Open Source
Explore the latest AI open-source projects from GitHub and HuggingFace.
Explore the latest AI open-source projects from GitHub and HuggingFace.
World Monitor is an AI-powered, real-time global intelligence dashboard that aggregates news, geopolitical data, and infrastructure tracking into a unified situational awareness platform. With over 34,000 GitHub stars, it has emerged as one of the most popular open-source intelligence tools available, offering a comprehensive free and self-hostable alternative to expensive commercial OSINT platforms. The project synthesizes information from 435+ RSS feeds across 15 categories, displaying geopolitical events, military movements, infrastructure threats, and market signals on interactive 2D and 3D maps. Released under the AGPL v3 license, it supports native desktop apps through Tauri for macOS, Windows, and Linux, as well as progressive web app deployment with offline map support. ## Core Intelligence and Analysis World Monitor integrates LLM-synthesized summaries through local Ollama support with a 4-tier fallback chain covering Ollama, Groq, OpenRouter, and Transformers.js. The system calculates a Country Instability Index (CII) scoring system that rates geopolitical stability across all nations, providing at-a-glance assessments of global hotspots. The Headline Memory RAG system runs entirely client-side using ONNX embeddings, meaning no data leaves the user's machine. This approach addresses both privacy concerns and the need for historical context in intelligence analysis. ## Visualization Engine The dashboard features a dual map engine combining a photorealistic 3D globe (globe.gl with Three.js) and a WebGL flat map (deck.gl). Users can toggle between 45 data layers covering conflicts, military bases, infrastructure, vessel tracking (AIS), and flight tracking (ADS-B). Eight regional presets with 1-hour to 7-day time filtering allow focused monitoring of specific areas of interest. | Feature | Specification | |---------|---------------| | RSS Feeds | 435+ across 15 categories | | Data Layers | 45 toggleable layers | | Map Engines | globe.gl (3D) + deck.gl (2D) | | Live Streams | 30+ (Bloomberg, Sky News, Al Jazeera) | | Languages | 21 with native RSS feeds | | Financial Data | 92 stock exchanges, 19 financial centers | ## Multi-Variant Architecture One of World Monitor's most distinctive features is its single codebase producing five specialized dashboard variants through build-time tree-shaking: - **World Monitor**: Geopolitics and global events - **Tech Monitor**: Startup and AI ecosystem tracking - **Finance Monitor**: Markets and trading signals - **Commodity Monitor**: Mining and energy sector data - **Happy Monitor**: Positive news aggregation This architecture allows teams to deploy purpose-built dashboards without maintaining separate codebases. ## Technical Architecture The project uses vanilla TypeScript without any frontend framework, relying on custom DOM manipulation for maximum performance. The backend runs on Vercel Edge Functions (60+ endpoints) with a proto-first API architecture providing 22 typed services with auto-generated clients. Redis handles caching, while IndexedDB manages browser-side persistence. The API is publicly accessible at api.worldmonitor.app, making it possible to integrate World Monitor data into external applications and dashboards. ## Security and Testing Security features include CORS origin allowlisting, server-side API key isolation, input sanitization for XSS and SSRF protection, OS keychain storage for desktop credentials, and IP rate limiting on sensitive endpoints. The test suite includes 30 test files with 554 individual test cases covering server handlers, caching, data integrity, and map overlays, plus Playwright E2E tests. ## Limitations World Monitor depends on external RSS feeds and APIs that can change or become unavailable. The Vercel Edge Functions architecture means self-hosting requires adapting the deployment setup. The AGPL v3 license requires derivative works to also be open source, which may not suit all commercial use cases. The 45 data layers can be overwhelming for users who need simple news aggregation rather than full-spectrum intelligence monitoring.