# Omarchy Yeet AI: Generate Realtime System Dashboards with Natural Language Transform system queries into live visualizations instantly through AI-powered dashboards - **Source:** [Hacker News: Show HN](https://news.ycombinator.com) - **Published:** 2026-09-30 - **Topic:** [Software & Engineering](https://brifing.si/c/engineering) - **Original article:** - **On Brifing:** https://brifing.si/s/47325622-bb60-4d48-991b-5bb9bb2d4f9b ## AI-powered system visualization Omarchy Yeet AI transforms natural language queries into live system dashboards through generative AI. Users ask questions like 'show CPU core usage' or 'display memory-intensive processes' to instantly generate interactive visualizations. The system supports twelve chart types including gauges, heat maps, and scatter plots. Each visualization updates in real-time, with new data samples sliding in from the right and changed elements animating smoothly. The tool maintains a persistent connection to system metrics through yeet's typed GraphQL system graph. ## Architecture and execution model The plugin runs in an isolated V8 environment on the host machine, communicating through stdin/stdout without opening network ports. When a user submits a query, the system sends the prompt along with the host's system graph schema to the AI model (defaulting to Claude Opus 5). The model responds with markdown containing chart directives in :::chart blocks. These blocks are parsed and executed in real-time as they stream in, creating live subscriptions to system metrics. The isolate automatically terminates a few seconds after the last visualization is closed. ## System graph integration All metrics come from yeet's system graph, a comprehensive GraphQL interface covering processes, CPU, memory, network, sockets, containers, and sensors. Unlike traditional monitoring tools that poll for data, this system establishes subscriptions to specific metrics at defined intervals. The model has access to the complete schema through introspection and can query specific fields using graph_schema and graph_query tools. Nullable fields are automatically guarded against in queries, and the system includes a repair mechanism that automatically corrects failed queries up to two times. ## Visualization framework The dashboard uses a responsive grid layout that automatically adjusts based on the number of active charts. Starting with one column, it expands to two columns after the second chart, three after the fifth, and four after the tenth. Charts maintain consistent theming, automatically adopting the host system's color scheme. Each visualization includes hover functionality to display precise values and can be pinned in place. The panel's bar widget continues displaying the last eight samples of the most recent chart even when closed. ## Development and testing The plugin is built as a self-contained package requiring no build step, with all necessary components committed directly in the repository root. Development utilizes yeetkit-omarchy framework packaged in the vendor directory. Testing includes comprehensive unit tests for the parser, drawing functions, cell runtime, and agent loop. The system includes a terminal-based testing script (ask.js) that demonstrates the complete query-to-visualization pipeline without UI dependencies. Continuous development is supported through hot-reloading to the Omarchy plugins directory. ## Installation and authentication Installation requires a one-line curl command to install yeet followed by plugin activation. Users must authenticate through yeet login before accessing the visualization features. The authentication flow handles OAuth through a browser while maintaining existing visualizations during the process. The panel provides clear feedback for authentication states and errors. Multiple installation methods are documented, including manual setup for customized deployments. The plugin maintains security by running in isolated environments without requiring open network ports. ## Model interaction flow When a query is submitted, the system first introspects the host's system graph and includes the compact SDL schema in the prompt. The AI model uses this schema to construct valid queries, with the system automatically handling nullable fields. Responses stream in markdown format with embedded chart directives. Each :::chart block specifies visualization parameters like min/max values, units, and update intervals. The system executes these blocks immediately upon receiving the closing delimiter, even while the model continues generating additional content. ## Chart customization options Visualizations support extensive customization through block attributes. Users can specify axis ranges (min/max), measurement units (%, B, B/s), and display density (rows). Charts maintain history across rewrites when given consistent IDs. The system offers twelve visualization types: area, line, overlay, stacked, split, heat, sparks, gauge, stat, meters, bars, and pie. The AI model receives guidance on selecting appropriate chart types based on the data characteristics before generating the visualization code. ## Error handling and debugging Failed visualizations display error messages in the theme's urgent color. The system implements a repair mechanism that automatically sends errors back to the model for correction, limited to two attempts per query. Developers can inspect running chart code through the src option, crucial since visualizations are generated by AI. Terminal-based debugging is available through ask.js, which replicates the panel's functionality while exposing error output normally hidden in the UI flow. ## Operational implications This tool represents a significant shift in system monitoring by combining generative AI with real-time data visualization. It eliminates the need for pre-configured dashboards while providing deeper system visibility than traditional CLI tools. The architecture demonstrates how AI can be safely integrated into system administration through isolated execution environments. The project builds on yeet's system graph to create a new paradigm where monitoring interfaces are generated dynamically based on immediate operational needs. ## Sources covering this story - https://github.com/yeet-src/omarchy-yeet-ai — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._