flow-storm

FlowStorm

FlowStorm is an omniscient tracing debugger for Clojure and ClojureScript. It instruments code, records expression values and function calls as execution flows, then lets the user inspect the recorded timelines with stepping, searching, call trees, data windows, taps, and programmable APIs.

Artifacts:

• com.github.flow-storm/flow-storm-dbg {:mvn/version "4.5.9"} - full debugger UI plus runtime
• com.github.flow-storm/flow-storm-inst {:mvn/version "4.5.9"} - slim runtime for remote/debuggee processes
• com.github.flow-storm/clojure {:mvn/version "1.12.4"} - ClojureStorm compiler
• com.github.flow-storm/clojurescript {:mvn/version "1.12.134-3"} - ClojureScriptStorm compiler

Prerequisites: JDK 17+ for the UI and Clojure 1.11+ for normal artifact usage.

Quick Decision: What do you need?

Task	Go to
Set up ClojureStorm, vanilla FlowStorm, deps.edn, Leiningen, or start the UI	setup.md
Set up ClojureScriptStorm, shadow-cljs, cljs.main, or vanilla CLJS debugging	clojurescript.md
Choose instrumentation strategy, use #trace / #rtrace, instrument vars/namespaces, or tune prefixes	instrumentation.md
Use flows, stepping, Outputs, taps, data windows, datafy/nav, bookmarks, or thread breakpoints	debugging-workflow.md
Debug over nREPL, SSH tunnels, split runtime/UI processes, Docker, or remote ClojureScript	remote-and-runtime.md
Avoid too many traces, memory growth, mutable value surprises, macro issues, or unsupported instrumentation	anti-patterns.md

Core Mental Model

instrumented code -> execution -> recorded timelines -> FlowStorm UI / API

• Prefer ClojureStorm or ClojureScriptStorm for development when possible; they swap the compiler only in dev and automatically instrument code.
• Use vanilla FlowStorm when compiler swapping is not available or when you want manual #trace / #rtrace instrumentation.
• Keep recording off until just before the action you care about when the program is noisy.
• Limit instrumentation prefixes early. Do not instrument clojure.core, FlowStorm internals, or broad dependency trees without a reason.
• Clear recordings and use trace/heap limits when debugging loops, UI events, streams, or high-frequency functions.