feat: Phase 7.4.3 - Semantic classifier with material-aware search

Major improvements to HS code classification accuracy:

Backend - New Services:
- Add semantic-classifier.service.ts for embedding-based classification
- Add query-term-analyzer.service.ts for intelligent query parsing
- Add reranker.service.ts for function-over-material ranking
- Add input-specificity.service.ts for query complexity analysis
- Add attribute-classifier.service.ts for attribute-based classification

Backend - Query Analysis:
- Parse queries into product, variety, processing, material, packaging terms
- Use fullQueryWithoutPackaging to retain material terms (cotton, silicone)
- Exclude only packaging terms (1kg, bags, boxes) from search
- Add LED, LCD, bluetooth, wifi and other electronic variety terms

Backend - Classification Improvements:
- Implement function-over-material reranking (apparel > fabric)
- Add raw material code penalty (3910 silicone oil deprioritized)
- Add dominant chapter detection for direct classification
- Fix coffee classification (0901 not 0902 tea)
- Fix cotton t-shirts classification (6109 direct, not questions)
- Fix LED light bulbs classification (8539 direct, not questions)

Frontend - New Wizard UI:
- Add comprehensive wizard component system
- Add history page with filtering and stats
- Add classification tree visualization
- Add loading states with educational facts
- Add error handling with suggestions

Testing:
- Add comprehensive validation test suite (64 test cases)
- Achieve 67% chapter accuracy, 0% error rate
- Textiles category improved from 33% to 67%
- Metals category improved from 60% to 100%

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Author: AryanBV

backend/src/data/chapter-triggers.json

@@ -662,6 +662,28 @@
         "keywords": ["footwear", "shoes", "boots", "sandals", "sneakers", "slippers"],
         "forceChapter": "64",
         "reason": "Footwear is always Ch.64 regardless of material"
+      },
+      {
+        "keywords": [
+          "LED light", "LED bulb", "LED lamp", "LED tube", "LED strip", "LED panel",
+          "light bulb", "light bulbs", "electric lamp", "electric lamps",
+          "discharge lamp", "fluorescent lamp", "halogen lamp", "incandescent lamp",
+          "CFL bulb", "compact fluorescent", "filament lamp", "filament bulb",
+          "household lamp", "household bulb", "lighting bulb", "lighting lamp"
+        ],
+        "forceChapter": "85",
+        "reason": "Electric lamps and light bulbs are Ch.85 (electrical equipment), specifically 8539 for lamps"
+      },
+      {
+        "keywords": [
+          "luminaire", "luminaires", "lighting fixture", "lighting fixtures",
+          "ceiling light", "pendant light", "wall light", "chandelier",
+          "spotlight", "spotlights", "floodlight", "floodlights",
+          "table lamp", "floor lamp", "desk lamp", "reading lamp",
+          "street light", "street lamp", "garden light", "outdoor light"
+        ],
+        "forceChapter": "94",
+        "reason": "Luminaires and lighting fixtures (complete units) are Ch.94 (9405), not just the bulb"
       }
     ]
   },

backend/src/data/elimination-rules.json

@@ -225,5 +225,51 @@
       "exclude": ["flavored", "flavoured", "with flavor", "with flavour"],
       "description": "When user specifies unflavored, exclude flavored codes"
     }
+  },
+  "lampTypeExclusions": {
+    "led": {
+      "include": ["led", "led light", "led bulb", "led lamp", "led tube", "led panel", "led strip", "light-emitting diode", "light emitting diode"],
+      "exclude": ["filament", "tungsten", "halogen", "incandescent", "discharge", "fluorescent", "arc lamp", "arc-lamp", "sealed beam", "infrared", "infra-red", "ultraviolet", "mercury", "sodium", "neon"],
+      "description": "When user specifies LED, exclude other lamp types (filament, discharge, UV/IR)"
+    },
+    "filament": {
+      "include": ["filament", "filament lamp", "filament bulb", "incandescent", "tungsten", "halogen", "halogen lamp", "halogen bulb"],
+      "exclude": ["led", "light-emitting diode", "discharge", "fluorescent", "arc", "mercury", "sodium", "neon"],
+      "description": "When user specifies filament/halogen/incandescent, exclude LED and discharge lamps"
+    },
+    "halogen": {
+      "include": ["halogen", "halogen lamp", "halogen bulb", "tungsten halogen"],
+      "exclude": ["led", "light-emitting diode", "discharge", "fluorescent", "arc", "standard filament"],
+      "description": "When user specifies halogen, exclude LED and other lamp types"
+    },
+    "discharge": {
+      "include": ["discharge", "discharge lamp", "fluorescent", "cfl", "compact fluorescent", "mercury", "sodium", "metal halide", "arc lamp", "arc-lamp"],
+      "exclude": ["led", "light-emitting diode", "filament", "tungsten", "halogen", "incandescent"],
+      "description": "When user specifies discharge/fluorescent, exclude LED and filament lamps"
+    },
+    "fluorescent": {
+      "include": ["fluorescent", "fluorescent lamp", "fluorescent tube", "cfl", "compact fluorescent"],
+      "exclude": ["led", "light-emitting diode", "filament", "tungsten", "halogen", "incandescent"],
+      "description": "When user specifies fluorescent, exclude LED and filament lamps"
+    },
+    "uv": {
+      "include": ["uv", "uv lamp", "ultraviolet", "ultra-violet", "uv light", "blacklight", "black light"],
+      "exclude": ["led", "light-emitting diode", "filament", "halogen", "visible light", "household"],
+      "description": "When user specifies UV, exclude visible light lamp types"
+    },
+    "infrared": {
+      "include": ["infrared", "infra-red", "ir lamp", "heat lamp", "ir light"],
+      "exclude": ["led", "light-emitting diode", "filament", "visible light", "household", "ultraviolet"],
+      "description": "When user specifies infrared/IR, exclude visible light lamp types"
+    }
+  },
+  "semanticAliases": {
+    "description": "Semantic aliases for keyword matching - maps user terms to description terms",
+    "led": ["led", "light-emitting diode", "light emitting diode", "led light", "led lamp"],
+    "filament": ["filament", "tungsten", "incandescent"],
+    "halogen": ["halogen", "tungsten halogen"],
+    "discharge": ["discharge", "fluorescent", "mercury vapor", "sodium vapor"],
+    "uv": ["ultraviolet", "ultra-violet", "uv"],
+    "infrared": ["infrared", "infra-red", "ir"]
   }
 }

