agentic-qe 3.15.5 → 3.15.7

package/dist/workers/workers/defect-predictor.js

@@ -2,13 +2,24 @@
  * Agentic QE v3 - Defect Predictor Worker
  * ADR-014: Background Workers for QE Monitoring
  *
- * ML-based defect prediction including:
- * - Code change risk analysis
- * - Historical defect pattern matching
- * - Hotspot identification
- * - Predictive maintenance suggestions
+ * Surfaces the defect-intelligence domain's latest predictions:
+ * - High-risk files and their contributing factors
+ * - Targeted review / refactoring / coverage recommendations
+ *
+ * #1043: every prediction comes from a run of the defect-intelligence
+ * predictor (`defect-intelligence:predictor:prediction:<id>` in the
+ * `defect-intelligence` namespace, written by DefectPredictorService), and
+ * only files that exist in the project are kept. Code hotspots have no
+ * measured source and are reported unmeasured. Nothing predicted → nothing
+ * stored, an "unmeasured" info finding. A `defect:predictions` record without
+ * the measured stamp (the invented one written by <= 3.15.6) is ignored and
+ * replaced by the next measured run.
  */
+import * as fs from 'fs';
+import * as path from 'path';
+import { bareSearchKey } from '../namespaced-keys.js';
 import { BaseWorker } from '../base-worker.js';
+import { findProjectRoot } from '../../kernel/project-root.js';
 const CONFIG = {
     id: 'defect-predictor',
     name: 'Defect Predictor',
@@ -21,6 +32,82 @@ const CONFIG = {
     retryCount: 2,
     retryDelayMs: 15000,
 };
+const DEFECT_NAMESPACE = 'defect-intelligence';
+const PREDICTION_KEY_PATTERN = 'defect-intelligence:predictor:prediction:*';
+const PREDICTIONS_KEY = 'defect:predictions';
+const SCHEMA_VERSION = 1;
+/** The predictor keeps its records for 7 days; nothing older is current. */
+const MAX_PREDICTION_AGE_MS = 7 * 24 * 60 * 60 * 1000;
+const MAX_PREDICTION_RECORDS = 1000;
+const RISK_LEVELS = ['critical', 'high', 'medium', 'low', 'info'];
+/**
+ * Where a predicted path lies, resolved against the project root with
+ * symlinks followed: only a regular file whose real path is inside the
+ * project's real root is a project file (#1043 round 2b).
+ */
+function classifyPredictedFile(file, realRoot) {
+    let real;
+    try {
+        real = fs.realpathSync(path.resolve(realRoot, file));
+    }
+    catch {
+        // Unresolvable: missing when it would lie inside the root, else outside.
+        const lexical = path.resolve(realRoot, file);
+        const rel = path.relative(realRoot, lexical);
+        return rel && !rel.startsWith('..') && !path.isAbsolute(rel) ? 'missing' : 'outside';
+    }
+    const rel = path.relative(realRoot, real);
+    if (rel === '' || rel.startsWith('..') || path.isAbsolute(rel))
+        return 'outside';
+    try {
+        return fs.statSync(real).isFile() ? 'inside' : 'missing';
+    }
+    catch {
+        return 'missing';
+    }
+}
+function realProjectRoot() {
+    const root = findProjectRoot();
+    try {
+        return fs.realpathSync(root);
+    }
+    catch {
+        return path.resolve(root);
+    }
+}
+const isRecord = (v) => !!v && typeof v === 'object' && !Array.isArray(v);
+function parseFilePrediction(raw) {
+    if (!isRecord(raw))
+        return null;
+    const { file, probability, riskLevel } = raw;
+    if (typeof file !== 'string' || file.length === 0)
+        return null;
+    if (typeof probability !== 'number' || !Number.isFinite(probability) || probability < 0 || probability > 1)
+        return null;
+    if (!RISK_LEVELS.includes(riskLevel))
+        return null;
+    const factors = Array.isArray(raw.factors)
+        ? raw.factors.filter((f) => isRecord(f) && typeof f.name === 'string' && typeof f.contribution === 'number' && Number.isFinite(f.contribution))
+        : [];
+    const recommendations = Array.isArray(raw.recommendations)
+        ? raw.recommendations.filter((r) => typeof r === 'string')
+        : [];
+    return { file, probability, riskLevel: riskLevel, factors, recommendations };
+}
+/** A recorded prediction run, or `null` when the record is not a usable one. */
+function parsePredictionRecord(raw, now) {
+    if (!isRecord(raw) || typeof raw.id !== 'string' || typeof raw.timestamp !== 'string')
+        return null;
+    const t = Date.parse(raw.timestamp);
+    if (!Number.isFinite(t) || t > now + 60_000 || now - t > MAX_PREDICTION_AGE_MS)
+        return null;
+    if (!Array.isArray(raw.predictions) || raw.predictions.length === 0)
+        return null;
+    const predictions = raw.predictions.map(parseFilePrediction);
