agentic-flow 2.0.7 → 3.0.0-alpha.2

package/dist/agentic-flow/src/mcp/fastmcp/tools/quantization-tools.d.ts

@@ -0,0 +1,7 @@
+/**
+ * Quantization MCP Tools
+ *
+ * 8 tools for model quantization, distillation, pruning, and caching
+ */
+export declare function registerQuantizationTools(server: any): void;
+//# sourceMappingURL=quantization-tools.d.ts.map

package/dist/agentic-flow/src/mcp/fastmcp/tools/quantization-tools.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"quantization-tools.d.ts","sourceRoot":"","sources":["../../../../../../src/mcp/fastmcp/tools/quantization-tools.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAKH,wBAAgB,yBAAyB,CAAC,MAAM,EAAE,GAAG,GAAG,IAAI,CAoY3D"}

package/dist/agentic-flow/src/mcp/fastmcp/tools/quantization-tools.js

@@ -0,0 +1,325 @@
+/**
+ * Quantization MCP Tools
+ *
+ * 8 tools for model quantization, distillation, pruning, and caching
+ */
+import { z } from 'zod';
+import QuantizationService from '../../../services/quantization-service.js';
+export function registerQuantizationTools(server) {
+    const service = QuantizationService;
+    // ==========================================================================
+    // Tool 1: quantize_int8
+    // ==========================================================================
+    server.addTool({
+        name: 'quantize_int8',
+        description: 'Quantize a model to INT8 (4x memory reduction, 2-4x faster inference)',
+        parameters: z.object({
+            modelPath: z.string().describe('Path to original model'),
+            outputPath: z.string().describe('Path to save quantized model'),
+            calibrationSize: z.number().positive().optional().default(100).describe('Calibration dataset size'),
+            symmetric: z.boolean().optional().default(true).describe('Use symmetric quantization'),
+            perChannel: z.boolean().optional().default(true).describe('Per-channel quantization'),
+        }),
+        execute: async ({ modelPath, outputPath, calibrationSize, symmetric, perChannel, }) => {
+            try {
+                const result = await service.quantizeINT8(modelPath, outputPath, {
+                    type: 'int8',
+                    calibrationSize,
+                    symmetric,
+                    perChannel,
+                });
+                return JSON.stringify({
+                    success: true,
+                    data: {
+                        ...result,
+                        message: `Model quantized successfully: ${result.memoryReduction.toFixed(2)}x smaller, ${result.inferenceSpeedup.toFixed(2)}x faster, ${result.accuracyLoss.toFixed(2)}% accuracy loss`,
+                    },
+                    timestamp: new Date().toISOString(),
+                }, null, 2);
+            }
+            catch (error) {
+                return JSON.stringify({
+                    success: false,
+                    error: error.message,
+                    timestamp: new Date().toISOString(),
+                }, null, 2);
+            }
+        },
+    });
+    // ==========================================================================
+    // Tool 2: quantize_int4
+    // ==========================================================================
+    server.addTool({
+        name: 'quantize_int4',
+        description: 'Quantize a model to INT4 (8x memory reduction for embeddings)',
+        parameters: z.object({
+            modelPath: z.string().describe('Path to original model'),
+            outputPath: z.string().describe('Path to save quantized model'),
+            calibrationSize: z.number().positive().optional().default(200).describe('Calibration dataset size'),
+            symmetric: z.boolean().optional().default(true).describe('Use symmetric quantization'),
+        }),
+        execute: async ({ modelPath, outputPath, calibrationSize, symmetric, }) => {
+            try {
+                const result = await service.quantizeINT4(modelPath, outputPath, {
+                    type: 'int4',
+                    calibrationSize,
+                    symmetric,
+                });
+                return JSON.stringify({
+                    success: true,
+                    data: {
+                        ...result,
+                        message: `Model quantized to INT4: ${result.memoryReduction.toFixed(2)}x smaller, ${result.inferenceSpeedup.toFixed(2)}x faster, ${result.accuracyLoss.toFixed(2)}% accuracy loss`,
+                    },
+                    timestamp: new Date().toISOString(),
+                }, null, 2);
+            }
+            catch (error) {
+                return JSON.stringify({
+                    success: false,
+                    error: error.message,
+                    timestamp: new Date().toISOString(),
+                }, null, 2);
+            }
+        },
+    });
+    // ==========================================================================
+    // Tool 3: quantize_dynamic
+    // ==========================================================================
+    server.addTool({
+        name: 'quantize_dynamic',
+        description: 'Dynamic quantization - adapt precision based on task complexity (high/medium/low)',
+        parameters: z.object({
+            modelPath: z.string().describe('Path to original model'),
