KloudMate now includes a Profiling module: continuous CPU and memory profiling that turns sampled data into an interactive flame graph, so you can go from "this service is slow" to a specific function and line number, without attaching a debugger or shipping a new build.
Five views, each answering a different question:
1. All Services:
a grid of every service reporting a given profile type, with a peak value and mini trend chart per card, your entry point2. Profile Types:
pick one service and see every type it reports (CPU, heap, goroutines, etc.) side by side3. Labels:
break a service's profile down by resource attributes (region, pod, version) to isolate where a spike is coming from4. Flame Graph:
the actual hotspot-finding view, zoomable, with sandwich mode (see a function's callers and callees), a sortable function table, and an Ask AI button that hands the top 10 functions by self time to the KloudMate Assistant5. Difference View:
compare two time windows of the same service, merged (one tree colored by change) or side by side, the view for "did that deploy make things better or worse?"Getting data in:
KloudMate accepts OpenTelemetry Profiles or the Pyroscope protocol, with dedicated SDK guides for Go, Java, Python, Node.js, Ruby, .NET, and Rust.
Check out the Profiling documentation for full details, including per-language SDK setup guides.