Fluxtion Audit Log Analyser¶
A desktop canvas for Fluxtion audit evidence. A Mongoose/Fluxtion EventProcessor writes down every cycle it runs:
the event that arrived, and each node that logged, in order, with what it wrote. The analyser indexes that record,
so multi-GB logs stay responsive. You, or an AI you connect, can then browse it, filter it, graph it, step through it
on the processor's own topology, and report on it.
You do not have to have written the system. Nobody wrote the log up afterwards: the processor emitted it as it ran. So it answers a support engineer in their first week as readily as the person who built the graph, and an AI's conclusion arrives as something on screen you can check, not prose you have to trust.
Choose your door¶
-
Diagnose an incident
Something went wrong in a running system. Point Claude Code, Codex or another MCP client at the analyser you have open, ask in your own words, and watch the client filter, plot and flag the evidence on your screen. Or ask the assistant built into the analyser.
-
Build with Claude
You state the intent and the acceptance criteria. The AI structures the components, Fluxtion writes the mechanics, and you check your criteria against the recorded evidence of your runs. Whether the criteria capture what you meant stays your judgement.
-
Get started
Install with one command, open the built-in DEMO (Explore DEMO ▸ Guided tour on the start page), and find the route that fits what you came to do. No AI and no key are needed to begin.
-
Administer and secure
What each assistant profile can do, how external access is opened and closed, what is stored on the machine, and what leaves it.
Run it now¶
Install JBang once (it fetches Java 21 if you have none), then run the analyser:
With no log to hand, choose 1. Run me through a demo on the start page and press Play: it opens a recorded
DEMO run and its topology from inside the jar. Install & run has the plain-jar download, building from source, and --rest for an
agent.
One log, every view¶
The audit log is the single source of truth, and every feature reads the same records. Narrow the view once and the table, the charts and the topology all follow.
- Records, detail & filtering. A columnar index keeps browsing and filtering fast; files over a size threshold are memory-mapped rather than loaded into the heap.
- Graphs. Any top-level numeric or boolean node value over time, with formula series, guides, bands and markers.
- Topology & step-through. The processor's own node graph, which nodes ran in a cycle, and which never logged at all.
- Source navigation. A node log line opens the class and method that wrote it.
- Investigation reports and evidence bundles. An account of what you found, and a verifiable package for someone who was not watching.
Why it matters¶
- Minutes to root cause, with evidence. An incident in an event-driven system usually means grepping a large log and reconstructing by hand what fired, in what order. Here the log is indexed in seconds, any value plots over time, two cycles diff in a click, and a log line opens the source method that wrote it.
- The audit log becomes an asset. The bytes you already keep become incident forensics (what actually ran, with its state), a compliance answer (why this order was amended, cycle by cycle, traceable to source) and a shared team artefact.
- Only possible with Fluxtion. A Fluxtion processor is deterministic and its graph is generated from source, so a record is not just a log line: it is a trace of one cycle with an address in the code.