+    if (predictions.some((p) => p === null))
+        return null;
+    return { id: raw.id, timestamp: raw.timestamp, predictions: predictions };
+}
 export class DefectPredictorWorker extends BaseWorker {
     constructor() {
         super(CONFIG);
@@ -30,23 +117,46 @@ export class DefectPredictorWorker extends BaseWorker {
         context.logger.info('Starting defect prediction analysis');
         const findings = [];
         const recommendations = [];
-        // Analyze code for defect predictions
-        const predictions = await this.analyzeCodeForDefects(context);
-        // Identify hotspots
-        const hotspots = await this.identifyHotspots(context);
-        // Generate findings from predictions
+        const { measured: latest, unusable } = await this.readLatestPrediction(context);
+        if (!latest) {
+            findings.push({
+                type: 'defect-prediction-unmeasured',
+                severity: 'info',
+                domain: 'defect-intelligence',
+                title: 'Defect Prediction Unmeasured',
+                description: (unusable
+                    ? `The latest defect prediction (${unusable.id}) names no file inside this project ` +
+                        `(${unusable.outside} outside the project root, ${unusable.missing} missing), so no risk is reported. `
+                    : 'No defect prediction for files in this project has been recorded in the last 7 days, so no risk is reported. ') +
+                    'Run defect_predict on the files to assess.',
+            });
+            if (unusable)
+                await this.markPreviousStale(context, unusable.id);
+            context.logger.info('Defect prediction complete', { measured: false });
+            return this.createResult(Date.now() - startTime, {
+                itemsAnalyzed: 0,
+                issuesFound: 0,
+                healthScore: 100,
+                trend: 'stable',
+                domainMetrics: {
+                    filesAnalyzed: 0,
+                    highRiskFiles: 0,
+                    hotspots: 'unmeasured',
+                    avgDefectProbability: 'unmeasured',
+                },
+            }, findings, recommendations);
+        }
+        const { predictions } = latest;
         this.analyzeHighRiskPredictions(predictions, findings, recommendations);
-        // Analyze hotspots
-        this.analyzeHotspots(hotspots, findings, recommendations);
-        // Store predictions
-        await context.memory.set('defect:predictions', predictions);
-        await context.memory.set('defect:hotspots', hotspots);
-        await context.memory.set('defect:lastAnalysis', new Date().toISOString());
-        const healthScore = this.calculateHealthScore(predictions, hotspots);
+        await context.memory.set(PREDICTIONS_KEY, latest);
+        await context.memory.set('defect:lastAnalysis', latest.evaluatedAt);
+        const healthScore = this.calculateHealthScore(predictions);
+        const highRisk = predictions.filter(p => p.riskLevel === 'high' || p.riskLevel === 'critical').length;
         context.logger.info('Defect prediction complete', {
+            measured: true,
             healthScore,
+            predictionId: latest.predictionId,
             predictionsCount: predictions.length,
-            hotspotsCount: hotspots.length,
         });
         return this.createResult(Date.now() - startTime, {
             itemsAnalyzed: predictions.length,
@@ -55,237 +165,134 @@ export class DefectPredictorWorker extends BaseWorker {
             trend: 'stable',
             domainMetrics: {
                 filesAnalyzed: predictions.length,
-                highRiskFiles: predictions.filter(p => p.riskLevel === 'high' || p.riskLevel === 'critical').length,
-                hotspots: hotspots.length,
+                highRiskFiles: highRisk,
+                missingFilesSkipped: latest.missingFilesSkipped,
+                outsideProjectSkipped: latest.outsideProjectSkipped,
+                hotspots: 'unmeasured',
                 avgDefectProbability: (predictions.reduce((sum, p) => sum + p.probability, 0) / predictions.length).toFixed(2),
             },
         }, findings, recommendations);
     }
-    async analyzeCodeForDefects(_context) {
-        // In a real implementation, this would:
-        // 1. Analyze git history for change patterns
-        // 2. Apply ML models for defect prediction
-        // 3. Consider code complexity metrics
-        // 4. Factor in historical defect data
