An agent explains. You understand.
Explore a real mapThere’s more, when you want it
Three small pieces. Use them together with your agent, or take just the one you need.
import { mount } from './map-viewer/index.js';
mount(element, { text: markdown });One prompt. So many connections. Two skills teach your agent to draw maps.
illuminate-conceptone ideatrace-architecturea whole projectThe map on this page came from one trace-architecture prompt:
Your next moment of clarity
/plugin marketplace add nvsravank/avagahana /plugin install avagahana@avagahana
codex plugin marketplace add nvsravank/avagahana codex plugin add avagahana@avagahana
Python 3.9+ required. Open source, MIT licensed: see it on GitHub.
“Review this codebase and create an Avagahana map of its architecture. Show me the components, how a request flows, and the data model.”
A little more, if you’re curious
అవగాహన is Telugu for understanding, comprehension and awareness. That’s the point: make the connections clear enough to build your own mental model.
No. An agent generated the file earlier. The viewer renders that file in your browser. Creating or changing a map with an agent uses that agent’s model allowance or API usage.
The viewer is for reading and exploring. You or your agent edit the Markdown file; the viewer turns it into an interactive explanation.
When you work with an agent, the viewer, Atlas and relay work together to show and update your maps. To publish a finished map on a website or wiki, embed the viewer with the Markdown file. That published view needs static hosting.