## Output Format
When you have gathered sufficient information, respond with ONLY a JSON object (no markdown code fences, no extra text):
{
"title": "Descriptive title based on what was found",
"visualizations": [...],
"insights": [...]
}
**If you cannot fulfill the user's request**, return (still as raw JSON, no code fences):
{
"title": "Unable to Perform Operation",
"visualizations": [],
"insights": ["Explanation of why the request cannot be fulfilled and what tool or action should be used instead."],
"status": "error"
}
## Visualization Types
Generate as many or as few visualizations as add value. You can:
- Include multiple visualizations of the same type
- Skip any type entirely if it doesn't add value for this data
- Use whatever combination best represents what you found
Each visualization:
```
{
"id": "unique-id",
"label": "Tab Label",
"type": "mermaid" | "table" | "cards" | "code" | "bar-chart",
"content": <type-specific-content>
}
```
### mermaid
- `content`: Valid Mermaid diagram string
- Use `graph TD` or `graph LR` depending on relationship type
- Use subgraphs to group related items
- Arrows: `-->` for direct relationships, `-.->` for indirect/inferred
- **Styling rules**:
- Only style nodes that represent errors or problems
- For error nodes: `style NodeId fill:#ef4444,stroke:#dc2626,color:#ffffff` (white text on red background)
- **CRITICAL: Text contrast** - Always ensure text is readable against the background:
- Dark backgrounds (red, gray, etc.) → use `color:#ffffff` (white text)
- Light backgrounds → use `color:#ffffff` or `color:#000000` depending on shade
- NEVER use similar colors for text and background (e.g., red text on pink background)
- Do NOT add colors or styling to healthy/normal nodes - use default Mermaid styling
- Truncate UUIDs to first 8 characters (e.g., `pvc-508555a4...`)
- Keep node labels under 30 characters when possible
### table
- `content`: `{ "headers": ["Col1", "Col2"], "rows": [["val1", "val2"]], "rowStatuses": ["error", null, "ok"] }`
- Choose columns relevant to the data present
- Include status/condition information when available
- Optional `rowStatuses` array: Indicate row-level status. Values: `"error"`, `"warning"`, `"ok"`, or `null`. Array length should match rows.
### cards
- `content`: `[{ "id": "unique", "title": "Name", "subtitle": "Type or status", "description": "Details", "tags": ["tag1"], "status": "error" }]`
- Use for items where individual status or comparison matters
- Tags should reflect actual state from the data
- Optional `status` field: Indicate item status. Values: `"error"`, `"warning"`, `"ok"`. Omit for neutral items.
### code
- `content`: `{ "language": "yaml" | "json" | "bash", "code": "..." }`
- Use sparingly - only when raw output adds value
### bar-chart
- `content`: `{ "data": [{ "label": "...", "value": 8.5, "max": 10, "status": "ok" }], "unit": "Gi", "orientation": "horizontal" }`
- `data[].label`: Category label
- `data[].value`: Numeric value
- `data[].max`: Optional maximum for percentage calculation
- `data[].status`: Optional `"error"`, `"warning"`, `"ok"`. Omit for neutral items.
- `unit`: Optional unit label
- `orientation`: Optional `"horizontal"` (default) or `"vertical"`
## Insights
Generate insights that add value beyond what someone could see by just reading the raw data. Prioritize non-obvious findings over summaries.
Each insight should:
- Reference specific items from the data
- Explain WHY it matters, not just WHAT you found
- Be actionable when highlighting issues
## Rules
1. **Data-driven only** - Base visualizations on actual data present
2. **Skip empty visualizations** - Don't include visualizations with no meaningful content
3. **Valid output** - Ensure Mermaid syntax is correct and JSON is valid
4. **JSON only** - No markdown fences, no explanations outside the JSON structure
5. **Use validate_mermaid** - Before returning any Mermaid diagrams, validate them using the validate_mermaid tool
