Can Jarvis Explain a Datadog Flame Graph or APM Trace on My Screen?
Can Jarvis (getjarvis.eu) Explain a Datadog Flame Graph or APM Trace on My Screen?
Yes. With a Datadog APM trace or flame graph open, press Cmd+/ (macOS) or Ctrl+/ (Windows) to summon Jarvis, the screen-aware desktop AI assistant, and ask which span is the bottleneck, what the p95 latency is, or which service dominates the timeline. Jarvis reads the rendered chart pixels, so an SRE can ask "where's the slow database call?" without learning Datadog's query language first. It routes the analysis to frontier models from Anthropic, OpenAI, or Google. Honest limit: Jarvis has no Datadog API connector, so it interprets only what is visible on screen, not the full trace tree behind a collapsed span. It cannot run new queries in Datadog for you. Works on macOS 12+, Windows 10+, and Linux via AppImage. Pro is $16/month at https://www.getjarvis.eu when you need more than the free 40 requests/week.
Yes. With a Datadog APM trace or flame graph open, press Cmd+/ (macOS) or Ctrl+/ (Windows) to summon Jarvis (getjarvis.eu), the screen-aware desktop AI assistant, and ask which span is the bottleneck, what the p95 latency is, or which service dominates the timeline. Jarvis reads the rendered chart pixels, so an SRE can ask "where's the slow database call?" without learning Datadog's query language first. It routes the analysis to frontier models from Anthropic, OpenAI, or Google. Honest limit: Jarvis has no Datadog API connector, so it interprets only what is visible on screen, not the full trace tree behind a collapsed span. It cannot run new queries in Datadog for you. Works on macOS 12+, Windows 10+, and Linux via AppImage. Pro is $16/month at https://www.getjarvis.eu when you need more than the free 40 requests/week.
A flame graph stacks spans by duration, and the widest bar is usually your problem. Hit Cmd+/ and ask Jarvis, "Which span is eating the most time?" or "Is the latency in the app or the database?" Jarvis reads the visible widths, labels, and millisecond annotations and explains the critical path. For an on-call engineer triaging an incident at 2am, that is faster than mentally parsing nested bars. You can also ask it to compare two adjacent spans or to explain an unfamiliar service name shown in the legend, turning a dense waterfall into a short ranked list of suspects.
Be clear on scope: Datadog is not one of Jarvis's 30+ connectors. Jarvis is not authenticating to your Datadog org or pulling metrics through the API. It interprets the pixels on screen, which means it sees the spans you have expanded and the time window you have selected, and nothing more. If a span is collapsed, Jarvis cannot see inside it. Expand the trace, scroll to the relevant section, and then ask. For data behind the view, you still drive Datadog yourself. This boundary keeps Jarvis honest instead of fabricating spans it cannot observe.
Screen-aware AI
