@claude-flow/aidefence 3.0.0-alpha.1

package/dist/index.js

@@ -0,0 +1,188 @@
+/**
+ * @claude-flow/aidefence
+ *
+ * AI Manipulation Defense System with self-learning capabilities.
+ *
+ * Features:
+ * - 50+ prompt injection patterns
+ * - HNSW-indexed threat pattern search (150x-12,500x faster with AgentDB)
+ * - ReasoningBank-style pattern learning
+ * - Adaptive mitigation with effectiveness tracking
+ * - Strange-loop meta-learning integration
+ *
+ * @example
+ * ```typescript
+ * import { createAIDefence } from '@claude-flow/aidefence';
+ *
+ * const aidefence = createAIDefence({ enableLearning: true });
+ *
+ * // Detect threats
+ * const result = await aidefence.detect('Ignore all previous instructions');
+ * console.log(result.safe); // false
+ *
+ * // Search similar patterns (uses HNSW when connected to AgentDB)
+ * const similar = await aidefence.searchSimilarThreats('system prompt injection');
+ *
+ * // Learn from feedback
+ * await aidefence.learnFromDetection(input, result, { wasAccurate: true });
+ * ```
+ */
+export { createThreat } from './domain/entities/threat.js';
+// Domain services
+export { ThreatDetectionService, createThreatDetectionService } from './domain/services/threat-detection-service.js';
+export { ThreatLearningService, createThreatLearningService, InMemoryVectorStore, } from './domain/services/threat-learning-service.js';
+// Import for internal use
+import { createThreatDetectionService } from './domain/services/threat-detection-service.js';
+import { createThreatLearningService } from './domain/services/threat-learning-service.js';
+/**
+ * Create an AIDefence instance with optional learning capabilities
+ *
+ * @example
+ * ```typescript
+ * // Simple usage (detection only)
+ * const simple = createAIDefence();
+ *
+ * // With learning enabled
+ * const learning = createAIDefence({ enableLearning: true });
+ *
+ * // With AgentDB for HNSW search (150x-12,500x faster)
+ * import { AgentDB } from 'agentdb';
+ * const agentdb = new AgentDB({ path: './data/aidefence' });
+ * const fast = createAIDefence({
+ *   enableLearning: true,
+ *   vectorStore: agentdb
+ * });
+ * ```
+ */
+export function createAIDefence(config = {}) {
+    const detectionService = createThreatDetectionService();
+    const learningService = config.enableLearning
+        ? createThreatLearningService(config.vectorStore)
+        : null;
+    return {
+        async detect(input) {
+            const result = detectionService.detect(input);
+            // Auto-learn if enabled
+            if (learningService && result.threats.length > 0) {
+                await learningService.learnFromDetection(input, result);
+            }
+            return result;
+        },
+        quickScan(input) {
+            return detectionService.quickScan(input);
+        },
+        hasPII(input) {
+            return detectionService.detectPII(input);
+        },
+        async searchSimilarThreats(query, options) {
+            if (!learningService) {
+                return [];
+            }
+            return learningService.searchSimilarThreats(query, options);
+        },
+        async learnFromDetection(input, result, feedback) {
+            if (!learningService) {
+                console.warn('Learning not enabled. Pass { enableLearning: true } to createAIDefence()');
+                return;
+            }
+            await learningService.learnFromDetection(input, result, feedback);
