Workflow Pulse

Live heartbeat view of every active pipeline — latency, throughput, and health in a single dashboard.

Overview

Workflow Pulse is a real-time health dashboard that surfaces the operational status of every pipeline in a compact, at-a-glance view. Unlike the Monitoring log (which shows individual events) or the Heatmap (which shows historical density), Pulse focuses on the current state of each workflow: is it healthy, degraded, or down?

Pulse Cards

Each workflow is represented by a pulse card showing:

Status indicator — a colour-coded dot: green (healthy), amber (degraded / elevated error rate), red (failing or stalled).
Throughput — executions per minute over the last 5 minutes.
P95 latency — 95th-percentile execution duration, updated every 30 seconds.
Error rate — percentage of executions that errored in the last 15 minutes.
Last run — timestamp of the most recent execution.

Use Pulse as an Ops Dashboard

Leave Pulse open on a secondary monitor to get immediate visual feedback whenever a workflow goes into an error state. The status dots update in real time without requiring a page refresh.

Health Thresholds

Pulse calculates workflow health automatically using these thresholds:

Green — error rate below 5%, P95 latency within expected range.
Amber — error rate 5–20%, or P95 latency more than 2× the 7-day average.
Red — error rate above 20%, or no successful execution in the last scheduled interval.

Thresholds Are Per-Workflow

Amber and Red thresholds are calculated relative to each workflow's own historical baseline, not a global fixed number. A workflow that normally takes 10 seconds turning amber at 20 seconds is different from a fast workflow that normally completes in 1 second.

AI Assistant

Workflow Pulse ships with a built-in AI Assistant. Click the vertical AI Assistant tab on the right edge of the screen (or the Sparkles toggle) to open a chat panel that reads the current health state of every monitored workflow — status, BPM, latency, error rate, success rate, and throughput. Because this is a dashboard/overview screen rather than a record-editing table, the assistant is scoped to dashboard-style capabilities only — it narrates and explains the metrics already shown, it never edits workflows:

Executive summary — a plain-language readout of overall fleet health, calling out how many workflows are healthy vs warning vs critical vs stopped, and naming the most concerning ones.
Root cause & anomaly detection — for workflows in warning or critical state, explains what is driving it (high error rate, high latency, low throughput) by name.
Recommendations — prioritizes which workflows need attention first, referencing the actual workflow names and numbers.

Every reply can be saved with the Save as Report button, which stores it in My Reports under a title derived from your question and the current month/year. The panel respects the Writer Engine selected in Settings → Agentic (Local CLI, Hosted API, Local LLM, or Ollama) — Ollama and other local/in-browser engines run entirely on your own machine and never proxy through the server.

Frequently Asked Questions

How is Workflow Pulse different from the Monitoring log and the Execution Heatmap?

Monitoring shows a chronological feed of individual execution events, and the Heatmap shows historical execution density over time. Pulse is different from both — it shows only the current health state of each workflow (healthy, degraded, or down) as a set of live-updating cards, so you can tell at a glance which pipelines need attention right now without scrolling through logs or historical charts.

How often do the pulse cards refresh?

Throughput is calculated over a rolling 5-minute window, P95 latency updates every 30 seconds, and error rate is calculated over the trailing 15 minutes. The status dot (green/amber/red) and "Last run" timestamp update in real time as new executions complete — you never need to manually refresh the page.

Why did a workflow turn amber even though its error rate is 0%?

A workflow can turn amber for two independent reasons: an error rate between 5–20%, or a P95 latency more than 2× its own 7-day average. Because thresholds are calculated per-workflow against its own historical baseline (not a fixed global number), a normally fast workflow can go amber purely on a latency spike even with zero errors.

What does a red status mean, and how quickly does it clear?

Red indicates either an error rate above 20% or no successful execution during the workflow's last scheduled interval. The status re-evaluates continuously, so as soon as new executions bring the error rate back under the amber threshold (or a successful run occurs), the card returns to amber or green automatically — no manual acknowledgement or reset is required.