feat: integrate impact analysis specification and initial types setup

.env.example

@@ -37,6 +37,10 @@ DEFAULT_REVIEW_MODE=fast     # fast | rlm
 # OpenAI-compatible endpoint (optional — for Azure OpenAI, local models, etc.)
 OPENAI_BASE_URL=             # e.g. "https://your-resource.openai.azure.com/openai/deployments/your-model"
 
+# Impact Analysis
+IMPACT_ENABLED=true          # Enable dependency impact analysis (default: true)
+IMPACT_DEPTH_THRESHOLD=10    # Max depth for transitive dependency tracing
+
 # GitHub (CLI mode)
 GITHUB_PAT=                  # Personal Access Token for CLI use
 

.pnpm-approve-builds

@@ -1 +1,7 @@
-["unrs-resolver@*"]
+[
+  "unrs-resolver@*",
+  "tree-sitter@*",
+  "tree-sitter-javascript@*",
+  "tree-sitter-python@*",
+  "tree-sitter-typescript@*"
+]

GETTING_STARTED.md

@@ -247,6 +247,8 @@ GITHUB_PAT=ghp_your-token-here
 | `EXCLUDE_GLOBS` | — | File patterns to exclude (e.g., `dist/**`) |
 | `OPENAI_BASE_URL` | — | Custom OpenAI-compatible endpoint |
 | `DEFAULT_REVIEW_MODE` | `fast` | Default: `fast` or `rlm` |
+| `IMPACT_ENABLED` | `true` | Enable dependency impact analysis |
+| `IMPACT_DEPTH_THRESHOLD` | `10` | Max depth for transitive dependency tracing |
 
 See [`.env.example`](.env.example) for the complete list with linter toggles and all options.
 
@@ -301,6 +303,9 @@ node cli/dist/main.mjs review --url https://github.com/owner/repo/pull/123 --out
 
