File size: 20,976 Bytes
29555e6
 
 
 
 
 
 
 
 
 
 
 
 
6fd3014
29555e6
 
 
 
6fd3014
29555e6
 
6fd3014
 
29555e6
 
6fd3014
29555e6
 
 
6fd3014
 
 
 
 
 
29555e6
 
 
6fd3014
29555e6
 
 
 
 
 
fa76df4
29555e6
 
5c03fe8
29555e6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6fd3014
29555e6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6fd3014
29555e6
 
 
 
6fd3014
 
 
29555e6
 
 
6fd3014
29555e6
 
 
6fd3014
29555e6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6fd3014
 
29555e6
 
 
 
 
 
 
 
 
 
 
 
 
6fd3014
29555e6
 
 
 
6fd3014
29555e6
 
 
 
 
6fd3014
29555e6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6fd3014
29555e6
6fd3014
29555e6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6fd3014
29555e6
 
 
6fd3014
 
29555e6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6fd3014
29555e6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6fd3014
29555e6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6fd3014
29555e6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6fd3014
29555e6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6fd3014
29555e6
 
 
 
6fd3014
29555e6
 
 
6fd3014
29555e6
6fd3014
29555e6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6fd3014
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
import os
import sys
from typing import List, Dict, Tuple, Generator, Set
from openai import OpenAI
from vectorize_knowledge_base import KnowledgeBaseVectorizer
import json
from datetime import datetime
import re


class RAGLearningAssistant:
    def __init__(self, api_key: str, model: str = "gpt-4.1-nano-2025-04-14", vector_db_path: str = ""):
        """
        初始化RAG学习助手(适配学生Space)
        
        Args:
            api_key: OpenAI API密钥(必需)
            model: 使用的模型名称
            vector_db_path: 向量数据库所在目录路径(数据存储仓库的本地目录)
        """
        self.client = OpenAI(api_key=api_key)
        
        # 使用修改后的KnowledgeBaseVectorizer,指定vector_db_dir
        self.vectorizer = KnowledgeBaseVectorizer(
            api_key=api_key,
            vector_db_dir=vector_db_path  # 传递数据存储仓库的本地目录
        )
        
        # 预加载向量数据库到缓存
        print("[RAGLearningAssistant] Preloading vector database...")
        load_result = self.vectorizer.load_vector_database()
        if load_result[0] is not None:
            print(f"[RAGLearningAssistant] Vector database loaded successfully")
        else:
            print(f"[RAGLearningAssistant] Warning: Failed to load vector database")
        
        # 模型配置
        self.model = model
        self.temperature = 0.1
        self.max_tokens = 2000
        
        # 系统提示词
        self.system_prompt = """You are a helpful learning assistant specializing in road engineering.
You have access to a knowledge base of course materials. When answering questions:
1. Stick to the provided context from the knowledge base.
2. At the end of your response, provide students the 'title' & 'from' fields of the chunks that were used to answer the question.
3. If the knowledge base doesn't contain relevant information, say so. Students can go to the teaching team for further assistance.

In the response, enclose mathematical formulas and parameters for proper Markdown rendering.
Bold key words if applicable.
"""
        
        # 查询重写的系统提示词 - 改进版本
        self.rewrite_prompt = """You are a query rewriting assistant. Your task is to provide a summary of the conversation history and then rewrite user queries based on conversation history to make them more clear and complete.

Please format your response as follows:
SUMMARY: [Brief summary of the conversation context. Include key points, user intent, and any relevant details]
REWRITTEN_QUERY: [The rewritten query that incorporates context]

Rules:
1. If there's relevant context from previous messages, incorporate it into the rewritten query
2. Make implicit references explicit
3. Maintain the original intent while adding clarity
4. If the query is already clear and complete, keep it as is
5. Always provide both SUMMARY and REWRITTEN_QUERY sections"""
        
        # 实体提取的系统提示词
        self.entity_extraction_prompt = """You are an expert in road engineering. Extract key entities from the given query.
Focus on:
1. Technical terms and jargon specific to road engineering
2. Formulas, equations, or mathematical concepts
3. Parameters, specifications, or measurements
4. Standards, methods, or procedures
5. Materials, equipment, or structures

Return the entities as a JSON array of strings. Only include the most important and specific entities."""
        