-        return [
-            {
-                file: 'src/coordination/workflow-orchestrator.ts',
-                probability: 0.78,
-                riskLevel: 'high',
-                factors: [
-                    { name: 'complexity', weight: 0.3, description: 'High cyclomatic complexity (28)' },
-                    { name: 'change-frequency', weight: 0.25, description: 'Changed 15 times in last 30 days' },
-                    { name: 'size', weight: 0.2, description: 'Large file (1500+ lines)' },
-                    { name: 'coupling', weight: 0.15, description: 'High coupling with 8 other modules' },
-                ],
-                recentChanges: 15,
-                historicalDefects: 3,
-                complexity: 28,
-            },
-            {
-                file: 'src/domains/security-compliance/services/security-scanner.ts',
-                probability: 0.65,
-                riskLevel: 'high',
-                factors: [
-                    { name: 'complexity', weight: 0.35, description: 'High cyclomatic complexity (32)' },
-                    { name: 'size', weight: 0.25, description: 'Very large file (1200+ lines)' },
-                    { name: 'test-coverage', weight: 0.2, description: 'Low test coverage (45%)' },
-                ],
-                recentChanges: 8,
-                historicalDefects: 2,
-                complexity: 32,
-            },
-            {
-                file: 'src/kernel/memory-backend.ts',
-                probability: 0.45,
-                riskLevel: 'medium',
-                factors: [
-                    { name: 'change-frequency', weight: 0.4, description: 'Changed 10 times in last 30 days' },
-                    { name: 'authors', weight: 0.3, description: 'Multiple authors (4)' },
-                ],
-                recentChanges: 10,
-                historicalDefects: 1,
-                complexity: 15,
-            },
-            {
-                file: 'src/shared/http/http-client.ts',
-                probability: 0.25,
-                riskLevel: 'low',
-                factors: [
-                    { name: 'stability', weight: 0.5, description: 'Stable code, few recent changes' },
-                ],
-                recentChanges: 2,
-                historicalDefects: 0,
-                complexity: 8,
-            },
-        ];
+    /**
+     * #1043 round 2b: the newest prediction names no usable file, so a stored
+     * measured record from an older prediction is no longer current — stamp it
+     * stale (once) instead of leaving it looking like the latest analysis.
+     */
+    async markPreviousStale(context, supersededBy) {
+        const stored = await context.memory.get(PREDICTIONS_KEY);
+        if (!stored || typeof stored !== 'object' || stored.measured !== true || stored.schemaVersion !== SCHEMA_VERSION)
+            return;
+        if (stored.stale === true || stored.predictionId === supersededBy)
+            return;
+        await context.memory.set(PREDICTIONS_KEY, {
+            ...stored, stale: true, staleSince: new Date().toISOString(), supersededBy,
+        });
     }
-    async identifyHotspots(_context) {
-        // In a real implementation, this would analyze git history
-        return [
-            {
-                file: 'src/coordination/workflow-orchestrator.ts',
-                changeFrequency: 15,
-                defectDensity: 0.02,
-                authors: 4,
-                lastModified: new Date(Date.now() - 2 * 24 * 60 * 60 * 1000),
-                score: 0.85,
-            },
-            {
-                file: 'src/domains/test-generation/services/test-generator.ts',
-                changeFrequency: 12,
-                defectDensity: 0.015,
-                authors: 3,
-                lastModified: new Date(Date.now() - 5 * 24 * 60 * 60 * 1000),
-                score: 0.72,
-            },
-            {
-                file: 'src/mcp/server.ts',
-                changeFrequency: 8,
-                defectDensity: 0.01,
-                authors: 2,
-                lastModified: new Date(Date.now() - 3 * 24 * 60 * 60 * 1000),
-                score: 0.55,
-            },
-        ];
+    /**
+     * #1043: the most recent prediction run the defect-intelligence predictor
+     * recorded, restricted to regular files whose real path is inside the
+     * project root. `measured` is `null` when there is no such prediction;
+     * `unusable` describes a newest prediction none of whose files qualify.