 # Post findings directly as GitHub PR comments
 node cli/dist/main.mjs review --url https://github.com/owner/repo/pull/123 --submit
+
+# Run with dependency impact analysis explicitly disabled
+node cli/dist/main.mjs review --url https://github.com/owner/repo/pull/123 --no-impact
 ```
 
 ### Step 6: Explore other commands
@@ -334,6 +339,7 @@ PR Diff → File Filtering → Diff Chunking → LLM Review (per chunk) → Cita
 - Large diffs are chunked by file (~40K chars per LLM call) for reliable results
 - Lock files, generated code, and binaries are skipped automatically
 - Built-in linters (ESLint, Ruff, Semgrep, ShellCheck, Gitleaks) run in parallel
+- **Impact Analysis** computes the blast radius of changes across the codebase to prioritize findings that break critical dependents
 
 ### Deep / RLM Mode
 

README.md

@@ -15,6 +15,7 @@ Open-source, agentic code review tool. AI-powered bug detection, sandboxed code
 - **Codebase-aware chat** — ask questions about your PR with full repo context via `@openreview`.
 - **Learns from feedback** — persistent learnings database avoids repeating false positives.
 - **Language & framework aware** — detects TypeScript, Python, Go, Rust, React, Next.js, and more for targeted review.
+- **Impact Analysis** — computes the blast radius of changes, scoring transitive dependents to prioritize high-impact findings.
 
 ## Quick Start
 
@@ -100,6 +101,7 @@ npx openreview review --url <PR-URL> [options]
 | `--expert`       | off      | Comprehensive SOLID, security, and quality review  |
 | `--submit`       | off      | Post findings as GitHub PR comment (inline + summary) |
 | `--quiet`        | off      | Suppress progress output (only print findings)     |
+| `--impact`       | auto     | Enable or disable (`--no-impact`) dependency impact analysis |
 
 **Expert mode** (`--expert`) adds deep analysis covering:
 - SOLID principles (single responsibility, open/closed, Liskov, interface segregation, dependency inversion)
@@ -242,8 +244,8 @@ See [GETTING_STARTED.md](GETTING_STARTED.md) for the full setup guide, [CONTRIBU
 
 ## Roadmap
 
-- **Phase 1 (MVP)** ✅ — CLI + GitHub Action, Fast + RLM review, codebase chat, learnings, trace logging
-- **Phase 2 (Growth)** — Web UI, auto-fix, Jira/Linear integration, 30+ linters, Docker sandbox, Impact Analysis
+- **Phase 1 (MVP)** ✅ — CLI + GitHub Action, Fast + RLM review, codebase chat, learnings, trace logging, Impact Analysis
+- **Phase 2 (Growth)** — Web UI, auto-fix, Jira/Linear integration, 30+ linters, Docker sandbox
 - **Phase 3 (Enterprise)** — Multi-platform (GitLab, Azure DevOps, Bitbucket), IDE extension, cloud hosting, analytics
 
 ## Contributing

cli/src/commands/review.ts

@@ -1,3 +1,4 @@
+import { createInterface } from 'node:readline/promises';
 import {
   CommentPoster,
   GitHubClient,
@@ -24,6 +25,8 @@
     .option('--expert', 'Comprehensive SOLID/security/quality review')
     .option('--submit', 'Post findings as GitHub PR comment')
     .option('--quiet', 'Suppress progress output')
+    .option('--impact', 'Include impact analysis')
+    .option('--no-impact', 'Skip impact analysis')
     .action(async (opts) => {
       try {
         const cfg = loadConfig();
@@ -33,6 +36,17 @@
           (cfg as unknown as Record<string, unknown>).mainModel = opts.model;
         }
 
+        if (opts.impact === true) {
+          cfg.impactEnabled = true;
+        } else if (opts.impact === false) {
+          cfg.impactEnabled = false;
+        } else if (!process.env.CI && process.stdout.isTTY) {
+          const rl = createInterface({ input: process.stdin, output: process.stdout });
+          const answer = await rl.question('Would you like to include impact analysis? (y/n) ');
+          rl.close();
+          cfg.impactEnabled = answer.trim().toLowerCase() === 'y';
+        }
+
         const { client, prNumber } = GitHubClient.fromPRUrl(opts.url);
 
         if (!opts.quiet) {
@@ -73,6 +87,8 @@
         let findings: ReviewFinding[];
         let summary: ReviewSummary;
 
+        let impactResult: any;
+
         if (opts.mode === 'rlm') {
           if (!opts.quiet) {
             process.stderr.write('Running RLM deep review...\n');
@@ -109,9 +125,10 @@
             process.stderr.write('Running fast review...\n');
           }
 
-          const result = await runFastReview(prContext);
+          const result = await runFastReview(prContext, process.cwd());
