Topology, root cause and blast radius
The floor of the city shows who talks to whom as lines of light. Pulses flow from the user to the used: a web server's link to its database flows toward the database.
Where links come from
Links are never guessed from names or districts. There are three sources:
| Kind | Source | Looks like |
|---|---|---|
| declared | dependsOn: [ids] on a server or resource (you wrote it) | faint, slow |
| observed | a TCP connection really seen on a server, to another building's address and port | brighter, bows aside |
| balances | a target an AWS load balancer reports | bright, bows the other way |
| probes | dependsOn on an HTTP/TCP probe: what it really hits | faint |
Observed connections come from service discovery. They fade out after four discovery rounds without the connection. Links touching the focused building turn cyan.
hosts:
- id: lekhix-api
address: 10.0.1.12
user: ops
dependsOn: [orders-db, rate-limit-cache]
Blast radius
Every building's panel shows Blast radius: N depend on it: everything that uses it, directly or through others, nearest first. Before you restart a database, you see what goes down with it.
Root cause
When two or more things fail at once, a banner at the top names the likely root cause: a failing building whose own dependencies are healthy, with how many other failures it explains. The Doctor panel lists every root cause with a Diagnose button.
Here orders-db explains the four failures upstream of it: fix it first.
For AI agents
incident_report returns the same analysis in one call (root causes, what each explains, what is
still up but at risk), and get_topology the links and blast radius of any building. See
MCP tools.