🌊 The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, federation, vector RAG integration, and native Claude Code / Codex / Hermes and many more Integrated
TypeScript agentic-aiagentic-frameworkagentic-workflowagentsai-agentsai-assistantai-skillsautonomous-agentsclaudecodexdsh-pluginharnessmcp-servermulti-agentmulti-agent-systemsnpmskillsswarmswarm-intelligencetypescript
/**
* Diff Classifier for Change Analysis
*/
import { execFileSync, execFile } from 'node:child_process';
import { createHash } from 'node:crypto';
import { promisify } from 'node:util';
export interface DiffClassifierConfig {
maxDiffSize: number;
classifyByImpact: boolean;
detectRefactoring: boolean;
minConfidence: number;
}
export interface DiffHunk {
oldStart: number;
oldLines: number;
newStart: number;
newLines: number;
content: string;
changes: DiffChange[];
}
export interface DiffChange {
type: 'add' | 'remove' | 'context';
lineNumber: number;
content: string;
}
export interface DiffClassification {
primary: 'feature' | 'bugfix' | 'refactor' | 'docs' | 'test' | 'config' | 'style' | 'unknown';
secondary: string[];
confidence: number;
impactLevel: 'low' | 'medium' | 'high' | 'critical';
suggestedReviewers: string[];
testingStrategy: string[];
riskFactors: string[];
}
export interface FileDiff {
path: string;
hunks: DiffHunk[];
additions: number;
deletions: number;
classification: DiffClassification;
}
export interface DiffAnalysis {
files: FileDiff[];
overall: DiffClassification;
stats: {
totalAdditions: number;
totalDeletions: number;
filesChanged: number;
avgConfidence: number;
};
timestamp: number;
}
const DEFAULT_CONFIG: DiffClassifierConfig = {
maxDiffSize: 10000,
classifyByImpact: true,
detectRefactoring: true,
minConfidence: 0.5,
};
const COMMIT_CLASSIFICATION_PATTERNS: Record<string, RegExp[]> = {
feature: [/^feat/, /add.*feature/, /implement/, /new.*functionality/i],
bugfix: [/^fix/, /bug/, /patch/, /resolve.*issue/i, /hotfix/i],
refactor: [/^refactor/, /restructure/, /reorganize/, /cleanup/i, /rename/i],
docs: [/^docs?/, /documentation/, /readme/i, /comment/i, /\.md$/i],
test: [/^test/, /spec/, /\.test\.[jt]sx?$/, /\.spec\.[jt]sx?$/, /__tests__/],
config: [/^config/, /\.config\./, /package\.json/, /tsconfig/, /\.env/],
style: [/^style/, /format/, /lint/, /prettier/, /eslint/],
};
// File paths and commit subjects are different signals. Prefixes such as
// "fix" and words such as "bug" are useful in subjects, but match unrelated
// paths like src/debug/logger.ts and src/dispatcher.ts.
const PATH_CLASSIFICATION_PATTERNS: Record<string, RegExp[]> = {
docs: [/(^|\/)docs?\//i, /(^|\/)readme(?:\.[^/]*)?$/i, /\.mdx?$/i],
test: [/(^|\/)(?:__tests__|tests?|specs?)\//i, /\.(?:test|spec)\.[^/]+$/i],
config: [/(^|\/)(?:configs?\/|config(?:\.[^/]*)?$)/i, /\.config\.[^/]+$/i,
/(^|\/)(?:package|tsconfig)(?:\.[^/]*)?\.json$/i, /(^|\/)\.env(?:\.[^/]*)?$/i],
style: [/\.(?:css|scss|sass|less)$/i],
};
const IMPACT_KEYWORDS: Record<string, number> = {
payment: 3, database: 2, api: 2, core: 2,
util: 1, helper: 1, test: 0, mock: 0, fixture: 0,
};
function pathHasTerm(path: string, term: string): boolean {
const tokens = path.replace(/([a-z])([A-Z])/g, '$1-$2').toLowerCase().split(/[^a-z0-9]+/);
const aliases: Record<string, string[]> = {
auth: ['auth', 'authentication', 'authorization', 'authenticator', 'oauth'],
util: ['util', 'utils', 'utility', 'utilities'],
test: ['test', 'tests'],
mock: ['mock', 'mocks'],
fixture: ['fixture', 'fixtures'],
config: ['config', 'configuration'],
};
return (aliases[term] ?? [term]).some(value => tokens.includes(value));
}
// Match path terms rather than substrings: `encryption` must trigger review,
// while unrelated names such as `author` and `tokenizer` must not.