+            outputPath: z.string().describe('Path to save quantized model'),
+            precisionLevel: z.enum(['high', 'medium', 'low']).describe('Precision level: high (FP16), medium (INT8), low (INT4)'),
+        }),
+        execute: async ({ modelPath, outputPath, precisionLevel, }) => {
+            try {
+                const result = await service.quantizeDynamic(modelPath, outputPath, precisionLevel);
+                return JSON.stringify({
+                    success: true,
+                    data: {
+                        ...result,
+                        message: `Dynamic quantization (${precisionLevel}): ${result.memoryReduction.toFixed(2)}x smaller, ${result.inferenceSpeedup.toFixed(2)}x faster`,
+                    },
+                    timestamp: new Date().toISOString(),
+                }, null, 2);
+            }
+            catch (error) {
+                return JSON.stringify({
+                    success: false,
+                    error: error.message,
+                    timestamp: new Date().toISOString(),
+                }, null, 2);
+            }
+        },
+    });
+    // ==========================================================================
+    // Tool 4: distill_knowledge
+    // ==========================================================================
+    server.addTool({
+        name: 'distill_knowledge',
+        description: 'Transfer knowledge from large teacher model to small student model',
+        parameters: z.object({
+            teacherModel: z.string().describe('Path to teacher model (large, accurate)'),
+            studentModel: z.string().describe('Path to student model (small, fast)'),
+            temperature: z.number().positive().optional().default(3.0).describe('Softmax temperature (higher = softer)'),
+            alpha: z.number().min(0).max(1).optional().default(0.5).describe('Loss weighting (teacher vs ground truth)'),
+            epochs: z.number().positive().optional().default(10).describe('Training epochs'),
+            learningRate: z.number().positive().optional().default(0.001).describe('Learning rate'),
+        }),
+        execute: async ({ teacherModel, studentModel, temperature, alpha, epochs, learningRate, }) => {
+            try {
+                const result = await service.distill({
+                    teacherModel,
+                    studentModel,
+                    temperature,
+                    alpha,
+                    epochs,
+                    learningRate,
+                });
+                return JSON.stringify({
+                    success: true,
+                    data: {
+                        ...result,
+                        message: `Distillation complete: ${result.accuracyRetention.toFixed(1)}% accuracy retained, ${result.sizeReduction.toFixed(2)}x smaller, ${result.speedup.toFixed(2)}x faster`,
+                    },
+                    timestamp: new Date().toISOString(),
+                }, null, 2);
+            }
+            catch (error) {
+                return JSON.stringify({
+                    success: false,
+                    error: error.message,
+                    timestamp: new Date().toISOString(),
+                }, null, 2);
+            }
+        },
+    });
+    // ==========================================================================
+    // Tool 5: prune_model
+    // ==========================================================================
+    server.addTool({
+        name: 'prune_model',
+        description: 'Remove unnecessary weights to reduce model size (magnitude pruning)',
+        parameters: z.object({
+            modelPath: z.string().describe('Path to model to prune'),
+            outputPath: z.string().describe('Path to save pruned model'),
+            sparsity: z.number().min(0).max(1).describe('Target sparsity (0-1, e.g., 0.5 = 50% weights removed)'),
+            method: z.enum(['magnitude', 'structured', 'unstructured', 'l1']).optional().default('magnitude').describe('Pruning method'),
+            fineTune: z.boolean().optional().default(true).describe('Fine-tune after pruning'),
+            fineTuneEpochs: z.number().positive().optional().default(5).describe('Fine-tuning epochs'),
+        }),
+        execute: async ({ modelPath, outputPath, sparsity, method, fineTune, fineTuneEpochs, }) => {
+            try {
+                const result = await service.prune(modelPath, outputPath, {
+                    sparsity,
+                    method,
+                    fineTune,
+                    fineTuneEpochs,
+                });
+                return JSON.stringify({
+                    success: true,
+                    data: {
+                        ...result,
+                        message: `Pruning complete: ${(result.sparsity * 100).toFixed(1)}% sparsity, ${result.sizeReduction.toFixed(2)}x smaller, ${result.speedup.toFixed(2)}x faster, ${result.accuracyLoss.toFixed(2)}% accuracy loss`,
+                    },
+                    timestamp: new Date().toISOString(),
+                }, null, 2);
+            }
+            catch (error) {
+                return JSON.stringify({
+                    success: false,
+                    error: error.message,
+                    timestamp: new Date().toISOString(),