+     */
+    async readLatestPrediction(context) {
+        const keys = await context.memory.search(PREDICTION_KEY_PATTERN, MAX_PREDICTION_RECORDS, { namespace: DEFECT_NAMESPACE });
+        const now = Date.now();
+        let latest = null;
+        for (const found of keys) {
+            // #1049: the in-memory backend returns `<namespace>:<key>` from a namespaced search.
+            const key = bareSearchKey(found, DEFECT_NAMESPACE, PREDICTION_KEY_PATTERN);
+            const parsed = parsePredictionRecord(await context.memory.get(key, { namespace: DEFECT_NAMESPACE }), now);
+            if (parsed && (!latest || Date.parse(parsed.timestamp) > Date.parse(latest.timestamp)))
+                latest = parsed;
+        }
+        if (!latest)
+            return { measured: null, unusable: null };
+        const realRoot = realProjectRoot();
+        const disposition = latest.predictions.map((p) => classifyPredictedFile(p.file, realRoot));
+        const existing = latest.predictions.filter((_, i) => disposition[i] === 'inside');
+        const outside = disposition.filter((d) => d === 'outside').length;
+        const missing = disposition.filter((d) => d === 'missing').length;
+        if (outside > 0) {
+            context.logger.warn('Ignoring defect predictions for paths outside the project root', { predictionId: latest.id, outside });
+        }
+        if (existing.length === 0) {
+            context.logger.info('Latest defect prediction names no file inside the project', { predictionId: latest.id });
+            return { measured: null, unusable: { id: latest.id, outside, missing } };
+        }
+        return { unusable: null, measured: {
+                measured: true,
+                schemaVersion: SCHEMA_VERSION,
+                source: 'defect-intelligence',
+                predictionId: latest.id,
+                predictedAt: latest.timestamp,
+                evaluatedAt: new Date().toISOString(),
+                predictions: existing,
+                missingFilesSkipped: missing,
+                outsideProjectSkipped: outside,
+            } };
     }
     analyzeHighRiskPredictions(predictions, findings, recommendations) {
         const critical = predictions.filter(p => p.riskLevel === 'critical');
         const high = predictions.filter(p => p.riskLevel === 'high');
-        for (const prediction of critical) {
+        for (const prediction of [...critical, ...high]) {
+            const name = prediction.file.split(/[\\/]/).pop();
             findings.push({
-                type: 'critical-defect-risk',
-                severity: 'critical',
+                type: `${prediction.riskLevel}-defect-risk`,
+                severity: prediction.riskLevel,
                 domain: 'defect-intelligence',
-                title: `Critical Defect Risk: ${prediction.file.split('/').pop()}`,
-                description: `${(prediction.probability * 100).toFixed(0)}% probability of defect based on code analysis`,
+                title: `${prediction.riskLevel === 'critical' ? 'Critical' : 'High'} Defect Risk: ${name}`,
+                description: `${(prediction.probability * 100).toFixed(0)}% predicted defect probability (defect-intelligence predictor)`,
                 resource: prediction.file,
                 context: {
                     probability: prediction.probability,
                     factors: prediction.factors,
-                    recentChanges: prediction.recentChanges,
-                    historicalDefects: prediction.historicalDefects,
+                    recommendations: prediction.recommendations,
                 },
             });
         }
-        for (const prediction of high) {
-            findings.push({
-                type: 'high-defect-risk',
-                severity: 'high',
-                domain: 'defect-intelligence',
-                title: `High Defect Risk: ${prediction.file.split('/').pop()}`,
-                description: `${(prediction.probability * 100).toFixed(0)}% probability of defect`,
-                resource: prediction.file,
-                context: {
-                    probability: prediction.probability,
-                    topFactors: prediction.factors.slice(0, 3),
-                    complexity: prediction.complexity,
-                },
-            });
-        }
-        if (critical.length > 0 || high.length > 0) {
+        if (critical.length === 0 && high.length === 0)
