Skip to content
Impact Gate Impact Gate Impact Gate

Route Families

Knowledge model

A route family ties a user-facing capability to the routes, source paths, specs, and flows that own it. Build the manifest with train or bootstrap, then refine it until impact analysis stays trustworthy.

What it captures

One family groups the code and tests behind a product capability

  • URL routes
  • frontend paths
  • backend paths
  • spec directories
  • user flows
  • priority
Manifest shape

The manifest lives under <testsRoot>/.e2e-ai-agents and uses a top-level families array

{ "families": [ { "id": "feature_a", "routes": ["/{workspace}/feature-a/{item}"], "webappPaths": ["src/features/feature-a/**"], "serverPaths": ["server/feature-a/**"], "specDirs": ["feature-a"] } ] }

For Playwright repos, specDirs are relative to --tests-root. For Cypress repos, use cypressSpecDirs instead.

Create it Refine it Why it matters When to update

Create It

Train

Scan your codebase and let the scanner propose families

npx impact-gate train --no-enrich --path . npx impact-gate train --path . npx impact-gate train --validate --since HEAD~50 --path .

Use --no-enrich for the deterministic offline path, train for enrichment, and --validate to measure coverage against real git history. The default --budget-usd is $0.50 and the hard cap is $10.

Bootstrap

Import route families from an Understand-Anything knowledge graph

npx impact-gate bootstrap --path . npx impact-gate bootstrap --kg-path ./knowledge-graph.json npx impact-gate bootstrap --dry-run --max-families 30 --path .

Use --kg-path when the graph lives elsewhere, --dry-run to preview the manifest, and --max-families to keep the first pass focused.

Refine It

Review loop

Check the manifest against the real product areas

  • Open <testsRoot>/.e2e-ai-agents/route-families.json
  • Add or tighten webappPaths, serverPaths, and specDirs for unmapped files
  • Re-run train --validate until the coverage and confidence look healthy
  • Commit the manifest so impact, plan, and the optional AI flows stay grounded
Helpful fields

Use the optional fields to make routing more precise

  • priority distinguishes critical flows from lower-risk ones
  • userFlows turns raw file mappings into readable scenarios
  • components and pageObjects make generation and review easier
  • features let one family break into finer-grained bindings

Why It Matters

Better decisions

A strong manifest improves the whole diff-to-test workflow

  • impact analysis accuracy
  • release-diff planning
  • run-set precision
  • gap detection quality
AI grounding

The same manifest keeps generation and QA scoped to the right areas

  • AI generation grounding
  • QA-agent scoping
  • readable user flows
  • more useful coverage guidance

When To Update

Update signals

Refine route families whenever the manifest stops matching reality

  • changed files show up as unmapped
  • confidence is repeatedly low
  • generated plans miss obvious flows
  • release diffs feel too broad
  • new product areas or test directories appear
Takeaway

A strong manifest is one of the biggest predictors of whether `impact-gate` feels useful on real diffs.

Fallback truth

Heuristic grouping is useful for orientation, but a maintained manifest is still the trusted path for release-ready decisions.

Impact Analysis Zero Config CLI Reference