const SECURITY_PATH_TERMS = new Set([
'security', 'secure', 'auth', 'oauth', 'authenticate', 'authentication',
'authorization', 'authorized', 'authenticator', 'crypto', 'cryptography',
'encrypt', 'encryption', 'encrypted', 'decrypt', 'decryption', 'decrypted',
'cipher', 'secret', 'secrets', 'password', 'passwords', 'token', 'tokens',
'vault', 'kms',
]);
function isSecuritySensitivePath(path: string): boolean {
const terms = path.replace(/([a-z])([A-Z])/g, '$1-$2').toLowerCase().split(/[^a-z0-9]+/);
return terms.some(term => SECURITY_PATH_TERMS.has(term));
}
export class DiffClassifier {
private config: DiffClassifierConfig;
private ruvectorEngine: unknown = null;
private useNative = false;
private classificationCache: Map<string, DiffClassification> = new Map();
constructor(config: Partial<DiffClassifierConfig> = {}) {
this.config = { ...DEFAULT_CONFIG, ...config };
}
async initialize(): Promise<void> {
try {
// @ruvector/diff is optional - gracefully fallback if not installed
const ruvector = await import('@ruvector/diff' as string).catch(() => null);
if (ruvector) {
this.ruvectorEngine = (ruvector as any).createDiffClassifier?.(this.config);
this.useNative = !!this.ruvectorEngine;
}
} catch { this.useNative = false; }
}
parseDiff(diffContent: string): FileDiff[] {
const files: FileDiff[] = [];
const fileBlocks = diffContent.split(/^diff --git/m).filter(Boolean);
for (const block of fileBlocks) {
const headerEnd = block.indexOf('\n');
const header = block.slice(0, headerEnd < 0 ? undefined : headerEnd);
const destination = header.lastIndexOf(' b/');
if (!header.startsWith(' a/') || destination < 0) continue;
const path = header.slice(destination + 3);
const hunks = this.parseHunks(block);
const additions = hunks.reduce((sum, h) => sum + h.changes.filter(c => c.type === 'add').length, 0);
const deletions = hunks.reduce((sum, h) => sum + h.changes.filter(c => c.type === 'remove').length, 0);
const classification = this.classifyFile(path, hunks);
files.push({ path, hunks, additions, deletions, classification });
}
return files;
}
classify(files: FileDiff[]): DiffAnalysis {
const overall = this.computeOverallClassification(files);
const stats = {
totalAdditions: files.reduce((sum, f) => sum + f.additions, 0),
totalDeletions: files.reduce((sum, f) => sum + f.deletions, 0),
filesChanged: files.length,
avgConfidence: files.length > 0 ? files.reduce((sum, f) => sum + f.classification.confidence, 0) / files.length : 0,
};
return { files, overall, stats, timestamp: Date.now() };
}
classifyCommitMessage(message: string): DiffClassification['primary'] {
const lowerMessage = message.toLowerCase();
const conventional = lowerMessage.match(/^([a-z]+)(?:\([^)]*\))?!?:/);
const conventionalTypes: Record<string, DiffClassification['primary']> = {
feat: 'feature', feature: 'feature', fix: 'bugfix', refactor: 'refactor',
docs: 'docs', doc: 'docs', test: 'test', config: 'config', style: 'style',
};
if (conventional && conventionalTypes[conventional[1]]) {
return conventionalTypes[conventional[1]];
}
for (const [type, patterns] of Object.entries(COMMIT_CLASSIFICATION_PATTERNS)) {
for (const pattern of patterns) {
if (pattern.test(lowerMessage)) return type as DiffClassification['primary'];
}
}
return 'unknown';
}
getStats(): Record<string, number | boolean> {
return { useNative: this.useNative, cacheSize: this.classificationCache.size };
}
clearCache(): void { this.classificationCache.clear(); }
private parseHunks(block: string): DiffHunk[] {
const hunks: DiffHunk[] = [];
const readRange = (value: string, prefix: string): [number, number] | null => {
if (!value.startsWith(prefix)) return null;
const parts = value.slice(1).split(',');
if (parts.length > 2 || parts.some(part => !/^\d+$/.test(part))) return null;
const start = Number(parts[0]);
const count = parts.length === 2 ? Number(parts[1]) : 1;
return Number.isSafeInteger(start) && Number.isSafeInteger(count) ? [start, count] : null;
};
let current: { oldStart: number; oldLines: number; newStart: number; newLines: number; lines: string[] } | null = null;
const finish = () => {
if (!current) return;
const content = current.lines.join('\n');
hunks.push({
oldStart: current.oldStart, oldLines: current.oldLines,
newStart: current.newStart, newLines: current.newLines,