+            return;
+        recommendations.push({
+            priority: 'p0',
+            domain: 'defect-intelligence',
+            action: 'Prioritize Code Review for High-Risk Files',
+            description: `${critical.length + high.length} files have elevated predicted defect risk. Schedule focused code reviews.`,
+            estimatedImpact: 'high',
+            effort: 'medium',
+            autoFixable: false,
+        });
+        const risky = [...critical, ...high];
+        const complexity = risky.filter(p => p.factors.some(f => f.name === 'codeComplexity' && f.contribution > 0.1));
+        if (complexity.length > 0) {
             recommendations.push({
-                priority: 'p0',
+                priority: 'p1',
                 domain: 'defect-intelligence',
-                action: 'Prioritize Code Review for High-Risk Files',
-                description: `${critical.length + high.length} files have elevated defect risk. Schedule focused code reviews.`,
+                action: 'Reduce Code Complexity',
+                description: `${complexity.length} high-risk files have code complexity as a major risk factor. Consider refactoring.`,
                 estimatedImpact: 'high',
-                effort: 'medium',
-                autoFixable: false,
-            });
-            // Specific recommendations based on factors
-            const complexityIssues = predictions.filter(p => p.factors.some(f => f.name === 'complexity' && f.weight > 0.25));
-            if (complexityIssues.length > 0) {
-                recommendations.push({
-                    priority: 'p1',
-                    domain: 'defect-intelligence',
-                    action: 'Reduce Code Complexity',
-                    description: `${complexityIssues.length} files have high complexity as a defect risk factor. Consider refactoring.`,
-                    estimatedImpact: 'high',
-                    effort: 'high',
-                    autoFixable: false,
-                });
-            }
-            const coverageIssues = predictions.filter(p => p.factors.some(f => f.name === 'test-coverage'));
-            if (coverageIssues.length > 0) {
-                recommendations.push({
-                    priority: 'p1',
-                    domain: 'defect-intelligence',
-                    action: 'Improve Test Coverage for Risky Files',
-                    description: `${coverageIssues.length} high-risk files have low test coverage. Add targeted tests.`,
-                    estimatedImpact: 'high',
-                    effort: 'medium',
-                    autoFixable: true,
-                });
-            }
-        }
-    }
-    analyzeHotspots(hotspots, findings, recommendations) {
-        const criticalHotspots = hotspots.filter(h => h.score > 0.7);
-        if (criticalHotspots.length > 0) {
-            findings.push({
-                type: 'code-hotspots',
-                severity: 'medium',
-                domain: 'defect-intelligence',
-                title: `${criticalHotspots.length} Code Hotspots Detected`,
-                description: 'Files with high change frequency and defect density identified',
-                context: {
-                    hotspots: criticalHotspots.map(h => ({
-                        file: h.file,
-                        changeFrequency: h.changeFrequency,
-                        score: h.score.toFixed(2),
-                    })),
-                },
-            });
-            recommendations.push({
-                priority: 'p2',
-                domain: 'defect-intelligence',
-                action: 'Address Code Hotspots',
-                description: 'Consider refactoring frequently-changed, defect-prone files to improve stability.',
-                estimatedImpact: 'medium',
                 effort: 'high',
                 autoFixable: false,
             });
         }
-        // Check for knowledge silos (files with single author and high change frequency)
-        const silos = hotspots.filter(h => h.authors === 1 && h.changeFrequency > 5);
-        if (silos.length > 0) {
-            findings.push({
-                type: 'knowledge-silo',
-                severity: 'medium',
-                domain: 'defect-intelligence',
-                title: 'Knowledge Silos Detected',
-                description: `${silos.length} frequently-changed files have only one author`,