+        },
+        async recordMitigation(threatType, strategy, success) {
+            if (!learningService)
+                return;
+            await learningService.recordMitigation(threatType, strategy, success);
+        },
+        async getBestMitigation(threatType) {
+            if (!learningService)
+                return null;
+            return learningService.getBestMitigation(threatType);
+        },
+        startTrajectory(sessionId, task) {
+            learningService?.startTrajectory(sessionId, task);
+        },
+        async endTrajectory(sessionId, verdict) {
+            await learningService?.endTrajectory(sessionId, verdict);
+        },
+        async getStats() {
+            const detectionStats = detectionService.getStats();
+            const learningStats = learningService
+                ? await learningService.getStats()
+                : { learnedPatterns: 0, mitigationStrategies: 0, avgEffectiveness: 0 };
+            return {
+                detectionCount: detectionStats.detectionCount,
+                avgDetectionTimeMs: detectionStats.avgDetectionTimeMs,
+                learnedPatterns: learningStats.learnedPatterns,
+                mitigationStrategies: learningStats.mitigationStrategies,
+                avgMitigationEffectiveness: learningStats.avgEffectiveness,
+            };
+        },
+    };
+}
+/**
+ * Singleton instance for convenience
+ */
+let defaultInstance = null;
+/**
+ * Get the default AIDefence instance (singleton, learning enabled)
+ */
+export function getAIDefence() {
+    if (!defaultInstance) {
+        defaultInstance = createAIDefence({ enableLearning: true });
+    }
+    return defaultInstance;
+}
+/**
+ * Convenience function for quick threat check
+ */
+export function isSafe(input) {
+    const service = createThreatDetectionService();
+    return service.detect(input).safe;
+}
+/**
+ * Convenience function for quick threat check with details
+ */
+export function checkThreats(input) {
+    const service = createThreatDetectionService();
+    return service.detect(input);
+}
+/**
+ * Calculate security consensus from multiple agent assessments
+ * Uses attention-based weighting for flash attention integration
+ */
+export function calculateSecurityConsensus(assessments) {
+    if (assessments.length === 0) {
+        return { consensus: 'uncertain', confidence: 0, criticalThreats: [] };
+    }
+    // Normalize weights
+    const totalWeight = assessments.reduce((sum, a) => sum + a.weight, 0);
+    const normalized = assessments.map(a => ({
+        ...a,
+        weight: a.weight / totalWeight,
+    }));
+    // Calculate weighted threat score
+    let threatScore = 0;
+    const allThreats = [];
+    for (const assessment of normalized) {
+        if (!assessment.threatAssessment.safe) {
+            threatScore += assessment.weight;
+            allThreats.push(...assessment.threatAssessment.threats);
+        }
+    }
+    // Determine consensus
+    const criticalThreats = allThreats.filter(t => t.severity === 'critical');
+    if (criticalThreats.length > 0) {
+        return {
+            consensus: 'threat',
+            confidence: Math.max(...criticalThreats.map(t => t.confidence)),
+            criticalThreats,
+        };
+    }
+    if (threatScore > 0.5) {
+        return { consensus: 'threat', confidence: threatScore, criticalThreats: [] };
+    }