content, changes: this.parseChanges(content, current.newStart),
});
};
for (const line of block.split('\n')) {
if (line.startsWith('@@ -')) {
const end = line.indexOf(' @@', 3);
const coordinates = end < 0 ? [] : line.slice(3, end).split(' ');
const oldRange = coordinates.length === 2 ? readRange(coordinates[0], '-') : null;
const newRange = coordinates.length === 2 ? readRange(coordinates[1], '+') : null;
if (oldRange && newRange) {
finish();
current = { oldStart: oldRange[0], oldLines: oldRange[1], newStart: newRange[0], newLines: newRange[1], lines: [] };
continue;
}
}
if (current) current.lines.push(line);
}
finish();
return hunks;
}
private parseChanges(content: string, startLine: number): DiffChange[] {
const changes: DiffChange[] = [];
const lines = content.split('\n');
let lineNumber = startLine;
for (const line of lines) {
if (line.startsWith('+')) { changes.push({ type: 'add', lineNumber, content: line.substring(1) }); lineNumber++; }
else if (line.startsWith('-')) { changes.push({ type: 'remove', lineNumber: -1, content: line.substring(1) }); }
else if (line.startsWith(' ') || line === '') { changes.push({ type: 'context', lineNumber, content: line.substring(1) || '' }); lineNumber++; }
}
return changes;
}
private classifyFile(path: string, hunks: DiffHunk[]): DiffClassification {
const cacheKey = this.getCacheKey(path, hunks);
const cached = this.classificationCache.get(cacheKey);
if (cached) return cached;
const primary = this.determinePrimaryClassification(path, hunks);
const secondary = this.determineSecondaryClassifications(path, hunks, primary);
const confidence = this.calculateConfidence(path, hunks, primary);
const impactLevel = this.determineImpactLevel(path, hunks);
const suggestedReviewers = this.suggestReviewers(path, primary, impactLevel);
const testingStrategy = this.determineTestingStrategy(path, primary, impactLevel);
const riskFactors = this.identifyRiskFactors(path, hunks, impactLevel);
const classification: DiffClassification = { primary, secondary, confidence, impactLevel, suggestedReviewers, testingStrategy, riskFactors };
this.classificationCache.set(cacheKey, classification);
return classification;
}
private getCacheKey(path: string, hunks: DiffHunk[]): string {
const digest = createHash('sha256');
digest.update(path);
for (const hunk of hunks) {
digest.update(`${hunk.oldStart}:${hunk.newStart}:`);
for (const change of hunk.changes) digest.update(`${change.type}:${change.content}\n`);
}
return digest.digest('hex');
}
private determinePrimaryClassification(path: string, hunks: DiffHunk[]): DiffClassification['primary'] {
for (const [type, patterns] of Object.entries(PATH_CLASSIFICATION_PATTERNS)) {
for (const pattern of patterns) { if (pattern.test(path)) return type as DiffClassification['primary']; }
}
const allContent = hunks.flatMap(h => h.changes.filter(c => c.type !== 'context').map(c => c.content)).join('\n').toLowerCase();
if (/function|class|interface|type\s+\w+/.test(allContent) && hunks.some(h => h.changes.filter(c => c.type === 'add').length > 10)) return 'feature';
if (/\b(?:fix(?:ed|es)?|bug|issue|exception)\b/.test(allContent)) return 'bugfix';
if (this.config.detectRefactoring && this.isRefactoring(hunks)) return 'refactor';
return 'unknown';
}
private isRefactoring(hunks: DiffHunk[]): boolean {
let totalAdds = 0, totalRemoves = 0;
for (const hunk of hunks) { for (const change of hunk.changes) { if (change.type === 'add') totalAdds++; else if (change.type === 'remove') totalRemoves++; } }
const ratio = totalAdds > 0 ? totalRemoves / totalAdds : 0;
return ratio > 0.7 && ratio < 1.4 && totalAdds > 5;
}
private determineSecondaryClassifications(path: string, hunks: DiffHunk[], primary: DiffClassification['primary']): string[] {
const secondary: string[] = [];
for (const [type, patterns] of Object.entries(PATH_CLASSIFICATION_PATTERNS)) {
if (type === primary) continue;
for (const pattern of patterns) { if (pattern.test(path)) { secondary.push(type); break; } }
}
return secondary.slice(0, 3);
}
private calculateConfidence(path: string, hunks: DiffHunk[], primary: DiffClassification['primary']): number {
let confidence = 0.5;
for (const patterns of Object.values(PATH_CLASSIFICATION_PATTERNS)) { for (const pattern of patterns) { if (pattern.test(path)) { confidence += 0.2; break; } } }