+                }, null, 2);
+            }
+        },
+    });
+    // ==========================================================================
+    // Tool 6: get_cached_model
+    // ==========================================================================
+    server.addTool({
+        name: 'get_cached_model',
+        description: 'Get cached quantized model or return null if not cached',
+        parameters: z.object({
+            modelName: z.string().describe('Model name'),
+            type: z.enum(['int8', 'int4', 'fp16', 'dynamic']).describe('Quantization type'),
+        }),
+        execute: async ({ modelName, type, }) => {
+            try {
+                const path = await service.getCachedModel(modelName, type);
+                if (path) {
+                    return JSON.stringify({
+                        success: true,
+                        data: {
+                            cached: true,
+                            path,
+                            message: `Found cached model: ${path}`,
+                        },
+                        timestamp: new Date().toISOString(),
+                    }, null, 2);
+                }
+                else {
+                    return JSON.stringify({
+                        success: true,
+                        data: {
+                            cached: false,
+                            path: null,
+                            message: 'Model not in cache',
+                        },
+                        timestamp: new Date().toISOString(),
+                    }, null, 2);
+                }
+            }
+            catch (error) {
+                return JSON.stringify({
+                    success: false,
+                    error: error.message,
+                    timestamp: new Date().toISOString(),
+                }, null, 2);
+            }
+        },
+    });
+    // ==========================================================================
+    // Tool 7: clear_model_cache
+    // ==========================================================================
+    server.addTool({
+        name: 'clear_model_cache',
+        description: 'Clear all cached quantized models',
+        parameters: z.object({
+            confirm: z.boolean().default(false).describe('Confirm cache clearing'),
+        }),
+        execute: async ({ confirm }) => {
+            try {
+                if (!confirm) {
+                    return JSON.stringify({
+                        success: false,
+                        error: 'Confirmation required: set confirm=true to clear cache',
+                        timestamp: new Date().toISOString(),
+                    }, null, 2);
+                }
+                await service.clearCache();
+                return JSON.stringify({
+                    success: true,
+                    data: {
+                        message: 'Model cache cleared successfully',
+                    },
+                    timestamp: new Date().toISOString(),
+                }, null, 2);
+            }
+            catch (error) {
+                return JSON.stringify({
+                    success: false,
+                    error: error.message,
+                    timestamp: new Date().toISOString(),
+                }, null, 2);
+            }
+        },
+    });
+    // ==========================================================================
+    // Tool 8: quantization_metrics
+    // ==========================================================================
+    server.addTool({
+        name: 'quantization_metrics',
+        description: 'Get quantization service metrics (jobs, cache stats, performance)',
+        parameters: z.object({}),
+        execute: async () => {
+            try {
+                const metrics = service.getMetrics();
+                const cachedModels = service.listCachedModels();
+                return JSON.stringify({
+                    success: true,
+                    data: {
+                        metrics,
+                        cachedModels: cachedModels.map(m => ({
+                            name: m.name,
+                            type: m.type,
+                            size: `${(m.size / (1024 * 1024)).toFixed(2)} MB`,
+                            hitCount: m.hitCount,
+                            lastAccessed: new Date(m.lastAccessed).toISOString(),
+                        })),
+                        summary: {
+                            totalCached: metrics.cachedModels,
+                            totalMemory: `${(metrics.totalMemory / (1024 * 1024 * 1024)).toFixed(2)} GB`,
+                            cacheHitRate: `${(metrics.cacheHitRate * 100).toFixed(1)}%`,
+                            avgInferenceTime: `${metrics.avgInferenceTime.toFixed(2)} ms`,
+                            quantizationJobs: metrics.quantizationJobs,
+                            uptime: `${(metrics.uptime / 1000 / 60).toFixed(1)} minutes`,
+                        },
+                    },
+                    timestamp: new Date().toISOString(),
+                }, null, 2);
+            }
+            catch (error) {
+                return JSON.stringify({
+                    success: false,
+                    error: error.message,