           findings = result.findings;
           summary = result.summary;
+          impactResult = result.impact;
         }
 
         // Format and print output

cli/src/formatter.ts

@@ -16,6 +16,9 @@ export function formatText(findings: ReviewFinding[], summary: ReviewSummary): s
 
   lines.push(`OpenReview — ${summary.mode} mode | ${summary.filesReviewed} files | ${summary.duration}`);
   lines.push(`Findings: ${summary.totalFindings}`);
+  if (summary.impactSummary) {
+    lines.push(`Impact Summary: ${summary.impactSummary.totalImpacted} total files impacted, ${summary.impactSummary.affectedPageCount} UI pages/routes affected`);
+  }
   lines.push('');
 
   if (findings.length === 0) {
@@ -26,7 +29,8 @@ export function formatText(findings: ReviewFinding[], summary: ReviewSummary): s
   for (const f of findings) {
     const icon = SEVERITY_ICONS[f.severity] ?? '❓';
     const source = f.source === 'linter' ? ` [${f.linterName}]` : '';
-    lines.push(`${icon} ${f.severity.toUpperCase()} ${f.file}:${f.startLine} — ${f.title}${source}`);
+    const impact = f.impactScope ? ` [Impact: ${f.impactScope.affectedPages} pages, ${f.impactScope.affectedFiles} files]` : '';
+    lines.push(`${icon} ${f.severity.toUpperCase()} ${f.file}:${f.startLine} — ${f.title}${source}${impact}`);
     lines.push(`  ${f.explanation}`);
     if (f.suggestedFix) {
       lines.push(`  Fix: ${f.suggestedFix}`);
@@ -47,6 +51,9 @@ export function formatMarkdown(findings: ReviewFinding[], summary: ReviewSummary
   lines.push(`## OpenReview Summary`);
   lines.push('');
   lines.push(`**Files reviewed:** ${summary.filesReviewed} | **Duration:** ${summary.duration} | **Mode:** ${summary.mode}`);
+  if (summary.impactSummary) {
+    lines.push(`**Impact:** ${summary.impactSummary.totalImpacted} total files impacted, ${summary.impactSummary.affectedPageCount} UI pages/routes affected`);
+  }
   lines.push('');
   lines.push('| Severity | Count |');
   lines.push('|---|---|');
@@ -74,7 +81,8 @@ export function formatMarkdown(findings: ReviewFinding[], summary: ReviewSummary
     lines.push(`### ${icon} ${severity}`);
     lines.push('');
     for (const f of group) {
-      lines.push(`**${f.title}** — \`${f.file}:${f.startLine}\``);
+      const impact = f.impactScope ? ` _[Impact: ${f.impactScope.affectedPages} pages, ${f.impactScope.affectedFiles} files]_` : '';
+      lines.push(`**${f.title}** — \`${f.file}:${f.startLine}\`${impact}`);
       lines.push('');
       lines.push(f.explanation);
       if (f.suggestedFix) {

cli/src/impact-formatter.ts

@@ -0,0 +1,76 @@
+import type { ImpactResult, ImpactNode } from '@openreview/core';
+
+export function formatImpactTree(result: ImpactResult): string {
+  if (!result || result.impactedFiles.length === 0) {
+    return 'No impact analysis data available or no files impacted.';
+  }
+
+  const lines: string[] = [];
+  lines.push('================================================================================');
+  lines.push('🌳 IMPACT ANALYSIS TREE');
+  lines.push('================================================================================');
+  lines.push('');
+
+  // Group by proximity
+  const byProximity = new Map<number, ImpactNode[]>();
+  for (const node of result.impactedFiles) {
+    const group = byProximity.get(node.proximity) || [];
+    group.push(node);
+    byProximity.set(node.proximity, group);
+  }
+
+  const sortedProximities = Array.from(byProximity.keys()).sort((a, b) => a - b);
+
+  for (const proximity of sortedProximities) {
+    const nodes = byProximity.get(proximity) || [];
+
+    // Determine section title
+    let sectionTitle = '';
+    if (proximity === 1) sectionTitle = 'Direct Dependents (Proximity 1)';
+    else if (proximity === 2) sectionTitle = '2nd Degree Dependents (Proximity 2)';
+    else if (proximity === 3) sectionTitle = '3rd Degree Dependents (Proximity 3)';
+    else sectionTitle = `Deeper Dependents (Proximity ${proximity})`;
+
+    lines.push(`## ${sectionTitle}`);
+
+    for (const node of nodes) {
+      // Score and File
+      const scoreStr = (node.relevanceScore * 100).toFixed(0);
+      lines.push(`- 📄 ${node.file} (Score: ${scoreStr})`);
+
+      // Import Chain Path
+      if (node.importChain && node.importChain.length > 1) {
+        // Show path starting from the changed file
+        const pathStr = node.importChain.join(' ➔ ');