const totalChanges = hunks.reduce((sum, h) => sum + h.changes.length, 0);
if (totalChanges > 10) confidence += 0.1;
if (totalChanges > 50) confidence += 0.1;
if (primary !== 'unknown') confidence += 0.1;
return Math.min(1, confidence);
}
private determineImpactLevel(path: string, hunks: DiffHunk[]): DiffClassification['impactLevel'] {
let score = 0;
for (const [keyword, weight] of Object.entries(IMPACT_KEYWORDS)) { if (pathHasTerm(path, keyword)) score = Math.max(score, weight); }
if (isSecuritySensitivePath(path)) score = Math.max(score, 3);
const totalChanges = hunks.reduce((sum, h) => sum + h.changes.filter(c => c.type !== 'context').length, 0);
if (totalChanges > 100) score = Math.max(score, 2);
if (totalChanges > 300) score = Math.max(score, 3);
if (score >= 3) return 'critical';
if (score >= 2) return 'high';
if (score >= 1) return 'medium';
return 'low';
}
private suggestReviewers(path: string, primary: DiffClassification['primary'], impact: DiffClassification['impactLevel']): string[] {
const reviewers: string[] = [];
const typeReviewers: Record<string, string[]> = { feature: ['tech-lead', 'product-owner'], bugfix: ['qa-engineer', 'developer'], refactor: ['senior-developer', 'architect'], docs: ['tech-writer', 'developer'], test: ['qa-engineer', 'developer'], config: ['devops', 'tech-lead'], style: ['developer'], unknown: ['developer'] };
reviewers.push(...(typeReviewers[primary] || typeReviewers.unknown));
if (impact === 'critical' || impact === 'high') reviewers.push('security-reviewer');
if (isSecuritySensitivePath(path)) reviewers.push('security-team');
if (pathHasTerm(path, 'database') || pathHasTerm(path, 'migration')) reviewers.push('dba');
return [...new Set(reviewers)].slice(0, 4);
}
private determineTestingStrategy(path: string, primary: DiffClassification['primary'], impact: DiffClassification['impactLevel']): string[] {
const strategies: string[] = [];
if (primary !== 'test') strategies.push('unit-tests');
if (primary === 'feature') strategies.push('integration-tests');
if (impact === 'high' || impact === 'critical') { strategies.push('regression-tests'); strategies.push('e2e-tests'); }
if (['api', 'endpoint', 'route', 'handler'].some(term => pathHasTerm(path, term))) strategies.push('api-contract-tests');
if (isSecuritySensitivePath(path)) strategies.push('security-audit');
return strategies.slice(0, 5);
}
private identifyRiskFactors(path: string, hunks: DiffHunk[], impact: DiffClassification['impactLevel']): string[] {
const risks: string[] = [];
const totalChanges = hunks.reduce((sum, h) => sum + h.changes.length, 0);
if (totalChanges > 200) risks.push('Large change set - increased review time needed');
if (impact === 'critical') risks.push('Critical system component - requires careful review');
if (impact === 'high') risks.push('High-impact area - monitor after deployment');
if (isSecuritySensitivePath(path)) risks.push('Security-sensitive code');
if (pathHasTerm(path, 'database') || pathHasTerm(path, 'migration')) risks.push('Database changes - ensure backup strategy');
if (pathHasTerm(path, 'config') || pathHasTerm(path, 'env')) risks.push('Configuration changes - verify all environments');
const allContent = hunks.flatMap(h => h.changes.map(c => c.content)).join('\n');
if (/TODO|FIXME|HACK/.test(allContent)) risks.push('Contains TODO/FIXME comments');
if (/password|secret|key|token/i.test(allContent)) risks.push('Potential secrets in code');
return risks.slice(0, 5);
}
private computeOverallClassification(files: FileDiff[]): DiffClassification {
if (files.length === 0) return { primary: 'unknown', secondary: [], confidence: 0, impactLevel: 'low', suggestedReviewers: [], testingStrategy: [], riskFactors: [] };
const primaryCounts: Record<string, number> = {};
for (const file of files) { const p = file.classification.primary; primaryCounts[p] = (primaryCounts[p] || 0) + 1; }
let primary: DiffClassification['primary'] = 'unknown';
let maxCount = 0;
for (const [type, count] of Object.entries(primaryCounts)) { if (count > maxCount) { maxCount = count; primary = type as DiffClassification['primary']; } }
const secondaryCounts: Record<string, number> = {};
for (const file of files) { for (const s of file.classification.secondary) { secondaryCounts[s] = (secondaryCounts[s] || 0) + 1; } }