+    if (threatScore < 0.2) {
+        return { consensus: 'safe', confidence: 1 - threatScore, criticalThreats: [] };
+    }
+    return { consensus: 'uncertain', confidence: 0.5, criticalThreats: [] };
+}
+//# sourceMappingURL=index.js.map

package/dist/index.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.js","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA4BG;AAYH,OAAO,EAAE,YAAY,EAAE,MAAM,6BAA6B,CAAC;AAE3D,kBAAkB;AAClB,OAAO,EAAE,sBAAsB,EAAE,4BAA4B,EAAE,MAAM,+CAA+C,CAAC;AASrH,OAAO,EACL,qBAAqB,EACrB,2BAA2B,EAC3B,mBAAmB,GACpB,MAAM,8CAA8C,CAAC;AAEtD,0BAA0B;AAC1B,OAAO,EAAE,4BAA4B,EAAE,MAAM,+CAA+C,CAAC;AAC7F,OAAO,EAAE,2BAA2B,EAAE,MAAM,8CAA8C,CAAC;AA6F3F;;;;;;;;;;;;;;;;;;;GAmBG;AACH,MAAM,UAAU,eAAe,CAAC,SAA0B,EAAE;IAC1D,MAAM,gBAAgB,GAAG,4BAA4B,EAAE,CAAC;IACxD,MAAM,eAAe,GAAG,MAAM,CAAC,cAAc;QAC3C,CAAC,CAAC,2BAA2B,CAAC,MAAM,CAAC,WAAW,CAAC;QACjD,CAAC,CAAC,IAAI,CAAC;IAET,OAAO;QACL,KAAK,CAAC,MAAM,CAAC,KAAa;YACxB,MAAM,MAAM,GAAG,gBAAgB,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC;YAE9C,wBAAwB;YACxB,IAAI,eAAe,IAAI,MAAM,CAAC,OAAO,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;gBACjD,MAAM,eAAe,CAAC,kBAAkB,CAAC,KAAK,EAAE,MAAM,CAAC,CAAC;YAC1D,CAAC;YAED,OAAO,MAAM,CAAC;QAChB,CAAC;QAED,SAAS,CAAC,KAAa;YACrB,OAAO,gBAAgB,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC;QAC3C,CAAC;QAED,MAAM,CAAC,KAAa;YAClB,OAAO,gBAAgB,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC;QAC3C,CAAC;QAED,KAAK,CAAC,oBAAoB,CAAC,KAAK,EAAE,OAAO;YACvC,IAAI,CAAC,eAAe,EAAE,CAAC;gBACrB,OAAO,EAAE,CAAC;YACZ,CAAC;YACD,OAAO,eAAe,CAAC,oBAAoB,CAAC,KAAK,EAAE,OAAO,CAAC,CAAC;QAC9D,CAAC;QAED,KAAK,CAAC,kBAAkB,CAAC,KAAK,EAAE,MAAM,EAAE,QAAQ;YAC9C,IAAI,CAAC,eAAe,EAAE,CAAC;gBACrB,OAAO,CAAC,IAAI,CAAC,0EAA0E,CAAC,CAAC;gBACzF,OAAO;YACT,CAAC;YACD,MAAM,eAAe,CAAC,kBAAkB,CAAC,KAAK,EAAE,MAAM,EAAE,QAAQ,CAAC,CAAC;QACpE,CAAC;QAED,KAAK,CAAC,gBAAgB,CAAC,UAAU,EAAE,QAAQ,EAAE,OAAO;YAClD,IAAI,CAAC,eAAe;gBAAE,OAAO;YAC7B,MAAM,eAAe,CAAC,gBAAgB,CAAC,UAAU,EAAE,QAAQ,EAAE,OAAO,CAAC,CAAC;QACxE,CAAC;QAED,KAAK,CAAC,iBAAiB,CAAC,UAAU;YAChC,IAAI,CAAC,eAAe;gBAAE,OAAO,IAAI,CAAC;YAClC,OAAO,eAAe,CAAC,iBAAiB,CAAC,UAAU,CAAC,CAAC;QACvD,CAAC;QAED,eAAe,CAAC,SAAS,EAAE,IAAI;YAC7B,eAAe,EAAE,eAAe,CAAC,SAAS,EAAE,IAAI,CAAC,CAAC;QACpD,CAAC;QAED,KAAK,CAAC,aAAa,CAAC,SAAS,EAAE,OAAO;YACpC,MAAM,eAAe,EAAE,aAAa,CAAC,SAAS,EAAE,OAAO,CAAC,CAAC;QAC3D,CAAC;QAED,KAAK,CAAC,QAAQ;YACZ,MAAM,cAAc,GAAG,gBAAgB,CAAC,QAAQ,EAAE,CAAC;YACnD,MAAM,aAAa,GAAG,eAAe;gBACnC,CAAC,CAAC,MAAM,eAAe,CAAC,QAAQ,EAAE;gBAClC,CAAC,CAAC,EAAE,eAAe,EAAE,CAAC,EAAE,oBAAoB,EAAE,CAAC,EAAE,gBAAgB,EAAE,CAAC,EAAE,CAAC;YAEzE,OAAO;gBACL,cAAc,EAAE,cAAc,CAAC,cAAc;gBAC7C,kBAAkB,EAAE,cAAc,CAAC,kBAAkB;gBACrD,eAAe,EAAE,aAAa,CAAC,eAAe;gBAC9C,oBAAoB,EAAE,aAAa,CAAC,oBAAoB;gBACxD,0BAA0B,EAAE,aAAa,CAAC,gBAAgB;aAC3D,CAAC;QACJ,CAAC;KACF,CAAC;AACJ,CAAC;AAED;;GAEG;AACH,IAAI,eAAe,GAAqB,IAAI,CAAC;AAE7C;;GAEG;AACH,MAAM,UAAU,YAAY;IAC1B,IAAI,CAAC,eAAe,EAAE,CAAC;QACrB,eAAe,GAAG,eAAe,CAAC,EAAE,cAAc,EAAE,IAAI,EAAE,CAAC,CAAC;IAC9D,CAAC;IACD,OAAO,eAAe,CAAC;AACzB,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,MAAM,CAAC,KAAa;IAClC,MAAM,OAAO,GAAG,4BAA4B,EAAE,CAAC;IAC/C,OAAO,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,IAAI,CAAC;AACpC,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,YAAY,CAAC,KAAa;IACxC,MAAM,OAAO,GAAG,4BAA4B,EAAE,CAAC;IAC/C,OAAO,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC;AAC/B,CAAC;AAYD;;;GAGG;AACH,MAAM,UAAU,0BAA0B,CACxC,WAA+B;IAM/B,IAAI,WAAW,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAC7B,OAAO,EAAE,SAAS,EAAE,WAAW,EAAE,UAAU,EAAE,CAAC,EAAE,eAAe,EAAE,EAAE,EAAE,CAAC;IACxE,CAAC;IAED,oBAAoB;IACpB,MAAM,WAAW,GAAG,WAAW,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,CAAC,EAAE,EAAE,CAAC,GAAG,GAAG,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;IACtE,MAAM,UAAU,GAAG,WAAW,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;QACvC,GAAG,CAAC;QACJ,MAAM,EAAE,CAAC,CAAC,MAAM,GAAG,WAAW;KAC/B,CAAC,CAAC,CAAC;IAEJ,kCAAkC;IAClC,IAAI,WAAW,GAAG,CAAC,CAAC;IACpB,MAAM,UAAU,GAAa,EAAE,CAAC;IAEhC,KAAK,MAAM,UAAU,IAAI,UAAU,EAAE,CAAC;QACpC,IAAI,CAAC,UAAU,CAAC,gBAAgB,CAAC,IAAI,EAAE,CAAC;YACtC,WAAW,IAAI,UAAU,CAAC,MAAM,CAAC;YACjC,UAAU,CAAC,IAAI,CAAC,GAAG,UAAU,CAAC,gBAAgB,CAAC,OAAO,CAAC,CAAC;QAC1D,CAAC;IACH,CAAC;IAED,sBAAsB;IACtB,MAAM,eAAe,GAAG,UAAU,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,QAAQ,KAAK,UAAU,CAAC,CAAC;IAE1E,IAAI,eAAe,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QAC/B,OAAO;YACL,SAAS,EAAE,QAAQ;YACnB,UAAU,EAAE,IAAI,CAAC,GAAG,CAAC,GAAG,eAAe,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,UAAU,CAAC,CAAC;YAC/D,eAAe;SAChB,CAAC;IACJ,CAAC;IAED,IAAI,WAAW,GAAG,GAAG,EAAE,CAAC;QACtB,OAAO,EAAE,SAAS,EAAE,QAAQ,EAAE,UAAU,EAAE,WAAW,EAAE,eAAe,EAAE,EAAE,EAAE,CAAC;IAC/E,CAAC;IAED,IAAI,WAAW,GAAG,GAAG,EAAE,CAAC;QACtB,OAAO,EAAE,SAAS,EAAE,MAAM,EAAE,UAAU,EAAE,CAAC,GAAG,WAAW,EAAE,eAAe,EAAE,EAAE,EAAE,CAAC;IACjF,CAAC;IAED,OAAO,EAAE,SAAS,EAAE,WAAW,EAAE,UAAU,EAAE,GAAG,EAAE,eAAe,EAAE,EAAE,EAAE,CAAC;AAC1E,CAAC"}

package/package.json

@@ -0,0 +1,76 @@
+{
+  "name": "@claude-flow/aidefence",
+  "version": "3.0.0-alpha.1",
+  "description": "AI Manipulation Defense System (AIMDS) with self-learning, prompt injection detection, and vector search integration",
+  "type": "module",
+  "main": "dist/index.js",
+  "types": "dist/index.d.ts",
+  "exports": {
+    ".": {
+      "import": "./dist/index.js",
+      "types": "./dist/index.d.ts"
+    },
+    "./detection": {
+      "import": "./dist/domain/services/threat-detection-service.js",
+      "types": "./dist/domain/services/threat-detection-service.d.ts"
+    },
+    "./learning": {
+      "import": "./dist/domain/services/threat-learning-service.js",
+      "types": "./dist/domain/services/threat-learning-service.d.ts"
+    }
+  },
+  "files": [
+    "dist",
+    "README.md"
+  ],
+  "scripts": {
+    "build": "tsc",
+    "prepublishOnly": "npm run build",
+    "test": "vitest",
+    "test:unit": "vitest run src",
+    "typecheck": "tsc --noEmit"
+  },
+  "peerDependencies": {
+    "agentdb": ">=2.0.0-alpha.1"
+  },
+  "peerDependenciesMeta": {
+    "agentdb": {
+      "optional": true
+    }
+  },
+  "devDependencies": {
+    "@types/node": "^20.10.0",
+    "typescript": "^5.3.3",
+    "vitest": "^1.1.0"
+  },
+  "keywords": [
+    "ai-security",
+    "prompt-injection",
+    "jailbreak-detection",
+    "threat-detection",
+    "pii-detection",
+    "claude-flow",
+    "vector-search",
+    "self-learning",
+    "aimds",
+    "llm-security"
+  ],
+  "author": "rUv <hello@ruv.io>",
+  "license": "MIT",
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/ruvnet/claude-flow.git",
+    "directory": "v3/@claude-flow/aidefence"
+  },
+  "bugs": {
+    "url": "https://github.com/ruvnet/claude-flow/issues"
+  },
+  "homepage": "https://github.com/ruvnet/claude-flow/tree/main/v3/@claude-flow/aidefence#readme",
+  "publishConfig": {
+    "access": "public",
+    "tag": "alpha"
+  },
+  "engines": {
+    "node": ">=18.0.0"
+  }
+}