Analytics

Analytics lets you connect KubeWatch directly to your own external data, a Postgres database, a MySQL database, a Redshift warehouse, or an S3 bucket, define a data-cleaning pipeline that runs natively inside KubeWatch, and visualize the cleaned result as a chart on the dashboard. No external BI tool required. Viewing the page is open to any org member; creating or changing connections, jobs, or saved charts is gated by admin role. Neither is gated by plan tier: Analytics is available on every plan.

What it does

• Connections: link a Postgres, MySQL, or Redshift database, or an S3 bucket, as a data source. Credentials are encrypted at rest.
• Jobs: define a cleaning pipeline against a connection and run it, on demand or on a recurring interval, to produce a cleaned dataset.
• Datasets: browse the output of each run: row counts, column types, and the cleaned data itself.
• Dashboards: build and save line, bar, or area charts straight from a cleaned dataset.

These map to the four tabs on the Analytics page: Connections, Jobs, Datasets, and Dashboards.

Connecting a data source

From the Connections tab, add a new connection and choose a type:

• Postgres / MySQL: host, port, database name, username, and password.
• Redshift: cluster endpoint, port, database name, username, and password, same shape as a Postgres connection.
• S3: bucket name, region, and an access key ID / secret access key pair with read access to the bucket.

Building a cleaning job

A job runs against one connection and applies a pipeline of steps, in order, to produce a cleaned dataset. Six step types are available:

• Deduplicate: drop duplicate rows.
• Fill missing values: replace missing values in a column with a fixed value.
• Drop missing values: drop rows where a column is missing.
• Coerce column type: convert a column to a different type (e.g. text to number).
• Filter: keep only rows matching a simple expression, e.g. amount > 100.
• Custom transform: compute a new or replacement column value from an expression, e.g. amount * 1.1.

The filter and custom transform steps use a small, safe expression language, not a full scripting language. An expression can only compute a value from that row's own columns: it has no file access, no network access, and no way to execute arbitrary code.

Triggering and viewing a run

A job can be run manually from the Jobs tab, or on a recurring interval (every N minutes). There's no cron syntax, just a plain interval.

A cleaning run reads up to 100,000 source rows (or up to 200MB from a single S3 object), and produces a cleaned dataset capped at 50,000 rows. Analytics is a dashboarding tool built for exploring and visualizing a working slice of your data, not a bulk-ETL system or a replacement for a data warehouse.

Once a run finishes, its output appears as a new dataset on the Datasets tab, alongside the row count and column types KubeWatch inferred.

Building a chart

From the Dashboards tab, pick a cleaned dataset, choose a chart type (line, bar, or area), and map columns to the chart's axes. Save the chart to keep it on the Analytics dashboard, where it re-renders against the latest run of its underlying dataset.

Access control

Any authenticated org member, including a viewer, can open the Analytics page and browse existing connections, jobs, datasets, and dashboards. Adding or editing a connection, creating, editing, deleting, or running a cleaning job, and creating or deleting a saved chart all require the admin role.