const secondary = Object.entries(secondaryCounts).sort((a, b) => b[1] - a[1]).slice(0, 3).map(([type]) => type);
const confidence = files.reduce((sum, f) => sum + f.classification.confidence, 0) / files.length;
const impactOrder: DiffClassification['impactLevel'][] = ['low', 'medium', 'high', 'critical'];
let impactLevel: DiffClassification['impactLevel'] = 'low';
for (const file of files) { if (impactOrder.indexOf(file.classification.impactLevel) > impactOrder.indexOf(impactLevel)) impactLevel = file.classification.impactLevel; }
const reviewers = [...new Set(files.flatMap(f => f.classification.suggestedReviewers))].slice(0, 5);
const testingStrategy = [...new Set(files.flatMap(f => f.classification.testingStrategy))].slice(0, 5);
const riskFactors = [...new Set(files.flatMap(f => f.classification.riskFactors))].slice(0, 5);
return { primary, secondary, confidence, impactLevel, suggestedReviewers: reviewers, testingStrategy, riskFactors };
}
}
export function createDiffClassifier(config?: Partial<DiffClassifierConfig>): DiffClassifier {
return new DiffClassifier(config);
}
// ============================================================================
// Additional Exports for MCP Tools
// ============================================================================
/**
* Risk level type for file risk assessment
*/
export type RiskLevel = 'low' | 'medium' | 'high' | 'critical';
/**
* Diff file interface for analyze tools
*/
export interface DiffFile {
path: string;
status: 'added' | 'modified' | 'deleted' | 'renamed';
additions: number;
deletions: number;
hunks: number;
binary: boolean;
}
/**
* File risk assessment result
*/
export interface FileRisk {
file: string;
risk: RiskLevel;
score: number;
reasons: string[];
}
/**
* Overall risk assessment result
*/
export interface OverallRisk {
overall: RiskLevel;
score: number;
breakdown: { low: number; medium: number; high: number; critical: number };
}
/**
* Diff analysis result
*/
export interface DiffAnalysisResult {
ref: string;
timestamp: number;
files: DiffFile[];
risk: OverallRisk;
classification: DiffClassification;
summary: string;
fileRisks?: FileRisk[];
recommendedReviewers?: string[];
}
// ============================================================================
// Optimized Git Diff Functions
// ============================================================================
// Cache for diff results (TTL-based)
const diffCache = new Map<string, { files: DiffFile[]; timestamp: number }>();
const CACHE_TTL_MS = 5000; // 5 seconds - short TTL since diffs change frequently
/**
* Validate git ref to prevent command injection
* Only allows safe characters: alphanumeric, -, _, /, ., ~, ^
*/
function validateGitRef(ref: string): void {
if (ref.length > 256) {
throw new Error(`Invalid git ref: too long`);
}
// A ref is passed as one argv element, but still must not be an option or
// contain shell metacharacters. A range may omit its right endpoint (HEAD).
if (!/^[a-zA-Z0-9_][a-zA-Z0-9_\-./~^@]*$/.test(ref)) {
throw new Error(`Invalid git ref: contains unsafe characters`);
}
if (ref.includes('..')) {
const range = ref.match(/^(.+?)(\.\.\.?)(.*)$/);
// Only one range separator is allowed. In particular, reject traversal
// inside either endpoint, while accepting branch paths and ancestry syntax.
if (!range || range[1].includes('..') || range[3].includes('..') ||
!/^[a-zA-Z0-9_][a-zA-Z0-9_\-./~^@]*$/.test(range[1]) ||
(range[3] !== '' && !/^[a-zA-Z0-9_][a-zA-Z0-9_\-./~^@]*$/.test(range[3]))) {
throw new Error(`Invalid git ref: suspicious pattern`);
}
}
}
/**
* Get git diff statistics using SINGLE combined command (optimized)
* Replaces two separate git commands with one
*/
export function getGitDiffNumstat(ref: string = 'HEAD'): DiffFile[] {
// SECURITY: Validate git ref to prevent command injection
validateGitRef(ref);
// Check cache first
const cacheKey = `numstat:${ref}`;
const cached = diffCache.get(cacheKey);
if (cached && Date.now() - cached.timestamp < CACHE_TTL_MS) {
return cached.files;
}
// execFileSync imported at top level
try {
// SECURITY: Use execFileSync with args array instead of shell string
// This prevents command injection via the ref parameter
const numstatOutput = execFileSync('git', [
'diff', '--numstat', '--diff-filter=ACDMRTUXB', ref