+                    timestamp: new Date().toISOString(),
+                }, null, 2);
+            }
+        },
+    });
+}
+//# sourceMappingURL=quantization-tools.js.map

package/dist/agentic-flow/src/mcp/fastmcp/tools/quantization-tools.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"quantization-tools.js","sourceRoot":"","sources":["../../../../../../src/mcp/fastmcp/tools/quantization-tools.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAEH,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AACxB,OAAO,mBAAmB,MAAM,2CAA2C,CAAC;AAE5E,MAAM,UAAU,yBAAyB,CAAC,MAAW;IACnD,MAAM,OAAO,GAAG,mBAAmB,CAAC;IAEpC,6EAA6E;IAC7E,wBAAwB;IACxB,6EAA6E;IAC7E,MAAM,CAAC,OAAO,CAAC;QACb,IAAI,EAAE,eAAe;QACrB,WAAW,EAAE,uEAAuE;QACpF,UAAU,EAAE,CAAC,CAAC,MAAM,CAAC;YACnB,SAAS,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,wBAAwB,CAAC;YACxD,UAAU,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,8BAA8B,CAAC;YAC/D,eAAe,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,QAAQ,CAAC,0BAA0B,CAAC;YACnG,SAAS,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,QAAQ,CAAC,4BAA4B,CAAC;YACtF,UAAU,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,QAAQ,CAAC,0BAA0B,CAAC;SACtF,CAAC;QACF,OAAO,EAAE,KAAK,EAAE,EACd,SAAS,EACT,UAAU,EACV,eAAe,EACf,SAAS,EACT,UAAU,GAOX,EAAE,EAAE;YACH,IAAI,CAAC;gBACH,MAAM,MAAM,GAAG,MAAM,OAAO,CAAC,YAAY,CAAC,SAAS,EAAE,UAAU,EAAE;oBAC/D,IAAI,EAAE,MAAM;oBACZ,eAAe;oBACf,SAAS;oBACT,UAAU;iBACX,CAAC,CAAC;gBAEH,OAAO,IAAI,CAAC,SAAS,CAAC;oBACpB,OAAO,EAAE,IAAI;oBACb,IAAI,EAAE;wBACJ,GAAG,MAAM;wBACT,OAAO,EAAE,iCAAiC,MAAM,CAAC,eAAe,CAAC,OAAO,CAAC,CAAC,CAAC,cAAc,MAAM,CAAC,gBAAgB,CAAC,OAAO,CAAC,CAAC,CAAC,aAAa,MAAM,CAAC,YAAY,CAAC,OAAO,CAAC,CAAC,CAAC,iBAAiB;qBACxL;oBACD,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;iBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;YACd,CAAC;YAAC,OAAO,KAAU,EAAE,CAAC;gBACpB,OAAO,IAAI,CAAC,SAAS,CAAC;oBACpB,OAAO,EAAE,KAAK;oBACd,KAAK,EAAE,KAAK,CAAC,OAAO;oBACpB,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;iBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;YACd,CAAC;QACH,CAAC;KACF,CAAC,CAAC;IAEH,6EAA6E;IAC7E,wBAAwB;IACxB,6EAA6E;IAC7E,MAAM,CAAC,OAAO,CAAC;QACb,IAAI,EAAE,eAAe;QACrB,WAAW,EAAE,+DAA+D;QAC5E,UAAU,EAAE,CAAC,CAAC,MAAM,CAAC;YACnB,SAAS,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,wBAAwB,CAAC;YACxD,UAAU,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,8BAA8B,CAAC;YAC/D,eAAe,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,QAAQ,CAAC,0BAA0B,CAAC;YACnG,SAAS,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,QAAQ,CAAC,4BAA4B,CAAC;SACvF,CAAC;QACF,OAAO,EAAE,KAAK,EAAE,EACd,SAAS,EACT,UAAU,EACV,eAAe,EACf,SAAS,GAMV,EAAE,EAAE;YACH,IAAI,CAAC;gBACH,MAAM,MAAM,GAAG,MAAM,OAAO,CAAC,YAAY,CAAC,SAAS,EAAE,UAAU,EAAE;oBAC/D,IAAI,EAAE,MAAM;oBACZ,eAAe;oBACf,SAAS;iBACV,CAAC,CAAC;gBAEH,OAAO,IAAI,CAAC,SAAS,CAAC;oBACpB,OAAO,EAAE,IAAI;oBACb,IAAI,EAAE;wBACJ,GAAG,MAAM;wBACT,OAAO,EAAE,4BAA4B,MAAM,CAAC,eAAe,CAAC,OAAO,CAAC,CAAC,CAAC,cAAc,MAAM,CAAC,gBAAgB,CAAC,OAAO,CAAC,CAAC,CAAC,aAAa,MAAM,CAAC,YAAY,CAAC,OAAO,CAAC,CAAC,CAAC,iBAAiB;qBACnL;oBACD,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;iBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;YACd,CAAC;YAAC,OAAO,KAAU,EAAE,CAAC;gBACpB,OAAO,IAAI,CAAC,SAAS,CAAC;oBACpB,OAAO,EAAE,KAAK;oBACd,KAAK,EAAE,KAAK,CAAC,OAAO;oBACpB,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;iBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;YACd,CAAC;QACH,CAAC;KACF,CAAC,CAAC;IAEH,6EAA6E;IAC7E,2BAA2B;IAC3B,6EAA6E;IAC7E,MAAM,CAAC,OAAO,CAAC;QACb,IAAI,EAAE,kBAAkB;QACxB,WAAW,EAAE,mFAAmF;QAChG,UAAU,EAAE,CAAC,CAAC,MAAM,CAAC;YACnB,SAAS,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,wBAAwB,CAAC;YACxD,UAAU,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,8BAA8B,CAAC;YAC/D,cAAc,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,MAAM,EAAE,QAAQ,EAAE,KAAK,CAAC,CAAC,CAAC,QAAQ,CAAC,yDAAyD,CAAC;SACtH,CAAC;QACF,OAAO,EAAE,KAAK,EAAE,EACd,SAAS,EACT,UAAU,EACV,cAAc,GAKf,EAAE,EAAE;YACH,IAAI,CAAC;gBACH,MAAM,MAAM,GAAG,MAAM,OAAO,CAAC,eAAe,CAAC,SAAS,EAAE,UAAU,EAAE,cAAc,CAAC,CAAC;gBAEpF,OAAO,IAAI,CAAC,SAAS,CAAC;oBACpB,OAAO,EAAE,IAAI;oBACb,IAAI,EAAE;wBACJ,GAAG,MAAM;wBACT,OAAO,EAAE,yBAAyB,cAAc,MAAM,MAAM,CAAC,eAAe,CAAC,OAAO,CAAC,CAAC,CAAC,cAAc,MAAM,CAAC,gBAAgB,CAAC,OAAO,CAAC,CAAC,CAAC,UAAU;qBAClJ;oBACD,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;iBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;YACd,CAAC;YAAC,OAAO,KAAU,EAAE,CAAC;gBACpB,OAAO,IAAI,CAAC,SAAS,CAAC;oBACpB,OAAO,EAAE,KAAK;oBACd,KAAK,EAAE,KAAK,CAAC,OAAO;oBACpB,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;iBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;YACd,CAAC;QACH,CAAC;KACF,CAAC,CAAC;IAEH,6EAA6E;IAC7E,4BAA4B;IAC5B,6EAA6E;IAC7E,MAAM,CAAC,OAAO,CAAC;QACb,IAAI,EAAE,mBAAmB;QACzB,WAAW,EAAE,oEAAoE;QACjF,UAAU,EAAE,CAAC,CAAC,MAAM,CAAC;YACnB,YAAY,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,yCAAyC,CAAC;YAC5E,YAAY,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,qCAAqC,CAAC;YACxE,WAAW,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,QAAQ,CAAC,uCAAuC,CAAC;YAC5G,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,QAAQ,CAAC,0CAA0C,CAAC;YAC5G,MAAM,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,EAAE,CAAC,CAAC,QAAQ,CAAC,iBAAiB,CAAC;YAChF,YAAY,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC,QAAQ,CAAC,eAAe,CAAC;SACxF,CAAC;QACF,OAAO,EAAE,KAAK,EAAE,EACd,YAAY,EACZ,YAAY,EACZ,WAAW,EACX,KAAK,EACL,MAAM,EACN,YAAY,GAQb,EAAE,EAAE;YACH,IAAI,CAAC;gBACH,MAAM,MAAM,GAAG,MAAM,OAAO,CAAC,OAAO,CAAC;oBACnC,YAAY;oBACZ,YAAY;oBACZ,WAAW;oBACX,KAAK;oBACL,MAAM;oBACN,YAAY;iBACb,CAAC,CAAC;gBAEH,OAAO,IAAI,CAAC,SAAS,CAAC;oBACpB,OAAO,EAAE,IAAI;oBACb,IAAI,EAAE;wBACJ,GAAG,MAAM;wBACT,OAAO,EAAE,0BAA0B,MAAM,CAAC,iBAAiB,CAAC,OAAO,CAAC,CAAC,CAAC,wBAAwB,MAAM,CAAC,aAAa,CAAC,OAAO,CAAC,CAAC,CAAC,cAAc,MAAM,CAAC,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,UAAU;qBAC/K;oBACD,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;iBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;YACd,CAAC;YAAC,OAAO,KAAU,EAAE,CAAC;gBACpB,OAAO,IAAI,CAAC,SAAS,CAAC;oBACpB,OAAO,EAAE,KAAK;oBACd,KAAK,EAAE,KAAK,CAAC,OAAO;oBACpB,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;iBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;YACd,CAAC;QACH,CAAC;KACF,CAAC,CAAC;IAEH,6EAA6E;IAC7E,sBAAsB;IACtB,6EAA6E;IAC7E,MAAM,CAAC,OAAO,CAAC;QACb,IAAI,EAAE,aAAa;QACnB,WAAW,EAAE,qEAAqE;QAClF,UAAU,EAAE,CAAC,CAAC,MAAM,CAAC;YACnB,SAAS,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,wBAAwB,CAAC;YACxD,UAAU,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,2BAA2B,CAAC;YAC5D,QAAQ,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,wDAAwD,CAAC;YACrG,MAAM,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,WAAW,EAAE,YAAY,EAAE,cAAc,EAAE,IAAI,CAAC,CAAC,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,WAAW,CAAC,CAAC,QAAQ,CAAC,gBAAgB,CAAC;YAC5H,QAAQ,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,QAAQ,CAAC,yBAAyB,CAAC;YAClF,cAAc,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,oBAAoB,CAAC;SAC3F,CAAC;QACF,OAAO,EAAE,KAAK,EAAE,EACd,SAAS,EACT,UAAU,EACV,QAAQ,EACR,MAAM,EACN,QAAQ,EACR,cAAc,GAQf,EAAE,EAAE;YACH,IAAI,CAAC;gBACH,MAAM,MAAM,GAAG,MAAM,OAAO,CAAC,KAAK,CAAC,SAAS,EAAE,UAAU,EAAE;oBACxD,QAAQ;oBACR,MAAM;oBACN,QAAQ;oBACR,cAAc;iBACf,CAAC,CAAC;gBAEH,OAAO,IAAI,CAAC,SAAS,CAAC;oBACpB,OAAO,EAAE,IAAI;oBACb,IAAI,EAAE;wBACJ,GAAG,MAAM;wBACT,OAAO,EAAE,qBAAqB,CAAC,MAAM,CAAC,QAAQ,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,eAAe,MAAM,CAAC,aAAa,CAAC,OAAO,CAAC,CAAC,CAAC,cAAc,MAAM,CAAC,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,aAAa,MAAM,CAAC,YAAY,CAAC,OAAO,CAAC,CAAC,CAAC,iBAAiB;qBAClN;oBACD,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;iBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;YACd,CAAC;YAAC,OAAO,KAAU,EAAE,CAAC;gBACpB,OAAO,IAAI,CAAC,SAAS,CAAC;oBACpB,OAAO,EAAE,KAAK;oBACd,KAAK,EAAE,KAAK,CAAC,OAAO;oBACpB,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;iBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;YACd,CAAC;QACH,CAAC;KACF,CAAC,CAAC;IAEH,6EAA6E;IAC7E,2BAA2B;IAC3B,6EAA6E;IAC7E,MAAM,CAAC,OAAO,CAAC;QACb,IAAI,EAAE,kBAAkB;QACxB,WAAW,EAAE,yDAAyD;QACtE,UAAU,EAAE,CAAC,CAAC,MAAM,CAAC;YACnB,SAAS,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,YAAY,CAAC;YAC5C,IAAI,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,SAAS,CAAC,CAAC,CAAC,QAAQ,CAAC,mBAAmB,CAAC;SAChF,CAAC;QACF,OAAO,EAAE,KAAK,EAAE,EACd,SAAS,EACT,IAAI,GAIL,EAAE,EAAE;YACH,IAAI,CAAC;gBACH,MAAM,IAAI,GAAG,MAAM,OAAO,CAAC,cAAc,CAAC,SAAS,EAAE,IAAI,CAAC,CAAC;gBAE3D,IAAI,IAAI,EAAE,CAAC;oBACT,OAAO,IAAI,CAAC,SAAS,CAAC;wBACpB,OAAO,EAAE,IAAI;wBACb,IAAI,EAAE;4BACJ,MAAM,EAAE,IAAI;4BACZ,IAAI;4BACJ,OAAO,EAAE,uBAAuB,IAAI,EAAE;yBACvC;wBACD,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;qBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;gBACd,CAAC;qBAAM,CAAC;oBACN,OAAO,IAAI,CAAC,SAAS,CAAC;wBACpB,OAAO,EAAE,IAAI;wBACb,IAAI,EAAE;4BACJ,MAAM,EAAE,KAAK;4BACb,IAAI,EAAE,IAAI;4BACV,OAAO,EAAE,oBAAoB;yBAC9B;wBACD,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;qBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;gBACd,CAAC;YACH,CAAC;YAAC,OAAO,KAAU,EAAE,CAAC;gBACpB,OAAO,IAAI,CAAC,SAAS,CAAC;oBACpB,OAAO,EAAE,KAAK;oBACd,KAAK,EAAE,KAAK,CAAC,OAAO;oBACpB,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;iBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;YACd,CAAC;QACH,CAAC;KACF,CAAC,CAAC;IAEH,6EAA6E;IAC7E,4BAA4B;IAC5B,6EAA6E;IAC7E,MAAM,CAAC,OAAO,CAAC;QACb,IAAI,EAAE,mBAAmB;QACzB,WAAW,EAAE,mCAAmC;QAChD,UAAU,EAAE,CAAC,CAAC,MAAM,CAAC;YACnB,OAAO,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC,QAAQ,CAAC,wBAAwB,CAAC;SACvE,CAAC;QACF,OAAO,EAAE,KAAK,EAAE,EAAE,OAAO,EAAwB,EAAE,EAAE;YACnD,IAAI,CAAC;gBACH,IAAI,CAAC,OAAO,EAAE,CAAC;oBACb,OAAO,IAAI,CAAC,SAAS,CAAC;wBACpB,OAAO,EAAE,KAAK;wBACd,KAAK,EAAE,wDAAwD;wBAC/D,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;qBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;gBACd,CAAC;gBAED,MAAM,OAAO,CAAC,UAAU,EAAE,CAAC;gBAE3B,OAAO,IAAI,CAAC,SAAS,CAAC;oBACpB,OAAO,EAAE,IAAI;oBACb,IAAI,EAAE;wBACJ,OAAO,EAAE,kCAAkC;qBAC5C;oBACD,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;iBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;YACd,CAAC;YAAC,OAAO,KAAU,EAAE,CAAC;gBACpB,OAAO,IAAI,CAAC,SAAS,CAAC;oBACpB,OAAO,EAAE,KAAK;oBACd,KAAK,EAAE,KAAK,CAAC,OAAO;oBACpB,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;iBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;YACd,CAAC;QACH,CAAC;KACF,CAAC,CAAC;IAEH,6EAA6E;IAC7E,+BAA+B;IAC/B,6EAA6E;IAC7E,MAAM,CAAC,OAAO,CAAC;QACb,IAAI,EAAE,sBAAsB;QAC5B,WAAW,EAAE,mEAAmE;QAChF,UAAU,EAAE,CAAC,CAAC,MAAM,CAAC,EAAE,CAAC;QACxB,OAAO,EAAE,KAAK,IAAI,EAAE;YAClB,IAAI,CAAC;gBACH,MAAM,OAAO,GAAG,OAAO,CAAC,UAAU,EAAE,CAAC;gBACrC,MAAM,YAAY,GAAG,OAAO,CAAC,gBAAgB,EAAE,CAAC;gBAEhD,OAAO,IAAI,CAAC,SAAS,CAAC;oBACpB,OAAO,EAAE,IAAI;oBACb,IAAI,EAAE;wBACJ,OAAO;wBACP,YAAY,EAAE,YAAY,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;4BACnC,IAAI,EAAE,CAAC,CAAC,IAAI;4BACZ,IAAI,EAAE,CAAC,CAAC,IAAI;4BACZ,IAAI,EAAE,GAAG,CAAC,CAAC,CAAC,IAAI,GAAG,CAAC,IAAI,GAAG,IAAI,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,KAAK;4BACjD,QAAQ,EAAE,CAAC,CAAC,QAAQ;4BACpB,YAAY,EAAE,IAAI,IAAI,CAAC,CAAC,CAAC,YAAY,CAAC,CAAC,WAAW,EAAE;yBACrD,CAAC,CAAC;wBACH,OAAO,EAAE;4BACP,WAAW,EAAE,OAAO,CAAC,YAAY;4BACjC,WAAW,EAAE,GAAG,CAAC,OAAO,CAAC,WAAW,GAAG,CAAC,IAAI,GAAG,IAAI,GAAG,IAAI,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,KAAK;4BAC5E,YAAY,EAAE,GAAG,CAAC,OAAO,CAAC,YAAY,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG;4BAC3D,gBAAgB,EAAE,GAAG,OAAO,CAAC,gBAAgB,CAAC,OAAO,CAAC,CAAC,CAAC,KAAK;4BAC7D,gBAAgB,EAAE,OAAO,CAAC,gBAAgB;4BAC1C,MAAM,EAAE,GAAG,CAAC,OAAO,CAAC,MAAM,GAAG,IAAI,GAAG,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,UAAU;yBAC7D;qBACF;oBACD,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;iBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;YACd,CAAC;YAAC,OAAO,KAAU,EAAE,CAAC;gBACpB,OAAO,IAAI,CAAC,SAAS,CAAC;oBACpB,OAAO,EAAE,KAAK;oBACd,KAAK,EAAE,KAAK,CAAC,OAAO;oBACpB,SAAS,EAAE,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE;iBACpC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;YACd,CAAC;QACH,CAAC;KACF,CAAC,CAAC;AACL,CAAC","sourcesContent":["/**\n * Quantization MCP Tools\n *\n * 8 tools for model quantization, distillation, pruning, and caching\n */\n\nimport { z } from 'zod';\nimport QuantizationService from '../../../services/quantization-service.js';\n\nexport function registerQuantizationTools(server: any): void {\n  const service = QuantizationService;\n\n  // ==========================================================================\n  // Tool 1: quantize_int8\n  // ==========================================================================\n  server.addTool({\n    name: 'quantize_int8',\n    description: 'Quantize a model to INT8 (4x memory reduction, 2-4x faster inference)',\n    parameters: z.object({\n      modelPath: z.string().describe('Path to original model'),\n      outputPath: z.string().describe('Path to save quantized model'),\n      calibrationSize: z.number().positive().optional().default(100).describe('Calibration dataset size'),\n      symmetric: z.boolean().optional().default(true).describe('Use symmetric quantization'),\n      perChannel: z.boolean().optional().default(true).describe('Per-channel quantization'),\n    }),\n    execute: async ({\n      modelPath,\n      outputPath,\n      calibrationSize,\n      symmetric,\n      perChannel,\n    }: {\n      modelPath: string;\n      outputPath: string;\n      calibrationSize: number;\n      symmetric: boolean;\n      perChannel: boolean;\n    }) => {\n      try {\n        const result = await service.quantizeINT8(modelPath, outputPath, {\n          type: 'int8',\n          calibrationSize,\n          symmetric,\n          perChannel,\n        });\n\n        return JSON.stringify({\n          success: true,\n          data: {\n            ...result,\n            message: `Model quantized successfully: ${result.memoryReduction.toFixed(2)}x smaller, ${result.inferenceSpeedup.toFixed(2)}x faster, ${result.accuracyLoss.toFixed(2)}% accuracy loss`,\n          },\n          timestamp: new Date().toISOString(),\n        }, null, 2);\n      } catch (error: any) {\n        return JSON.stringify({\n          success: false,\n          error: error.message,\n          timestamp: new Date().toISOString(),\n        }, null, 2);\n      }\n    },\n  });\n\n  // ==========================================================================\n  // Tool 2: quantize_int4\n  // ==========================================================================\n  server.addTool({\n    name: 'quantize_int4',\n    description: 'Quantize a model to INT4 (8x memory reduction for embeddings)',\n    parameters: z.object({\n      modelPath: z.string().describe('Path to original model'),\n      outputPath: z.string().describe('Path to save quantized model'),\n      calibrationSize: z.number().positive().optional().default(200).describe('Calibration dataset size'),\n      symmetric: z.boolean().optional().default(true).describe('Use symmetric quantization'),\n    }),\n    execute: async ({\n      modelPath,\n      outputPath,\n      calibrationSize,\n      symmetric,\n    }: {\n      modelPath: string;\n      outputPath: string;\n      calibrationSize: number;\n      symmetric: boolean;\n    }) => {\n      try {\n        const result = await service.quantizeINT4(modelPath, outputPath, {\n          type: 'int4',\n          calibrationSize,\n          symmetric,\n        });\n\n        return JSON.stringify({\n          success: true,\n          data: {\n            ...result,\n            message: `Model quantized to INT4: ${result.memoryReduction.toFixed(2)}x smaller, ${result.inferenceSpeedup.toFixed(2)}x faster, ${result.accuracyLoss.toFixed(2)}% accuracy loss`,\n          },\n          timestamp: new Date().toISOString(),\n        }, null, 2);\n      } catch (error: any) {\n        return JSON.stringify({\n          success: false,\n          error: error.message,\n          timestamp: new Date().toISOString(),\n        }, null, 2);\n      }\n    },\n  });\n\n  // ==========================================================================\n  // Tool 3: quantize_dynamic\n  // ==========================================================================\n  server.addTool({\n    name: 'quantize_dynamic',\n    description: 'Dynamic quantization - adapt precision based on task complexity (high/medium/low)',\n    parameters: z.object({\n      modelPath: z.string().describe('Path to original model'),\n      outputPath: z.string().describe('Path to save quantized model'),\n      precisionLevel: z.enum(['high', 'medium', 'low']).describe('Precision level: high (FP16), medium (INT8), low (INT4)'),\n    }),\n    execute: async ({\n      modelPath,\n      outputPath,\n      precisionLevel,\n    }: {\n      modelPath: string;\n      outputPath: string;\n      precisionLevel: 'high' | 'medium' | 'low';\n    }) => {\n      try {\n        const result = await service.quantizeDynamic(modelPath, outputPath, precisionLevel);\n\n        return JSON.stringify({\n          success: true,\n          data: {\n            ...result,\n            message: `Dynamic quantization (${precisionLevel}): ${result.memoryReduction.toFixed(2)}x smaller, ${result.inferenceSpeedup.toFixed(2)}x faster`,\n          },\n          timestamp: new Date().toISOString(),\n        }, null, 2);\n      } catch (error: any) {\n        return JSON.stringify({\n          success: false,\n          error: error.message,\n          timestamp: new Date().toISOString(),\n        }, null, 2);\n      }\n    },\n  });\n\n  // ==========================================================================\n  // Tool 4: distill_knowledge\n  // ==========================================================================\n  server.addTool({\n    name: 'distill_knowledge',\n    description: 'Transfer knowledge from large teacher model to small student model',\n    parameters: z.object({\n      teacherModel: z.string().describe('Path to teacher model (large, accurate)'),\n      studentModel: z.string().describe('Path to student model (small, fast)'),\n      temperature: z.number().positive().optional().default(3.0).describe('Softmax temperature (higher = softer)'),\n      alpha: z.number().min(0).max(1).optional().default(0.5).describe('Loss weighting (teacher vs ground truth)'),\n      epochs: z.number().positive().optional().default(10).describe('Training epochs'),\n      learningRate: z.number().positive().optional().default(0.001).describe('Learning rate'),\n    }),\n    execute: async ({\n      teacherModel,\n      studentModel,\n      temperature,\n      alpha,\n      epochs,\n      learningRate,\n    }: {\n      teacherModel: string;\n      studentModel: string;\n      temperature: number;\n      alpha: number;\n      epochs: number;\n      learningRate: number;\n    }) => {\n      try {\n        const result = await service.distill({\n          teacherModel,\n          studentModel,\n          temperature,\n          alpha,\n          epochs,\n          learningRate,\n        });\n\n        return JSON.stringify({\n          success: true,\n          data: {\n            ...result,\n            message: `Distillation complete: ${result.accuracyRetention.toFixed(1)}% accuracy retained, ${result.sizeReduction.toFixed(2)}x smaller, ${result.speedup.toFixed(2)}x faster`,\n          },\n          timestamp: new Date().toISOString(),\n        }, null, 2);\n      } catch (error: any) {\n        return JSON.stringify({\n          success: false,\n          error: error.message,\n          timestamp: new Date().toISOString(),\n        }, null, 2);\n      }\n    },\n  });\n\n  // ==========================================================================\n  // Tool 5: prune_model\n  // ==========================================================================\n  server.addTool({\n    name: 'prune_model',\n    description: 'Remove unnecessary weights to reduce model size (magnitude pruning)',\n    parameters: z.object({\n      modelPath: z.string().describe('Path to model to prune'),\n      outputPath: z.string().describe('Path to save pruned model'),\n      sparsity: z.number().min(0).max(1).describe('Target sparsity (0-1, e.g., 0.5 = 50% weights removed)'),\n      method: z.enum(['magnitude', 'structured', 'unstructured', 'l1']).optional().default('magnitude').describe('Pruning method'),\n      fineTune: z.boolean().optional().default(true).describe('Fine-tune after pruning'),\n      fineTuneEpochs: z.number().positive().optional().default(5).describe('Fine-tuning epochs'),\n    }),\n    execute: async ({\n      modelPath,\n      outputPath,\n      sparsity,\n      method,\n      fineTune,\n      fineTuneEpochs,\n    }: {\n      modelPath: string;\n      outputPath: string;\n      sparsity: number;\n      method: 'magnitude' | 'structured' | 'unstructured' | 'l1';\n      fineTune: boolean;\n      fineTuneEpochs: number;\n    }) => {\n      try {\n        const result = await service.prune(modelPath, outputPath, {\n          sparsity,\n          method,\n          fineTune,\n          fineTuneEpochs,\n        });\n\n        return JSON.stringify({\n          success: true,\n          data: {\n            ...result,\n            message: `Pruning complete: ${(result.sparsity * 100).toFixed(1)}% sparsity, ${result.sizeReduction.toFixed(2)}x smaller, ${result.speedup.toFixed(2)}x