worldmonitor: The Open-Source Palantir Running in Your Browser
How koala73/worldmonitor ditched heavy frameworks to render massive geopolitical data streams at 60 frames per second.
Real-time global intelligence dashboard — AI-powered news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface.
- World Monitor abandons React for Vanilla TypeScript to render 125,000 data points at 60 frames per second.
- A tiered hydration engine pipelines real-time and historical data through a single Redis request to prevent API bottlenecks.
- The architecture ensures data uptime by automatically rerouting blocked scrapers through secondary IP rotation relays.
- Local-first AI integration allows users to run intelligence classifiers privately via Ollama and Transformers.js.
The Density Problem
Building a global situational awareness tool requires rendering an obscene amount of data on a single screen. World Monitor tracks over 400 real-time intelligence feeds and plots 125,000 military bases simultaneously. To achieve this, the architecture relies on a dual map engine utilizing globe.gl for 3D global perspectives and deck.gl for high-density 2D data layers.
The developers quickly realized that a traditional modern frontend stack could not handle the rendering load. The overhead of React's virtual DOM reconciliation caused the browser to stutter when animating thousands of data points at 60 frames per second. The solution was a complete abandonment of heavy UI frameworks. World Monitor is built with Vanilla TypeScript and Vite, keeping the client as close to the metal as the browser allows.
The Hydration Engine
Getting 125,000 data points into the browser without triggering an API waterfall is a significant engineering challenge. If the client made a separate HTTP request for every news source, earthquake monitor, and financial ticker, the application would grind to a halt before the map even loaded.
World Monitor solves this with a highly optimized hydration engine located in api/bootstrap.js. When the application initializes, it makes a single, pipelined request to an Upstash Redis cache. The backend splits the required data into two distinct tiers. The FAST tier delivers real-time signals like breaking news and active alerts, while the SLOW tier handles historical and static data points. This ensures the user interface feels instantly responsive.
Evading the Blockades
Intelligence dashboards are entirely dependent on their data sources. World Monitor scrapes institutional news feeds and public APIs globally. However, major news networks frequently block IP ranges associated with cloud providers like Vercel to prevent automated scraping.
The system guarantees uptime through a dual-path ingestion network managed by api/rss-proxy.js and a secondary relay system. If a target server returns a 403 Forbidden error due to an IP block, the Edge function instantly reroutes the request through a long-lived Railway relay node. This node uses different IP rotation strategies to successfully retrieve the data. The architecture maintains a deterministic RELAY_ONLY_DOMAINS list, allowing the system to skip direct fetching for known hostile environments and save execution time.
Intelligence at the Edge (and the Desktop)
A major differentiator for World Monitor is its approach to artificial intelligence. Instead of relying solely on cloud-based LLMs like OpenAI, which introduces privacy concerns for sensitive intelligence gathering, the platform offers a local-first AI architecture.
The application runs AI classifiers locally in the browser using Transformers.js and supports full offline execution via Ollama. This setup correlates Convergence Alerts, flagging when unrelated signals like a natural disaster and a sudden currency dip occur in the same geographic cell. For users requiring maximum performance and privacy, World Monitor ships as a native desktop application using Tauri. The desktop version utilizes a Node.js sidecar to handle heavy local processing that the browser sandbox restricts.
Local AI — run everything with Ollama, no API keys required
| Feature | World Monitor | Traditional Dashboards |
|---|---|---|
| Rendering Engine | Vanilla TS & WebGL (60fps) | React or DOM (High overhead) |
| AI Execution | Local Ollama & Transformers.js | Cloud API (Privacy risks) |
| Data Fetching | Pipelined Redis & Edge Fallbacks | REST Monolith |
| Deployment | Browser & Tauri Desktop | Web Only |
Sources: koala73/worldmonitor on GitHub, AIBit Blog, v2.5.21 Release Notes.