-                context: {
-                    files: silos.map(s => s.file),
-                },
-            });
+        const coverage = risky.filter(p => p.factors.some(f => f.name === 'testCoverage' && f.contribution > 0.1));
+        if (coverage.length > 0) {
             recommendations.push({
-                priority: 'p2',
+                priority: 'p1',
                 domain: 'defect-intelligence',
-                action: 'Address Knowledge Silos',
-                description: 'Ensure knowledge sharing for files maintained by single developers.',
-                estimatedImpact: 'medium',
-                effort: 'low',
-                autoFixable: false,
+                action: 'Improve Test Coverage for Risky Files',
+                description: `${coverage.length} high-risk files have low test coverage as a major risk factor. Add targeted tests.`,
+                estimatedImpact: 'high',
+                effort: 'medium',
+                autoFixable: true,
             });
         }
     }
-    calculateHealthScore(predictions, hotspots) {
+    calculateHealthScore(predictions) {
         if (predictions.length === 0)
             return 100;
-        // Base score from average defect probability (inverted)
         const avgProbability = predictions.reduce((sum, p) => sum + p.probability, 0) / predictions.length;
-        let score = (1 - avgProbability) * 60;
-        // Penalty for high-risk files
-        const highRiskCount = predictions.filter(p => p.riskLevel === 'critical' || p.riskLevel === 'high').length;
-        score -= (highRiskCount / predictions.length) * 20;
-        // Penalty for hotspots
-        const criticalHotspots = hotspots.filter(h => h.score > 0.7).length;
-        score -= criticalHotspots * 3;
-        // Bonus for stable codebase
-        const lowRiskRatio = predictions.filter(p => p.riskLevel === 'low').length / predictions.length;
-        score += lowRiskRatio * 20;
+        const highRisk = predictions.filter(p => p.riskLevel === 'critical' || p.riskLevel === 'high').length;
+        const score = (1 - avgProbability) * 80 - (highRisk / predictions.length) * 20 + 20;
         return Math.max(0, Math.min(100, Math.round(score)));
     }
 }

package/dist/workers/workers/performance-baseline.d.ts

@@ -2,24 +2,39 @@
  * Agentic QE v3 - Performance Baseline Worker
  * ADR-014: Background Workers for QE Monitoring
  *
- * Tracks performance trends including:
- * - Test execution time baselines
- * - Build performance tracking
- * - Resource utilization trends
- * - Performance anomaly detection
+ * Tracks test execution time against a baseline built from measured runs.
+ *
+ * #1043: every figure comes from a recorded test run — the run behind the
+ * latest usable testsPassing evidence (`test-run:<runId>` in `test-execution`),
+ * read the same way the regression monitor reads it (#1028). Build and
+ * resource figures have no measured source and are reported unmeasured.
+ * Nothing measured → nothing stored or compared. A baseline without the
+ * measured stamp (the invented one written by <= 3.15.6) is ignored and
+ * replaced by the next measured run.
  */
 import { BaseWorker } from '../base-worker';
 import { WorkerContext, WorkerResult } from '../interfaces';
 export declare class PerformanceBaselineWorker extends BaseWorker {
     constructor();
     protected doExecute(context: WorkerContext): Promise<WorkerResult>;
-    private collectPerformanceMetrics;
-    private getOrCalculateBaseline;
+    private baselineStatus;
+    /**
+     * #1043: the run behind the latest usable testsPassing evidence, with its
+     * recorded wall time and test count. `null` when there is no such evidence
+     * or the run's result was not recorded.
+     */
+    private collectMeasuredRun;
+    /**
+     * Compare a measured run with the measured baseline of the same test set.
+     * The recorded duration is the run's wall time; per-test time across
+     * different test sets would mostly measure runner overhead (#1028 N1).
+     */
     private detectAnomalies;
     private analyzeAnomalies;
     private formatMetricName;
-    private updateBaseline;
-    private storeMetricsHistory;
+    /** Blend a measured run into a comparable baseline, or start a new one from it. */
+    private nextBaseline;
+    private appendHistory;
     private calculateHealthScore;
 }
 //# sourceMappingURL=performance-baseline.d.ts.map