backend/src/routes/classify-conversational.routes.ts

@@ -7,8 +7,10 @@
 
 import { Router, Request, Response } from 'express';
 import { logger } from '../utils/logger';
+// PHASE 7.1: Use Semantic Search as PRIMARY classifier
+import { classifyWithSemanticSearch } from '../services/semantic-classifier.service';
+// Keep legacy imports for backwards compatibility on other endpoints
 import {
-  classifyWithLLMNavigator,
   getConversation,
   abandonConversation,
   getConversationStats,
@@ -42,29 +44,21 @@ router.post('/', async (req: Request, res: Response): Promise<void> => {
   try {
     const { productDescription, sessionId, conversationId, answers } = req.body;
 
-    // Validation
-    if (!sessionId) {
-      res.status(400).json({
-        success: false,
-        error: 'Session ID is required',
-        timestamp: new Date().toISOString()
-      });
-      return;
-    }
-
+    // Validation - sessionId optional for semantic classifier
     if (!conversationId && !productDescription?.trim()) {
       res.status(400).json({
         success: false,
+        responseType: 'error',
         error: 'Product description is required for new conversations',
         timestamp: new Date().toISOString()
       });
       return;
     }
 
-    logger.info(`[CLASSIFY] ${conversationId ? 'Continuing' : 'New'} conversation`);
+    logger.info(`[CLASSIFY-SEMANTIC] ${conversationId ? 'Continuing' : 'New'} - "${productDescription?.substring(0, 50)}..."`);
 
-    // Call the LLM navigator classifier
-    const result = await classifyWithLLMNavigator({
+    // PHASE 7.1: Use Semantic Search as PRIMARY classifier
+    const result = await classifyWithSemanticSearch({
       productDescription: productDescription || '',
       sessionId,
       conversationId,
@@ -73,7 +67,7 @@ router.post('/', async (req: Request, res: Response): Promise<void> => {
 
     // Log performance
     const responseTime = Date.now() - startTime;
-    logger.info(`[CLASSIFY] Response in ${responseTime}ms - type: ${result.responseType}`);
+    logger.info(`[CLASSIFY-SEMANTIC] Response in ${responseTime}ms - type: ${result.responseType}`);
 
     if (result.success) {
       res.status(200).json(result);
@@ -83,43 +77,36 @@ router.post('/', async (req: Request, res: Response): Promise<void> => {
 
   } catch (error) {
     const errorMsg = error instanceof Error ? error.message : String(error);
-    logger.error(`[CLASSIFY] Error: ${errorMsg}`);
+    logger.error(`[CLASSIFY-SEMANTIC] Error: ${errorMsg}`);
 
     res.status(500).json({
       success: false,
+      responseType: 'error',
       error: 'Internal server error during classification',
       timestamp: new Date().toISOString()
     });
   }
 });
 
-// Legacy /llm endpoint - redirects to main endpoint for backward compatibility
+// Legacy /llm endpoint - now also uses semantic classifier
 router.post('/llm', async (req: Request, res: Response): Promise<void> => {
-  // Just forward to main handler
   const startTime = Date.now();
 
   try {
     const { productDescription, sessionId, conversationId, answers } = req.body;
 
-    if (!sessionId) {
-      res.status(400).json({
-        success: false,
-        error: 'Session ID is required',
-        timestamp: new Date().toISOString()
-      });
-      return;
-    }
-
     if (!conversationId && !productDescription?.trim()) {
       res.status(400).json({
         success: false,
+        responseType: 'error',
         error: 'Product description is required for new conversations',
         timestamp: new Date().toISOString()
       });
       return;
     }
 
-    const result = await classifyWithLLMNavigator({
+    // PHASE 7.1: Use Semantic Search for /llm endpoint too
+    const result = await classifyWithSemanticSearch({
       productDescription: productDescription || '',
       sessionId,
       conversationId,
@@ -141,6 +128,7 @@ router.post('/llm', async (req: Request, res: Response): Promise<void> => {
 
     res.status(500).json({
       success: false,
+      responseType: 'error',
       error: 'Internal server error during LLM classification',
       timestamp: new Date().toISOString()
     });