], { encoding: 'utf-8', maxBuffer: 10 * 1024 * 1024 });
const statusOutput = execFileSync('git', [
'diff', '--name-status', ref
], { encoding: 'utf-8', maxBuffer: 10 * 1024 * 1024 });
const output = numstatOutput + '---STATUS---' + statusOutput;
const [numstatPart, statusPart] = output.split('---STATUS---');
// Parse status (usually smaller, parse first)
const statusMap = new Map<string, string>();
if (statusPart) {
for (const line of statusPart.trim().split('\n')) {
if (!line) continue;
const [status, ...pathParts] = line.split('\t');
const path = pathParts[pathParts.length - 1] || pathParts[0];
if (path) statusMap.set(path, status.charAt(0));
}
}
// Parse numstat
const files: DiffFile[] = [];
if (numstatPart) {
for (const line of numstatPart.trim().split('\n')) {
if (!line) continue;
const [addStr, delStr, path] = line.split('\t');
if (!path) continue;
const additions = addStr === '-' ? 0 : parseInt(addStr, 10) || 0;
const deletions = delStr === '-' ? 0 : parseInt(delStr, 10) || 0;
const binary = addStr === '-' && delStr === '-';
const statusChar = statusMap.get(path) || 'M';
let status: DiffFile['status'] = 'modified';
switch (statusChar) {
case 'A': status = 'added'; break;
case 'D': status = 'deleted'; break;
case 'R': status = 'renamed'; break;
default: status = 'modified';
}
files.push({ path, status, additions, deletions, hunks: 1, binary });
}
}
// Cache the result
diffCache.set(cacheKey, { files, timestamp: Date.now() });
return files;
} catch {
return [];
}
}
/**
* Async version of getGitDiffNumstat for non-blocking operation
*/
export async function getGitDiffNumstatAsync(ref: string = 'HEAD'): Promise<DiffFile[]> {
// SECURITY: Validate git ref to prevent command injection
validateGitRef(ref);
// Check cache first
const cacheKey = `numstat:${ref}`;
const cached = diffCache.get(cacheKey);
if (cached && Date.now() - cached.timestamp < CACHE_TTL_MS) {
return cached.files;
}
// execFile + promisify imported at top level
const execFileAsync = promisify(execFile);
try {
// SECURITY: Use execFile with args array instead of shell string
const { stdout: numstatOutput } = await execFileAsync('git', [
'diff', '--numstat', '--diff-filter=ACDMRTUXB', ref
], { maxBuffer: 10 * 1024 * 1024 });
const { stdout: statusOutput } = await execFileAsync('git', [
'diff', '--name-status', ref
], { maxBuffer: 10 * 1024 * 1024 });
const stdout = numstatOutput + '---STATUS---' + statusOutput;
const [numstatPart, statusPart] = stdout.split('---STATUS---');
const statusMap = new Map<string, string>();
if (statusPart) {
for (const line of statusPart.trim().split('\n')) {
if (!line) continue;
const [status, ...pathParts] = line.split('\t');
const path = pathParts[pathParts.length - 1] || pathParts[0];
if (path) statusMap.set(path, status.charAt(0));
}
}
const files: DiffFile[] = [];
if (numstatPart) {
for (const line of numstatPart.trim().split('\n')) {
if (!line) continue;
const [addStr, delStr, path] = line.split('\t');
if (!path) continue;
const additions = addStr === '-' ? 0 : parseInt(addStr, 10) || 0;
const deletions = delStr === '-' ? 0 : parseInt(delStr, 10) || 0;
const binary = addStr === '-' && delStr === '-';
const statusChar = statusMap.get(path) || 'M';
let status: DiffFile['status'] = 'modified';
switch (statusChar) {
case 'A': status = 'added'; break;
case 'D': status = 'deleted'; break;
case 'R': status = 'renamed'; break;
default: status = 'modified';
}
files.push({ path, status, additions, deletions, hunks: 1, binary });
}
}
diffCache.set(cacheKey, { files, timestamp: Date.now() });
return files;
} catch {
return [];
}
}
/**
* Clear the diff cache (call when git state changes)
*/
export function clearDiffCache(): void {
diffCache.clear();
}
/**
* Assess risk for a single file
*/
export function assessFileRisk(file: DiffFile): FileRisk {
const reasons: string[] = [];
let score = 0;
// Size-based risk
const totalChanges = file.additions + file.deletions;
if (totalChanges > 300) {
score += 30;
reasons.push('Large change size (>300 lines)');
} else if (totalChanges > 100) {
score += 15;
reasons.push('Medium change size (>100 lines)');
}
// Path-based risk
if (isSecuritySensitivePath(file.path)) {
score += 40;
reasons.push('Security-sensitive file');
}
if (['payment', 'billing', 'transaction'].some(term => pathHasTerm(file.path, term))) {
