok-analysis.ts - Platform-Specific Analysis

• analyzeTikTokVideo: Extracts metadata, download links, and video content from TikTok URLs
• transcribeTikTokVideo: Converts TikTok audio to text using OpenAI Whisper
• compareTikTokVideos: Analyzes multiple videos for trends and patterns

Key technologies:

• @tobyg74/tiktok-api-dl for TikTok video extraction
• OpenAI Whisper via ai SDK for speech-to-text
• Real-time video processing and analysis

content-analysis.ts - Content Intelligence

• analyzeContentSentiment: NLP-powered sentiment analysis and theme extraction
• extractHashtagsAndMentions: Social media element extraction using regex patterns
• generateContentInsights: AI-driven recommendations and quality scoring
• generateVideoSummary: Automated content summarization

Advanced features:

• Multi-dimensional sentiment analysis
• Viral potential prediction algorithms
• Accessibility compliance checking
• Engagement metric calculations

fact-checking.ts - Misinformation Detection

• detectNewsContent: Identifies content requiring fact-checking using NLP
• researchAndFactCheck: Cross-references claims with credible sources
• analyzeCreatorCredibility: Calculates creator trustworthiness scores

Sophisticated algorithms:

• Domain credibility evaluation using LLM reasoning
• Multi-source claim verification
• Confidence scoring with uncertainty quantification
• Automated source reliability assessment

index.ts - Tool Orchestration

Exports organized tool collections:

export const allTiktokAnalysisTools = [...];
export const allFactCheckingTools = [...];
export const allTools = [...]; // Combined toolkit

Data Flow & Processing Pipeline

1. Content Ingestion

   User Input (URL) → Platform Detection → Content Extraction
   

2. Multi-Modal Analysis

   Video/Audio → Whisper Transcription → Text Analysis
   Text Content → NLP Processing → Claim Extraction
   

3. Fact-Checking Pipeline

   Claims → Web Research → Source Verification → Credibility Scoring
   

4. Result Synthesis

   Individual Results → Comprehensive Analysis → User Dashboard
   

API Architecture

/api/transcribe - Main Analysis Endpoint

Handles multi-platform content analysis:

// Request types supported
interface RequestBody {
  tiktokUrl?: string; // TikTok video URLs
  twitterUrl?: string; // Twitter/X post URLs
  webUrl?: string; // General web content
  videoUrl?: string; // Direct video URLs
}

// Response structure
interface AnalysisResult {
  transcription: TranscriptionData;
  metadata: ContentMetadata;
  newsDetection: NewsDetectionResult;
  factCheck: FactCheckData;
  creatorCredibilityRating: number;
}

Processing Flow:

1. URL validation and platform detection
2. Content extraction (TikTok API, Twitter Scraper, or Firecrawl)
3. Transcription (if video content exists)
4. News content detection using AI
5. Fact-checking pipeline execution
6. Creator credibility calculation
7. Result compilation and return

Frontend Architecture

Custom Hooks (lib/hooks/)

• use-tiktok-analysis.ts: Main analysis orchestration hook
• use-saved-analyses.ts: Database interaction for saved analyses
• use-credible-sources.ts: Credible source management
• use-all-analyses.ts: Comprehensive analysis data fetching

Component Structure

components/
├── ui/                    # Shadcn/UI base components
├── analysis-renderer.tsx  # Display analysis results
├── creator-credibility-display.tsx  # Credibility scoring UI
├── language-provider.tsx  # I18n support
└── theme-provider.tsx     # Dark/light mode

Security & Performance

Authentication & Authorization

• Clerk Integration: Secure user authentication with social logins
• Middleware Protection: Route-level authentication enforcement
• API Security: Request validation and rate limiting

Performance Optimizations

• Streaming Responses: Real-time analysis result delivery
• Caching Strategy: Convex-powered efficient data caching
• Lazy Loading: Component-based code splitting
• Image Optimization: Next.js automatic image optimization

Error Handling

• Graceful Degradation: Fallback mechanisms for API failures
• User Feedback: Clear error messages and retry mechanisms
• Logging: Comprehensive error tracking and monitoring

Scalability Considerations

Horizontal Scaling

• Serverless Architecture: Auto-scaling API routes
• Database Sharding: Convex handles automatic scaling
• CDN Integration: Global content delivery

Monitoring & Analytics

• Performance Metrics: Real-time application monitoring
• Usage Analytics: User behavior and feature adoption tracking
• Error Tracking: Automated error detection and alerting

🏆 Impact & Future Roadmap

Addressing the Malaysian Context

• Language Support: Prioritizing Bahasa Malaysia and regional dialects
• Cultural Sensitivity: Understanding local misinformation patterns
• Government Collaboration: Potential integration with official fact-checking bodies
• Educational Outreach: Community programs for digital literacy

Planned Enhancements

• Browser Extension: Real-time fact-checking while browsing (prototype available)
• Mobile App: Native iOS/Android applications via Flutter wrapper (prototype available)
• API for Partners: Integration capabilities for news organizations
• Advanced AI Models: Custom-trained models for Malaysian content
• Blockchain Verification: Immutable fact-check records

📄 License

MIT License - Open source for educational and research purposes

---

Built with ❤️ for the Great Malaysia AI Hackathon 2025 - Fighting misinformation through technology