faster, ${result.accuracyLoss.toFixed(2)}% accuracy loss`,\n          },\n          timestamp: new Date().toISOString(),\n        }, null, 2);\n      } catch (error: any) {\n        return JSON.stringify({\n          success: false,\n          error: error.message,\n          timestamp: new Date().toISOString(),\n        }, null, 2);\n      }\n    },\n  });\n\n  // ==========================================================================\n  // Tool 6: get_cached_model\n  // ==========================================================================\n  server.addTool({\n    name: 'get_cached_model',\n    description: 'Get cached quantized model or return null if not cached',\n    parameters: z.object({\n      modelName: z.string().describe('Model name'),\n      type: z.enum(['int8', 'int4', 'fp16', 'dynamic']).describe('Quantization type'),\n    }),\n    execute: async ({\n      modelName,\n      type,\n    }: {\n      modelName: string;\n      type: 'int8' | 'int4' | 'fp16' | 'dynamic';\n    }) => {\n      try {\n        const path = await service.getCachedModel(modelName, type);\n\n        if (path) {\n          return JSON.stringify({\n            success: true,\n            data: {\n              cached: true,\n              path,\n              message: `Found cached model: ${path}`,\n            },\n            timestamp: new Date().toISOString(),\n          }, null, 2);\n        } else {\n          return JSON.stringify({\n            success: true,\n            data: {\n              cached: false,\n              path: null,\n              message: 'Model not in cache',\n            },\n            timestamp: new Date().toISOString(),\n          }, null, 2);\n        }\n      } catch (error: any) {\n        return JSON.stringify({\n          success: false,\n          error: error.message,\n          timestamp: new Date().toISOString(),\n        }, null, 2);\n      }\n    },\n  });\n\n  // ==========================================================================\n  // Tool 7: clear_model_cache\n  // ==========================================================================\n  server.addTool({\n    name: 'clear_model_cache',\n    description: 'Clear all cached quantized models',\n    parameters: z.object({\n      confirm: z.boolean().default(false).describe('Confirm cache clearing'),\n    }),\n    execute: async ({ confirm }: { confirm: boolean }) => {\n      try {\n        if (!confirm) {\n          return JSON.stringify({\n            success: false,\n            error: 'Confirmation required: set confirm=true to clear cache',\n            timestamp: new Date().toISOString(),\n          }, null, 2);\n        }\n\n        await service.clearCache();\n\n        return JSON.stringify({\n          success: true,\n          data: {\n            message: 'Model cache cleared successfully',\n          },\n          timestamp: new Date().toISOString(),\n        }, null, 2);\n      } catch (error: any) {\n        return JSON.stringify({\n          success: false,\n          error: error.message,\n          timestamp: new Date().toISOString(),\n        }, null, 2);\n      }\n    },\n  });\n\n  // ==========================================================================\n  // Tool 8: quantization_metrics\n  // ==========================================================================\n  server.addTool({\n    name: 'quantization_metrics',\n    description: 'Get quantization service metrics (jobs, cache stats, performance)',\n    parameters: z.object({}),\n    execute: async () => {\n      try {\n        const metrics = service.getMetrics();\n        const cachedModels = service.listCachedModels();\n\n        return JSON.stringify({\n          success: true,\n          data: {\n            metrics,\n            cachedModels: cachedModels.map(m => ({\n              name: m.name,\n              type: m.type,\n              size: `${(m.size / (1024 * 1024)).toFixed(2)} MB`,\n              hitCount: m.hitCount,\n              lastAccessed: new Date(m.lastAccessed).toISOString(),\n            })),\n            summary: {\n              totalCached: metrics.cachedModels,\n              totalMemory: `${(metrics.totalMemory / (1024 * 1024 * 1024)).toFixed(2)} GB`,\n              cacheHitRate: `${(metrics.cacheHitRate * 100).toFixed(1)}%`,\n              avgInferenceTime: `${metrics.avgInferenceTime.toFixed(2)} ms`,\n              quantizationJobs: metrics.quantizationJobs,\n              uptime: `${(metrics.uptime / 1000 / 60).toFixed(1)} minutes`,\n            },\n          },\n          timestamp: new Date().toISOString(),\n        }, null, 2);\n      } catch (error: any) {\n        return JSON.stringify({\n          success: false,\n          error: error.message,\n          timestamp: new Date().toISOString(),\n        }, null, 2);\n      }\n    },\n  });\n}\n"]}

package/dist/agentic-flow/src/mcp/fastmcp/tools/quic-tools.d.ts

@@ -0,0 +1,2 @@
+export declare function registerQUICTools(server: any): void;
+//# sourceMappingURL=quic-tools.d.ts.map

package/dist/agentic-flow/src/mcp/fastmcp/tools/quic-tools.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"quic-tools.d.ts","sourceRoot":"","sources":["../../../../../../src/mcp/fastmcp/tools/quic-tools.ts"],"names":[],"mappings":"AAqBA,wBAAgB,iBAAiB,CAAC,MAAM,EAAE,GAAG,GAAG,IAAI,CA8anD"}