score += 35;
reasons.push('Payment-related file');
}
if (['database', 'migration', 'schema'].some(term => pathHasTerm(file.path, term))) {
score += 25;
reasons.push('Database-related file');
}
if (['core', 'main', 'index'].some(term => pathHasTerm(file.path, term))) {
score += 15;
reasons.push('Core module');
}
if (['config', 'env', 'settings'].some(term => pathHasTerm(file.path, term))) {
score += 20;
reasons.push('Configuration file');
}
// Status-based risk
if (file.status === 'deleted') {
score += 10;
reasons.push('File deleted');
}
// Binary file risk
if (file.binary) {
score += 5;
reasons.push('Binary file');
}
let risk: RiskLevel = 'low';
if (score >= 60) risk = 'critical';
else if (score >= 40) risk = 'high';
else if (score >= 20) risk = 'medium';
return { file: file.path, risk, score: Math.min(100, score), reasons };
}
/**
* Assess overall risk from files and file risks
*/
export function assessOverallRisk(files: DiffFile[], fileRisks: FileRisk[]): OverallRisk {
const breakdown = { low: 0, medium: 0, high: 0, critical: 0 };
let totalScore = 0;
for (const fr of fileRisks) {
breakdown[fr.risk]++;
totalScore += fr.score;
}
const avgScore = fileRisks.length > 0 ? totalScore / fileRisks.length : 0;
// Weight more heavily towards high/critical files
const weightedScore = avgScore + (breakdown.critical * 15) + (breakdown.high * 10);
let overall: RiskLevel = 'low';
if (weightedScore >= 60 || breakdown.critical > 0) overall = 'critical';
else if (weightedScore >= 40 || breakdown.high > 1) overall = 'high';
else if (weightedScore >= 20 || breakdown.medium > 2) overall = 'medium';
return { overall, score: Math.min(100, Math.round(weightedScore)), breakdown };
}
// Singleton classifier instance for reuse
let classifierInstance: DiffClassifier | null = null;
function getClassifier(): DiffClassifier {
if (!classifierInstance) {
classifierInstance = new DiffClassifier();
}
return classifierInstance;
}
/** Classify file metadata with the actual patch and, for one-commit ranges, its subject. */
export function classifyDiff(files: DiffFile[], patch = '', commitSubject = ''): DiffClassification {
const classifier = getClassifier();
const parsed = new Map(classifier.parseDiff(patch).map(file => [file.path, file]));
const fileDiffs: FileDiff[] = files.map(f => {
const patchFile = parsed.get(f.path);
const hunks = patchFile?.hunks ?? [];
return {
path: f.path,
hunks,
additions: f.additions,
deletions: f.deletions,
classification: patchFile?.classification ?? classifier['classifyFile'](f.path, hunks),
};
});
const overall = classifier['computeOverallClassification'](fileDiffs);
if (fileDiffs.length > 0 && commitSubject) {
const subjectType = classifier.classifyCommitMessage(commitSubject);
if (subjectType !== 'unknown') {
if (overall.primary !== subjectType && overall.primary !== 'unknown') {
overall.secondary = [overall.primary, ...overall.secondary].filter((type, index, all) =>
type !== subjectType && all.indexOf(type) === index).slice(0, 3);
}
overall.primary = subjectType;
overall.confidence = Math.max(overall.confidence, 0.85);
}
}
return overall;
}
function getSingleCommitSubjectSync(ref: string): string {
if (!ref.includes('..')) return '';
try {
const output = execFileSync('git', ['log', '-2', '--format=%s', ref],
{ encoding: 'utf-8', maxBuffer: 64 * 1024 });
const subjects = output.trim().split('\n').filter(Boolean);
return subjects.length === 1 ? subjects[0] : '';
} catch { return ''; }
}
function getDiffContextSync(ref: string): { patch: string; subject: string } {
validateGitRef(ref);
let patch = '';
try {
patch = execFileSync('git', ['diff', '--no-ext-diff', '--no-textconv', '--unified=0', ref],
{ encoding: 'utf-8', maxBuffer: 10 * 1024 * 1024 });
} catch { /* A large or unavailable patch falls back to path-only classification. */ }
return { patch, subject: getSingleCommitSubjectSync(ref) };
}
async function getDiffContextAsync(ref: string): Promise<{ patch: string; subject: string }> {
validateGitRef(ref);
const execFileAsync = promisify(execFile);
const patchPromise = execFileAsync('git', ['diff', '--no-ext-diff', '--no-textconv', '--unified=0', ref],
{ maxBuffer: 10 * 1024 * 1024 }).then(({ stdout }) => stdout).catch(() => '');
const subjectPromise = ref.includes('..')
