Data Visualization & Alerts

Turn check-in output into AI-generated charts and alert rules

Overview

Data Visualization & Alerts turns monitor check-ins into charts and alert rules. Send your job output; Telemetry.host can infer what to visualize and what alerts to suggest.

In simple words: send your output and let AI find what matters: values to chart, trends to track, and alerts to set.

This works best when a monitor has several recent check-ins with similar shape. The data can be structured JSON, plain text, command output, logs, or a mix of fields and messages.

Sending Data

You can send any text output directly to the ping endpoint:

/usr/local/bin/backup.sh 2>&1 | curl -X POST \
  https://telemetry.host/ping/{YOUR_MONITOR_ID} \
  -H "Content-Type: text/plain" \
  --data-binary @-

That output might look like this:

Backup completed
duration_seconds=143
backup_size_mb=2840
warnings=0

You can also send arbitrary JSON:

curl -X POST https://telemetry.host/ping/{YOUR_MONITOR_ID} \
  -H "Content-Type: application/json" \
  -d '{
    "status": "success",
    "job": "nightly-backup",
    "duration_seconds": 143,
    "backup_size_mb": 2840,
    "warnings": 0,
    "destination": "s3://prod-backups/db"
  }'

The AI-assisted workflow can inspect recent check-ins, extract useful values, generate chart data, and suggest alert rules such as “backup size dropped too low” or “duration increased above normal.”

Reliable Curl Delivery

For production scripts, use retry and timeout options so short network problems do not hide a real job result:

/usr/local/bin/backup.sh 2>&1 | curl -X POST \
  --fail-with-body \
  --retry 25 \
  --retry-all-errors \
  --retry-connrefused \
  --retry-delay 7 \
  --retry-max-time 180 \
  --connect-timeout 5 \
  --max-time 20 \
  https://telemetry.host/ping/{YOUR_MONITOR_ID} \
  -H "Content-Type: text/plain" \
  --data-binary @-

These options make curl fail clearly on HTTP errors, retry temporary failures, retry refused connections during deploys, and avoid hanging forever.

DSL Syntax

The DATA PREP DSL is JSON that describes how check-in data becomes structured data for charts and alerts. It is not Python, shell, or JavaScript.

A pipeline can:

• Extract fields from JSON paths or text patterns
• Derive values from extracted fields
• Emit time-series points, events, or intervals for visualization
• Define alert rules that trigger when extracted or derived values cross a threshold

Basic Shape

Every DSL document is a JSON object with these top-level sections:

• version: DSL version. Use 1.
• extract: Array of rules that pull raw values from JSON or text into named fields.
• derive: Array of rules that calculate new fields from extracted fields.
• emit: Object that chooses which fields become chart series, events, or intervals.
• alerts: Array of alert rules evaluated against extracted or derived fields.

Example:

{
  "version": 1,
  "extract": [
    {
      "name": "duration_seconds",
      "type": "number",
      "path": "$.duration_seconds"
    }
  ],
  "emit": {
    "series": [
      {
        "id": "duration_seconds",
        "label": "Duration",
        "kind": "line",
        "x": "record_timestamp",
        "y": "duration_seconds"
      }
    ],
    "events": [],
    "intervals": []
  },
  "alerts": [
    {
      "id": "slow_backup",
      "when": {
        "field": "duration_seconds",
        "op": ">",
        "value": 600
      },
      "severity": "warning",
      "message": "Backup took longer than expected"
    }
  ]
}

When you edit the DSL in the dashboard, Telemetry.host validates it and previews the extracted records before the pipeline is applied.

Top-Level Keywords

extract Rules

Each extract rule should set a field name and one extraction method.

Common examples:

{
  "extract": [
    { "field": "cpu_pct", "path": "$.metrics.cpu_pct", "type": "number" },
    { "field": "backup_size_raw", "regex": "\\((?P<value>[\\d.]+\\s*[KMGTPE]?i?B?)\\)", "type": "string" },
    { "field": "disk_pct", "key_value": "disk_usage", "type": "number" },
    { "field": "error_lines", "grep": "ERROR", "aggregate": "count", "type": "number" }
  ]
}

Regex safety notes:

• Patterns are case-insensitive and multiline.
• Backreferences, lookarounds, and nested quantified groups are rejected.
• Very long patterns and slow-running patterns are rejected.

Type Coercion

type controls how raw values become prepared values:

derive Rules

Derived fields are calculated after extract.

Operation details:

Examples:

{
  "derive": [
    { "field": "duration_hours", "op": "duration", "start": "started_at", "end": "finished_at", "unit": "hours", "precision": 2 },
    { "field": "backup_size_mb", "op": "parse_size", "source": "backup_size_raw", "unit": "mb" },
    { "field": "job_label", "op": "copy", "source": "job_name" },
    { "field": "insert_day", "op": "date", "source": "insert_datetime" }
  ]
}

emit Rules

emit decides what the charting layer can use.

Series keywords:

Event keywords:

Interval keywords:

alerts Rules

Alerts run against prepared fields for each check-in.

