Code Review with Impact Analysis

This example shows how an AI code reviewer can use kc-graph to understand the impact of code changes.

Scenario

A developer changes the validate function in utils.ts. The AI needs to know what else might break.

Implementation

import {
  CodeGraph,
  indexSourceFile,
  analyzeImpact,
  formatImpactSummary,
  getContextForSymbol,
} from 'kc-graph';
import { readFileSync } from 'fs';

// Build the graph from source files
const graph = new CodeGraph();
const files = ['src/utils.ts', 'src/auth.ts', 'src/api.ts', 'src/router.ts'];

for (const file of files) {
  indexSourceFile(graph, file, readFileSync(file, 'utf-8'));
}

// Developer changed 'validate' — analyze impact
const impact = analyzeImpact(graph, graph.resolve('validate')!.id, {
  maxDepth: 5,
  direction: 'dependents',
});

console.log(formatImpactSummary(impact));
// Output:
// Impact analysis for: validate (function)
// Total impacted: 3 nodes across 3 files
//   src/auth.ts:
//     login (function) - distance: 1, via: calls
//   src/api.ts:
//     handleRequest (function) - distance: 2, via: calls → calls
//   src/router.ts:
//     route (function) - distance: 3, via: calls → calls → calls

// Get full context for the changed function + its callers
const context = getContextForSymbol(graph, 'validate', {
  maxTokens: 3000,
  includeSignatures: true,
  includeDoc: true,
  depth: 2,
});

// Send to AI for review
console.log(context!.context);
console.log(`Context uses ~${context!.estimatedTokens} tokens`);

AI Review Workflow

1. Parse the changed files into the graph
2. For each changed function, run analyzeImpact
3. Get getContextForSymbol for each impacted area
4. Send the structured context to the AI reviewer
5. AI can see exactly what depends on the change and how