        # 对话历史
        self.conversation_history = []
        

    def rewrite_query(self, query: str) -> Tuple[str, str]:
        """
        基于对话历史重写查询,并返回历史总结
        
        Args:
            query: 原始查询
            
        Returns:
            (历史总结, 重写后的查询)
        """
        # 构建消息
        messages = [
            {"role": "system", "content": self.rewrite_prompt}
        ]
        
        # 添加对话历史上下文
        if self.conversation_history:
            context = "Previous conversation:\n"
            for msg in self.conversation_history[-6:]:  # 最近3轮对话
                role = "User" if msg["role"] == "user" else "Assistant"
                # 截取前200个字符避免过长
                content = msg["content"][:200] + "..." if len(msg["content"]) > 200 else msg["content"]
                context += f"{role}: {content}\n"
            
            messages.append({
                "role": "user", 
                "content": f"{context}\n\nCurrent query: {query}\n\nPlease provide summary and rewritten query following the specified format:"
            })
        else:
            # 没有历史时也要按格式返回
            messages.append({
                "role": "user",
                "content": f"Current query: {query}\n\nPlease provide summary and rewritten query following the specified format:"
            })
        
        try:
            response = self.client.chat.completions.create(
                model=self.model,
                messages=messages,
                temperature=0.1,  # 低温度确保一致性
                max_tokens=2000
            )
            
            content = response.choices[0].message.content.strip()
            
            # 改进的解析逻辑
            summary = ""
            rewritten = query  # 默认值
            
            # 使用正则表达式提取SUMMARY和REWRITTEN_QUERY
            summary_match = re.search(r'SUMMARY:\s*(.*?)(?=REWRITTEN_QUERY:|$)', content, re.DOTALL | re.IGNORECASE)
            rewritten_match = re.search(r'REWRITTEN_QUERY:\s*(.*?)$', content, re.DOTALL | re.IGNORECASE)

            
            if summary_match:
                summary = summary_match.group(1).strip()
            
            if rewritten_match:
                rewritten = rewritten_match.group(1).strip()
            
            # 备用解析方法 - 如果正则表达式失败
            if not summary and not rewritten_match:
                lines = content.split('\n')
                current_section = None
                summary_lines = []
                rewritten_lines = []
                
                for line in lines:
                    line = line.strip()
                    if line.upper().startswith("SUMMARY"):
                        current_section = "summary"
                        # 提取SUMMARY:后面的内容
                        summary_part = line[line.upper().find("SUMMARY"):].replace("SUMMARY:", "").strip()
                        if summary_part:
                            summary_lines.append(summary_part)
                    elif line.upper().startswith("REWRITTEN_QUERY") or line.upper().startswith("REWRITTEN QUERY"):
                        current_section = "rewritten"
                        # 提取REWRITTEN_QUERY:后面的内容
                        rewritten_part = re.sub(r'^REWRITTEN[_\s]*QUERY[:\s]*', '', line, flags=re.IGNORECASE).strip()
                        if rewritten_part:
                            rewritten_lines.append(rewritten_part)
                    elif current_section == "summary" and line:
                        summary_lines.append(line)
                    elif current_section == "rewritten" and line:
                        rewritten_lines.append(line)
                
                if summary_lines:
                    summary = " ".join(summary_lines)
                if rewritten_lines:
                    rewritten = " ".join(rewritten_lines)
            
            # 如果仍然没有获得有效结果,使用更简单的方法
            if not summary and self.conversation_history:
                summary = "Continue previous discussion"
            
            if not rewritten or rewritten == query:
                rewritten = query
            
            print(f"[rewrite_query] Raw query: {query}")
            print(f"[rewrite_query] Chat history summary: {summary}")
            print(f"[rewrite_query] Rewritten query: {rewritten}")
            return summary, rewritten
            
        except Exception as e:
            print(f"[rewrite_query] Query rewriting failed: {e}")
            # 生成简单的历史总结作为备用
            simple_summary = ""
            if self.conversation_history:
                simple_summary = "Based on previous conversation content"
            return simple_summary, query  # 失败时返回简单总结和原始查询
    
    def extract_entities(self, original_query: str, summary: str, rewritten_query: str) -> List[str]:
        """
        从原始查询、历史总结和重写查询中提取关键实体(专业术语、公式、参数等)
        
        Args:
            original_query: 原始用户查询
            summary: 历史总结
            rewritten_query: 重写后的查询文本
            
        Returns:
            提取的实体列表
        """
        # 合并所有文本作为实体提取的输入
        text_parts = []
        
        # 添加原始查询
        if original_query:
            text_parts.append(f"Original query: {original_query}")
        
        # 添加历史总结
        if summary:
            text_parts.append(f"Context summary: {summary}")
        
        # 添加重写查询
        if rewritten_query and rewritten_query != original_query:
            text_parts.append(f"Rewritten query: {rewritten_query}")
        
        combined_text = " | ".join(text_parts)
        
        messages = [
            {"role": "system", "content": self.entity_extraction_prompt},
            {"role": "user", "content": f"Text to extract entities from: {combined_text}\n\nExtract entities as JSON array:"}
        ]
        
        try:
            response = self.client.chat.completions.create(
                model=self.model,
                messages=messages,
                temperature=self.temperature,
                max_tokens=self.max_tokens
            )
            
            content = response.choices[0].message.content.strip()
            