? execFileAsync('git', ['log', '-2', '--format=%s', ref], { maxBuffer: 64 * 1024 })
.then(({ stdout }) => {
const subjects = stdout.trim().split('\n').filter(Boolean);
return subjects.length === 1 ? subjects[0] : '';
}).catch(() => '')
: Promise.resolve('');
const [patch, subject] = await Promise.all([patchPromise, subjectPromise]);
return { patch, subject };
}
/**
* Suggest reviewers based on files and risks
*/
export function suggestReviewers(files: DiffFile[], fileRisks: FileRisk[]): string[] {
const reviewers = new Set<string>();
for (const file of files) {
const hasAny = (...terms: string[]) => terms.some(term => pathHasTerm(file.path, term));
if (isSecuritySensitivePath(file.path)) reviewers.add('security-team');
if (hasAny('database', 'migration')) reviewers.add('dba');
if (hasAny('api', 'endpoint', 'route')) reviewers.add('api-owner');
if (hasAny('test', 'spec')) reviewers.add('qa-engineer');
if (hasAny('config', 'deploy', 'ci')) reviewers.add('devops');
if (hasAny('ui', 'component', 'style')) reviewers.add('frontend-lead');
if (hasAny('model', 'service', 'repository')) reviewers.add('backend-lead');
}
// Add based on risk
const hasHighRisk = fileRisks.some(fr => fr.risk === 'high' || fr.risk === 'critical');
if (hasHighRisk) {
reviewers.add('tech-lead');
reviewers.add('senior-developer');
}
// Default reviewer
if (reviewers.size === 0) {
reviewers.add('developer');
}
return Array.from(reviewers).slice(0, 5);
}
// Analysis result cache
const analysisCache = new Map<string, { result: DiffAnalysisResult; timestamp: number }>();
const ANALYSIS_CACHE_TTL_MS = 3000; // 3 seconds
/**
* Analyze a diff with full analysis (optimized with caching)
*/
export async function analyzeDiff(options: {
ref?: string;
useRuVector?: boolean;
skipCache?: boolean;
}): Promise<DiffAnalysisResult> {
const ref = options.ref || 'HEAD';
// Check analysis cache (unless skipCache is true)
if (!options.skipCache) {
const cached = analysisCache.get(ref);
if (cached && Date.now() - cached.timestamp < ANALYSIS_CACHE_TTL_MS) {
return cached.result;
}
}
// Use async git diff for non-blocking operation
const [files, context] = await Promise.all([getGitDiffNumstatAsync(ref), getDiffContextAsync(ref)]);
// Parallel file risk assessment for large diffs
const fileRisks = files.length > 20
? await Promise.all(files.map(f => Promise.resolve(assessFileRisk(f))))
: files.map(assessFileRisk);
const risk = assessOverallRisk(files, fileRisks);
const classification = classifyDiff(files, context.patch, context.subject);
const recommendedReviewers = suggestReviewers(files, fileRisks);
const totalAdditions = files.reduce((sum, f) => sum + f.additions, 0);
const totalDeletions = files.reduce((sum, f) => sum + f.deletions, 0);
const result: DiffAnalysisResult = {
ref,
timestamp: Date.now(),
files,
risk,
classification,
summary: `${files.length} files changed (+${totalAdditions}/-${totalDeletions}), ${risk.overall} risk`,
fileRisks,
recommendedReviewers,
};
// Cache the result
analysisCache.set(ref, { result, timestamp: Date.now() });
return result;
}
/**
* Synchronous version of analyzeDiff for backward compatibility
*/
export function analyzeDiffSync(options: {
ref?: string;
useRuVector?: boolean;
}): DiffAnalysisResult {
const ref = options.ref || 'HEAD';
// Check analysis cache
const cached = analysisCache.get(ref);
if (cached && Date.now() - cached.timestamp < ANALYSIS_CACHE_TTL_MS) {
return cached.result;
}
const files = getGitDiffNumstat(ref);
const context = getDiffContextSync(ref);
const fileRisks = files.map(assessFileRisk);
const risk = assessOverallRisk(files, fileRisks);
const classification = classifyDiff(files, context.patch, context.subject);
const recommendedReviewers = suggestReviewers(files, fileRisks);
const totalAdditions = files.reduce((sum, f) => sum + f.additions, 0);
const totalDeletions = files.reduce((sum, f) => sum + f.deletions, 0);
const result: DiffAnalysisResult = {
ref,
timestamp: Date.now(),
files,
risk,
classification,
summary: `${files.length} files changed (+${totalAdditions}/-${totalDeletions}), ${risk.overall} risk`,
fileRisks,
recommendedReviewers,
};
analysisCache.set(ref, { result, timestamp: Date.now() });
return result;
}
/**
* Clear all diff-related caches
*/
export function clearAllDiffCaches(): void {
diffCache.clear();
analysisCache.clear();
classifierInstance?.clearCache();
}