New in Psyclone AIOS v2.3 — shipped 19 November 2026.
PsyProbe 2.0 is the monitoring and control application built into every Psyclone node. Point a browser at the node’s port, log in, and you see the whole running system live — every node, process, component and connection — with ten minutes of history, the messages of the last three minutes, logs, alerts and a console. Nothing has to be instrumented: the runtime already knows what is running, and PsyProbe shows it. Since 2.3 it is a complete rewrite, with accounts, HTTPS, a permission-checked API and room for your own views.
Why it matters
A real-time system that misbehaves in production rarely does so on a developer’s laptop. The question at 2am is always the same: which component is backing up, which message stopped arriving, and what changed? PsyProbe answers it from the live system itself, without restarting anything or attaching a debugger, so you look inside the system that has the problem instead of trying to reproduce it in a lab.
It is also built to be left running in front of real users. Logins have roles, every change is audited, destructive actions ask for confirmation, and secret-looking values stay masked. Monitoring tools, scripts and AI agents use machine accounts with API keys, so the same live data can feed a Prometheus/Grafana stack or an automated operator with exactly the rights it needs.
How it works
Psyclone samples every component, node and connection once a second, keeps ten minutes of history and a masked summary of recent messages, and streams only what changed over a WebSocket — a typical system costs tens of bytes per second per viewer, and the sampler about half a millisecond per second. The application then shows that live model in several ways:
- Graph, 2D or 3D: nodes and processes as boxes, components inside, connections as arrows whose width follows the bytes per second, coloured by CPU, message rate, bytes or queue. By function regroups components by the
group=you give them in the psySpec. - Components: one dense row per component — rates in and out, totals, queue, cycles, CPU, memory, triggers and a one-minute sparkline.
- Inspector: select anything for its charts, live messages, triggers and posts, private data and logs.
- Operate: edit parameters (with the descriptions and units from your psySpec), test any trigger or post with one click, stop, start or restart a component, acknowledge alerts, and run console commands.
- Federation: several Psyclone systems can appear side by side, with signed identity assertions, CA pinning and role caps deciding what each one lets you do.
How to use it
An SSL build — every binary distribution is one — serves PsyProbe 2 on the main port. The first visit asks for the setup code printed in the Psyclone log and creates the first admin. For access from other machines, add a TLS port; PsyProbe issues its own certificates from a built-in CA, or a free ACME certificate for a public host name. Alerts, views and apps are declared in the psySpec:
<psyprobe tlsport="10443">
<alert name="asr_backlog" entity="ASR" metric="queue" above="50" for="5"/>
</psyprobe>
<module name="Tuner" group="Dialogue/Voice">
<parameter name="Gain" type="Integer" value="[0:100:5]=50"
description="How loud the reply is" unit="%" /> <!-- a slider in PsyProbe -->
<view name="Timeline" path="views/timeline/" /> <!-- your own inspector tab -->
</module>
- Open
https://<host>:10443/v2/, enter the setup code, create the admin account. - Create machine accounts and API keys for monitoring (
/metricsfor Prometheus) and scripts (/api/v2/...). - Give components a
group=, parameter descriptions and units, and alert rules, so the live view speaks your system’s language. - Add a
<view>for what only your component knows — a map, a timeline, a chart — using the small probe SDK, or a whole application page for end users.
When to use it / when not
- Use it for development, commissioning and day-to-day operation: watching load, finding the stalled component, tuning parameters live, and giving operators a safe control surface.
- Use it as the data source for your existing monitoring:
/metricsand the REST API are permission-checked and API-key based. - Not a long-term store: live history covers ten minutes. Keep longer metrics in Prometheus, and longer message history in recordings, which PsyProbe can open.
- Note: PsyProbe 1.x and its
/api/...calls are gone in 2.3; old URLs answer410 Gonenaming the PsyProbe 2 call that replaces them.
Read the docs
- User Guide 12 — PsyProbe — the application, parameters, your own views and apps, live updates.
- System Guide 3 — PsyProbe configuration — accounts and roles, federation, metrics and alerts.
- System Guide 5 — Security & SSL — HTTPS, the built-in CA and ACME certificates.
- The full PsyProbe 2 source (Svelte 5 + TypeScript) ships in the LGPL source distribution on the Download page.
PsyProbe 2.0 is one of the two headline 2.3 capabilities. See also Edge & Device Services for connecting phones, browsers and devices, Whiteboards & Dataflow for what flows through a system, or return to the Psyclone AIOS overview.