            # 尝试解析JSON
            try:
                # 提取JSON数组(处理可能的markdown格式)
                json_match = re.search(r'\[.*?\]', content, re.DOTALL)
                if json_match:
                    entities = json.loads(json_match.group())
                else:
                    entities = json.loads(content)
                
                print(f"[extract_entities] Extracted entities: {entities}")
                return entities
                
            except json.JSONDecodeError:
                # 如果JSON解析失败,尝试简单的文本处理
                print(f"[extract_entities] JSON parsing failed, using backup method")
                # 查找引号中的内容
                entities = re.findall(r'"([^"]+)"', content)
                return entities if entities else self.simple_entity_extraction(combined_text)
                
        except Exception as e:
            print(f"[extract_entities] Entity extraction failed: {e}")
            # 失败时使用简单的关键词提取
            return self.simple_entity_extraction(combined_text)
    
    def simple_entity_extraction(self, query: str) -> List[str]:
        """
        简单的实体提取备用方法
        
        Args:
            query: 查询文本
            
        Returns:
            提取的关键词列表
        """
        # 移除常见停用词
        stop_words = {'the', 'a', 'an', 'and', 'or', 'but', 'in', 'on', 'at', 'to', 'for', 
                     'of', 'with', 'by', 'from', 'what', 'how', 'when', 'where', 'why',
                     'is', 'are', 'was', 'were', 'been', 'be', 'have', 'has', 'had',
                     'original', 'query', 'context', 'summary', 'rewritten'}  # 添加新的停用词
        
        # 分词并过滤
        words = query.lower().split()
        entities = [w for w in words if w not in stop_words and len(w) > 2]
        
        # 查找可能的专业术语(包含大写字母或数字)
        special_terms = re.findall(r'\b[A-Z][a-zA-Z]*\b|\b\w*\d+\w*\b', query)
        entities.extend(special_terms)
        
        # 去重并返回
        return list(set(entities))[:10]  # 最多返回5个实体
    
    def enhanced_search(self, query: str, top_k: int = 5) -> Tuple[str, str, List[str], List[Tuple[Dict, float, Dict]]]:
        """
        增强搜索:重写查询 -> 提取实体 -> 基于实体搜索(优化版本)
        
        Args:
            query: 原始查询
            top_k: 返回的结果数
            
        Returns:
            (历史总结, 重写后的查询, 提取的实体, 搜索结果)
        """
        # 1. 重写查询并获取历史总结
        summary, rewritten_query = self.rewrite_query(query)
        
        # 2. 基于原始查询、总结和重写查询提取实体
        entities = self.extract_entities(query, summary, rewritten_query)
        
        # 3. 基于实体搜索(使用优化的批量搜索)
        if entities:
            # 使用优化的批量搜索方法
            search_results = self.vectorizer.search_with_entities_optimized(entities, top_k)
        else:
            # 如果没有提取到实体,使用重写后的查询进行搜索
            print("[enhanced_search] No entities extracted, using full query search")
            search_results = self.vectorizer.search_similar(
                rewritten_query, 
                top_k=top_k,
                title_weight=0.2,
                content_weight=0.5,
                full_weight=0.3
            )
        
        return summary, rewritten_query, entities, search_results         

    def format_context(self, search_results: List[Tuple[Dict, float, Dict]]) -> str:
        """
        格式化搜索结果作为上下文
        
        Args:
            search_results: 搜索结果列表
            
        Returns:
            格式化的上下文字符串
        """
        if not search_results:
            return ""
        
        context_parts = []
        for i, result in enumerate(search_results, 1):
            entry, combined_score, details = result
            # 只显示 title, source, content,不显示 id
            context_parts.append(
                f"Title: {entry['title']}\n"
                f"From: {entry['source']}\n"
                f"Content: {entry['content']}\n"
            )
        
        return "RELEVANT KNOWLEDGE BASE CONTENT:\n" + "\n---\n".join(context_parts)
    
    def build_messages(self, query: str, context: str) -> List[Dict[str, str]]:
        """
        构建消息列表,包含系统提示、上下文和用户查询
        
        Args:
            query: 用户查询
            context: 知识库上下文
            
        Returns:
            消息列表
        """
        messages = [
            {"role": "system", "content": self.system_prompt}
        ]
        