+        lines.push(`    Chain: ${pathStr}`);
+      }
+    }
+    lines.push('');
+  }
+
+  // Component-to-Page Mapping Highlight
+  if (result.affectedPages.length > 0 || result.affectedComponents.length > 0) {
+    lines.push('--------------------------------------------------------------------------------');
+    lines.push('🎯 AFFECTED UI PAGES & ROUTES');
+    lines.push('--------------------------------------------------------------------------------');
+    lines.push('');
+
+    if (result.affectedPages.length > 0) {
+      lines.push('Affected Pages/Routes:');
+      for (const page of result.affectedPages) {
+        lines.push(`  - 🌐 ${page}`);
+      }
+      lines.push('');
+    }
+
+    if (result.affectedComponents.length > 0) {
+      lines.push('Modified Components triggering these updates:');
+      for (const comp of result.affectedComponents) {
+        lines.push(`  - 🧩 ${comp}`);
+      }
+      lines.push('');
+    }
+  }
+
+  return lines.join('\n');
+}

cli/src/impact-report.ts

@@ -0,0 +1,38 @@
+import fs from 'node:fs/promises';
+import path from 'node:path';
+import type { ImpactResult } from '@openreview/core';
+
+/**
+ * Writes the full ImpactResult to a JSON file.
+ * 
+ * @param result The impact analysis result.
+ * @param outputPath The directory or specific file path to write to. If it's a directory, writes to `impact-report.json`.
+ */
+export async function writeImpactReport(result: ImpactResult, outputPath: string): Promise<void> {
+  let targetPath = outputPath;
+
+  try {
+    const stats = await fs.stat(outputPath);
+    if (stats.isDirectory()) {
+      targetPath = path.join(outputPath, 'impact-report.json');
+    }
+  } catch (error: any) {
+    // If the path doesn't exist, we check if it looks like a file extension
+    if (error.code === 'ENOENT') {
+      if (!path.extname(outputPath)) {
+        // Assume it's a directory that needs to be created
+        await fs.mkdir(outputPath, { recursive: true });
+        targetPath = path.join(outputPath, 'impact-report.json');
+      } else {
+        // Ensure parent directory exists
+        const parentDir = path.dirname(outputPath);
+        await fs.mkdir(parentDir, { recursive: true });
+      }
+    } else {
+      throw error;
+    }
+  }
+
+  const outputData = JSON.stringify(result, null, 2);
+  await fs.writeFile(targetPath, outputData, 'utf-8');
+}

cli/tests/formatter.test.ts

@@ -0,0 +1,59 @@
+import { describe, it, expect } from 'vitest';
+import { formatText, formatMarkdown } from '../src/formatter.js';
+import type { ReviewFinding, ReviewSummary } from '@openreview/core';
+
+describe('CLI Formatter', () => {
+  const mockSummary: ReviewSummary = {
+    filesReviewed: 5,
+    duration: '2s',
+    mode: 'fast',
+    findingsBySeverity: { severe: 1, 'non-severe': 0, investigate: 0, informational: 0 },
+    totalFindings: 1,
+    impactSummary: {
+      totalImpacted: 10,
+      affectedPageCount: 3
+    }
+  };
+
+  const mockFinding: ReviewFinding = {
+    id: '1',
+    category: 'bug',
+    severity: 'severe',
+    file: 'src/components/Button.tsx',
+    startLine: 10,
+    endLine: 10,
+    title: 'Null pointer exception',
+    explanation: 'Could be null',
+    source: 'ai',
+    citations: [],
+    impactScope: {
+      affectedFiles: 5,
+      affectedPages: 2
+    }
+  };
+
+  describe('formatText', () => {
+    it('should format finding with impact scope', () => {
+      const output = formatText([mockFinding], mockSummary);
+      expect(output).toContain('[Impact: 2 pages, 5 files]');
+      expect(output).toContain('🔴 SEVERE src/components/Button.tsx:10 — Null pointer exception');
+    });
+
+    it('should include impact summary', () => {
+      const output = formatText([mockFinding], mockSummary);
+      expect(output).toContain('Impact Summary: 10 total files impacted, 3 UI pages/routes affected');
+    });
+  });
+
+  describe('formatMarkdown', () => {
+    it('should format finding with impact scope', () => {
+      const output = formatMarkdown([mockFinding], mockSummary);
+      expect(output).toContain('_[Impact: 2 pages, 5 files]_');
+    });
+
+    it('should include impact summary', () => {
+      const output = formatMarkdown([mockFinding], mockSummary);
+      expect(output).toContain('**Impact:** 10 total files impacted, 3 UI pages/routes affected');
+    });
+  });
+});