Condition keywords:

Example:

{
  "alerts": [
    {
      "id": "disk_almost_full",
      "when": { "field": "disk_pct", "op": ">=", "value": 90 },
      "severity": "critical",
      "message": "Disk usage is above 90%"
    }
  ]
}

Complete DSL Example

This example works well for text output like:

status=success
backup_size=762K
disk_usage=91
started_at=23:53:02
finished_at=00:12:02

DSL:

{
  "version": 1,
  "extract": [
    { "field": "status", "key_value": "status", "type": "string" },
    { "field": "backup_size_raw", "key_value": "backup_size", "type": "string" },
    { "field": "disk_pct", "key_value": "disk_usage", "type": "number" },
    { "field": "started_at", "key_value": "started_at", "type": "string" },
    { "field": "finished_at", "key_value": "finished_at", "type": "string" }
  ],
  "derive": [
    { "field": "backup_size_mb", "op": "parse_size", "source": "backup_size_raw", "unit": "mb" },
    { "field": "duration_minutes", "op": "duration", "start": "started_at", "end": "finished_at", "unit": "minutes", "precision": 1 }
  ],
  "emit": {
    "series": [
      { "id": "backup_size_mb", "label": "Backup Size (MB)", "kind": "bar", "x": "record_timestamp", "y": "backup_size_mb" },
      { "id": "duration_minutes", "label": "Duration (minutes)", "kind": "line", "x": "record_timestamp", "y": "duration_minutes" },
      { "id": "disk_pct", "label": "Disk Usage (%)", "kind": "line", "x": "record_timestamp", "y": "disk_pct" }
    ],
    "events": [
      { "ts": "record_timestamp", "label_field": "status" }
    ],
    "intervals": []
  },
  "alerts": [
    {
      "id": "disk_almost_full",
      "when": { "field": "disk_pct", "op": ">=", "value": 90 },
      "severity": "critical",
      "message": "Disk usage is above 90%"
    }
  ]
}

ECharts Configuration

Charts are rendered with Apache ECharts option JSON. The configuration controls chart type, axes, labels, legends, and series.

Telemetry.host uses safe ECharts option objects, not custom JavaScript. Keep the configuration as JSON data: no functions, scripts, or browser code.

For the complete upstream reference, see the official Apache ECharts docs:

• Chart Configuration reference
• Apache ECharts Handbook

ECharts Option Shape

An ECharts configuration is usually called an option. Telemetry.host stores and edits that option as JSON.

Common top-level keywords:

Minimal example:

{
  "tooltip": { "trigger": "axis" },
  "xAxis": { "type": "time" },
  "yAxis": { "type": "value", "name": "Duration (seconds)" },
  "series": [
    {
      "name": "Backup duration",
      "type": "line",
      "data": []
    }
  ]
}

Axes

Most Data Visualization & Alerts charts use x/y axes.

Typical monitor timeline axes:

{
  "xAxis": { "type": "time" },
  "yAxis": { "type": "value", "name": "Backup Size (MB)" }
}

Use category when the x values are labels instead of timestamps:

{
  "xAxis": {
    "type": "category",
    "data": ["backup", "verify", "upload"]
  },
  "yAxis": { "type": "value" }
}

Series

series tells ECharts what to draw. Each item is one line, bar set, scatter plot, or other chart layer.

Common series keywords:

Examples:

{
  "series": [
    {
      "name": "Backup duration",
      "type": "line",
      "smooth": true,
      "data": [
        ["2026-05-01T02:00:00Z", 143],
        ["2026-05-02T02:00:00Z", 151]
      ]
    },
    {
      "name": "Warnings",
      "type": "bar",
      "data": [
        ["2026-05-01T02:00:00Z", 0],
        ["2026-05-02T02:00:00Z", 2]
      ]
    }
  ]
}

Tooltips And Legends

Use tooltip.trigger: "axis" for time-series charts so hovering one timestamp shows all nearby values.

{
  "tooltip": {
    "trigger": "axis"
  },
  "legend": {
    "top": 0
  }
}

Dataset And Encode

For more complex charts, ECharts can read tabular data from dataset.source and map columns with encode.

{
  "dataset": {
    "source": [
      ["timestamp", "duration_seconds", "backup_size_mb"],
      ["2026-05-01T02:00:00Z", 143, 762],
      ["2026-05-02T02:00:00Z", 151, 801]
    ]
  },
  "xAxis": { "type": "time" },
  "yAxis": { "type": "value" },
  "series": [
    {
      "name": "Duration",
      "type": "line",
      "encode": { "x": "timestamp", "y": "duration_seconds" }
    }
  ]
}

Useful Monitor Chart Patterns

Line chart for a numeric value over time:

{
  "tooltip": { "trigger": "axis" },
  "xAxis": { "type": "time" },
  "yAxis": { "type": "value", "name": "Duration (seconds)" },
  "series": [
    {
      "name": "Duration",
      "type": "line",
      "data": []
    }
  ]
}

Bar chart for per-run counts:

{
  "tooltip": { "trigger": "axis" },
  "xAxis": { "type": "time" },
  "yAxis": { "type": "value", "name": "Warnings" },
  "series": [
    {
      "name": "Warnings",
      "type": "bar",
      "data": []
    }
  ]
}

Two metrics with separate y-axes:

{
  "tooltip": { "trigger": "axis" },
  "legend": {},
  "xAxis": { "type": "time" },
  "yAxis": [
    { "type": "value", "name": "Duration (seconds)" },
    { "type": "value", "name": "Size (MB)" }
  ],
  "series": [
    {
      "name": "Duration",
      "type": "line",
      "yAxisIndex": 0,
      "data": []
    },
    {
      "name": "Backup size",
      "type": "bar",
      "yAxisIndex": 1,
      "data": []
    }
  ]
}

Telemetry.host Safety Rules

When editing ECharts JSON in Telemetry.host:

• Use JSON only. Do not add JavaScript functions, callbacks, or scripts.
• Keep series present and non-empty.
• Prefer line, bar, and scatter for monitor data.
• Prefer xAxis.type: "time" for check-in timelines.
• Keep data points as JSON arrays or dataset rows.

When you edit the ECharts configuration in the dashboard, Telemetry.host validates that it can still render a usable chart.