        # 添加对话历史(保留最近5轮对话)
        for msg in self.conversation_history[-10:]:  # 最多保留5轮对话(10条消息)
            messages.append(msg)
        
        # 构建用户消息,包含上下文
        user_message = query
        if context:
            user_message = f"{context}\n\nUSER QUESTION: {query}"
        
        messages.append({"role": "user", "content": user_message})
        
        return messages
    
    def generate_response_stream(self, query: str) -> Generator[str, None, None]:
        """
        生成流式响应
        
        Args:
            query: 用户查询
            
        Yields:
            响应文本片段
        """
        # 1. 增强搜索(现在使用优化版本)
        print("[generate_response_stream] Processing query...")
        summary, rewritten_query, entities, search_results = self.enhanced_search(query)
        
        # 2. 格式化上下文
        context = self.format_context(search_results)
        
        # 3. 构建消息(使用原始查询,但包含基于实体搜索的上下文)
        messages = self.build_messages(query, context)
        
        # 4. 调用OpenAI API进行流式生成
        try:
            stream = self.client.chat.completions.create(
                model=self.model,
                messages=messages,
                temperature=self.temperature,
                max_tokens=self.max_tokens,
                stream=True
            )
            
            # 收集完整响应用于保存到历史
            full_response = ""
            
            # 首先返回搜索信息
            search_info = f"\n**Query Analysis:**\n"
            search_info += f"- Query: {query}\n"
            if summary:
                search_info += f"- Summary of history: {summary}\n"
            if rewritten_query != query:
                search_info += f"- Rewritten query: {rewritten_query}\n"
            search_info += f"- Key entities: {', '.join(entities) if entities else 'No specific entities extracted'}\n"
            
            if search_results:
                search_info += f"\n**Relevant Sources:**\n"
                for result in search_results:
                    entry, combined_score, details = result
                    # 给用户显示时包含 ID 和相关度分数
                    search_info += f"- [{entry['id']}] {entry['title']} (Relevance: {combined_score:.3f})\n"
                search_info += "\n**Response:**\n"
            else:
                search_info += "\n**Response:** (No relevant knowledge base content found, answering based on general knowledge)\n"
            
            # 添加缓存信息(调试用)
            cache_info = self.vectorizer.get_cache_info()
            if cache_info['is_cached']:
                search_info += f"💡 Vector database cached with {cache_info['cache_size']} entries\n\n"

            yield search_info
            
            # 流式返回生成的内容
            for chunk in stream:
                if chunk.choices[0].delta.content is not None:
                    content = chunk.choices[0].delta.content
                    full_response += content
                    yield content
            
            # 保存到对话历史
            self.conversation_history.append({"role": "user", "content": query})
            self.conversation_history.append({"role": "assistant", "content": full_response})
            
        except Exception as e:
            yield f"\n\nError: Problem occurred while generating response - {str(e)}"
    
    def generate_response(self, query: str) -> str:
        """
        生成完整响应(非流式)
        
        Args:
            query: 用户查询
            
        Returns:
            完整的响应文本
        """
        response_parts = []
        for part in self.generate_response_stream(query):
            response_parts.append(part)
        return "".join(response_parts)
    
    def clear_history(self):
        """清除对话历史"""
        self.conversation_history = []
        print("[clear_history] Conversation history cleared")
    
    def clear_vector_cache(self):
        """清除向量数据库缓存"""
        self.vectorizer.clear_cache()
        print("[clear_vector_cache] Vector database cache cleared")
    
    def reload_vector_database(self):
        """重新加载向量数据库"""
        print("[reload_vector_database] Reloading vector database...")
        self.vectorizer.load_vector_database(force_reload=True)
        print("[reload_vector_database] Vector database reload completed")
    
    def get_system_status(self) -> Dict:
        """
        获取系统状态信息
        
        Returns:
            系统状态字典
        """
        cache_info = self.vectorizer.get_cache_info()
        return {
            'model': self.model,
            'conversation_turns': len(self.conversation_history) // 2,
            'vector_cache': cache_info,
            'last_update': datetime.now().isoformat()
        }
    
    def save_conversation(self, filepath: str = None):
        """
        保存对话历史
        
        Args:
            filepath: 保存路径
        """
        if filepath is None:
            timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
            filepath = f"conversation_{timestamp}.json"
        
        conversation_data = {
            "timestamp": datetime.now().isoformat(),
            "model": self.model,
            "system_status": self.get_system_status(),
            "history": self.conversation_history
        }
        
        with open(filepath, 'w', encoding='utf-8') as f:
            json.dump(conversation_data, f, ensure_ascii=False, indent=2)
        
        print(f"[save_conversation] Conversation saved to: {filepath}")