Skip to content
Osuite Osuite Docs

Visualizer Agent

The Visualizer Agent builds dashboards from plain English. Describe what you want to see and it figures out which metrics to use, writes the PromQL, picks the right chart types and time ranges, and assembles the dashboard — no query language or metric expertise required. Like the other Osuite agents, it runs inside your AI IDE.

Building a dashboard the manual way means first knowing which metric carries the signal you want, then knowing how to express it in PromQL, then choosing a sensible visualization. The Visualizer Agent collapses all three steps into one sentence.

Prerequisites

Section titled “Prerequisites”
  • The Osuite CLI installed and the agent skills set up — follow the one-time agent setup. Once done, the Visualizer Agent is available as /osuite-edit-dashboard.

Using the agent

Section titled “Using the agent”

In your AI IDE, describe the dashboard or widget you want:

Terminal window
/osuite-edit-dashboard add a panel showing p99 latency by service for the payments namespace

The agent works against your real metrics, so it never invents a metric name or a label:

  1. Discovers the right metric with semantic search over your metric catalog (osuite metrics search-embedding), then inspects its labels and label values.
  2. Composes the PromQL — applying the correct function for the metric type (rate() for counters, histogram_quantile() for histograms, aggregation for gauges) and the labels you described.
  3. Builds the widget — choosing a chart type that fits the data (time series, gauge, stat, pie, bar) and a unit (bytes, percentage, duration).
  4. Shows you a plan of the change before anything is applied.
  5. Applies it to your Osuite dashboard once you approve.

You always keep full control: the agent is a fast starting point, and every dashboard it produces remains fully editable in the Custom Dashboard builder.

Dashboards as code

Section titled “Dashboards as code”

Under the hood the agent treats dashboards as versionable JSON and drives them through the CLI:

Terminal window
osuite dashboard pull            # fetch existing dashboards as local JSON
osuite dashboard plan board.json # preview the change
osuite dashboard apply board.json

This means dashboards built by the agent can be reviewed, committed to source control, and applied like any other config — not just clicked together in a UI.

What it’s good for

Section titled “What it’s good for”
  • Standing up a service dashboard in seconds instead of an afternoon of PromQL
  • Letting any engineer — regardless of Prometheus experience — build production-ready views
  • Turning a finding from the Investigation Agent into a dashboard you can monitor going forward

If you only need the query text itself — for an alert rule or an ad-hoc lookup, not a full panel — use the Write PromQL Agent instead.

Next steps

Section titled “Next steps”
  • Dashboards — the manual builder, pre-built dashboards, and how dashboards work in Osuite
  • Write PromQL Agent — compose a query without building a whole dashboard
  • Investigation Agent — find the root cause of an incident, then visualize the metric that exposed it