Dashboard
Build a widget from a prompt
Describe a chart in plain words and Prodgator builds a dashboard widget from your data.
Availability
Enterprise plan (AI insights), with AI features turned on for your organization. Anyone in the organization can generate widgets, readers included.
Describe the widget you want in plain language. Prodgator's AI turns the prompt into a widget specification, checks it against your data and permissions, and draws the result on your dashboard.
Each generation spends 1 AI credit from your organization's shared monthly allowance, and counts toward the per-person and per-organization request limits.
How widget generation works
- You describe the widget you want, for example "Show me the count of failed runs by repository over the last 7 days".
- Prodgator sends your prompt to the AI model. None of your data is sent.
- The model returns a widget specification: the chart type, data source, filters and time range.
- Prodgator checks the specification against the data sources and fields it knows about, your organization's data retention, and your permissions (you only see data you already have access to).
- Prodgator runs the query against your data and draws the chart.
Data sources
Widgets can be built from five data sources:
| Source | Includes | Dimensions | Numbers |
|---|---|---|---|
| Runs | Pipeline runs from your workflows | Status (passed, failed, skipped and others), repository, workflow, branch, provider, failure category | Duration (seconds), importance score |
| Deployments | Deployment records from your deployment environments | Status (pending, approved, deployed, failed and others), environment, repository, workflow, branch | Time to deploy, time to first approval decision |
| Findings | Security findings (GHAS, Dependabot, SARIF) | Severity, type, status, repository, package | Age (days since detection), time to fix |
| Approvals | Approval decisions on deployments | Decision (approved or rejected), environment, repository, approver, approval type | Wait time (seconds from deployment creation to approval) |
| Notifications | Org-wide and your personal notifications | Type, severity, category | None |
Supported chart types
| Chart | Use it for | Needs |
|---|---|---|
| Number | One value, such as "Total failed runs this week" | No grouping, no time buckets |
| Line | A time series, such as "Failed runs per day" | Time buckets |
| Bar | Values by a dimension or over time | Grouping, time buckets, or both |
| Stacked bar | Several categories over time | Grouping and time buckets |
| Pie | Shares of a category, such as "Failed runs by repository" | Grouping with a count or sum |
| Table | Grouped data | Grouping by one or more dimensions |
Widget examples
Requests that work
- "Show failed runs by repository in the last 7 days"
- "What is the average time to deploy by environment this month?"
- "Pie chart of security findings by severity"
- "Line graph of deployment counts per week over the last 30 days"
- "How many runs were skipped vs passed this week?"
Requests that do not work
- "Show test counts per run": Prodgator does not store test counts or coverage.
- "Cost per deployment": Prodgator does not track deployment costs.
- "Number of reviewed PRs by developer": Prodgator does not track review data.
- "Cross-reference findings with test failures": a widget cannot join data across sources.
- "Calculate the deployment success rate and remaining budget in one widget": a widget cannot run formulas.
When a request asks for something Prodgator cannot do, the answer says "unsupported" and explains why. Change the request, or edit the generated specification.