Skip to main content

Impact Analysis

The Impact Analysis page at /impact answers: "If I drop / rename / restructure this table, what breaks?" It's a pre-flight blast-radius report covering views, functions, jobs, queries, and dashboards that read from the target object.

What it analyses​

For a given table or schema, the analyser checks six dimensions:

DimensionSource
Affected viewsUC dependency graph
Affected functionsUC dependency graph
Downstream tablesClone-Xs lineage + UC system lineage
Referencing jobsDatabricks Jobs API + task config scan
Active queriesQuery history (last 24h by default)
Dashboard referencesLakeview dashboard catalog scan

Each dimension reports a count card and an expandable detail table.

Workflow​

  1. Pick Catalog → Schema → optional Table
  2. Click Analyze Impact
  3. Watch the six count cards populate
  4. Optionally click Explain with AI to get a markdown narrative summarising risk

Risk level​

A top-of-page card shows an overall risk level computed from the spread of impact:

  • Low — only views/functions affected, no live consumers
  • Medium — < 5 jobs or dashboards reference it
  • High — > 5 jobs/dashboards, or recent active queries detected

The risk level adjusts the Explain with AI prompt context so the narrative emphasises the right concerns.

AI narrative​

The Explain with AI button calls:

POST /ai/summarize
{ "context_type": "report", "data": { ...impact JSON... } }

Returns a markdown summary suitable for sharing in a PR description or change ticket. Dismiss the panel to hide it.

Detail tables​

Each affected category has its own datatable with grouping, filtering, search, and CSV export. Hover any FQN to reveal a copy-to-clipboard button.

API​

POST /impact
{
  "catalog": "prod_warehouse",
  "schema": "sales",
  "table": "orders"
}

Returns one block per dimension plus an overall_risk field.