更新专业术语索引文件，优化意图识别逻辑，添加后缀项更新功能，调整重排序参数以提高相关性，同时修正文档中的描述信息。

data/wiki_extracted_nouns/主网电力建设计价通_nouns.json

@@ -9815,7 +9815,6 @@
   {
     "name": "措施项目一",
     "synonymous": [
-      "措施费"
     ],
     "description": "费用项目分类，包含特殊地区施工增加费等临时性工程费用；在建安预算中手动添加的不按费率计取的措施项目费，用于计算一笔性费用。"
   },

data/wiki_extracted_nouns/新型储能电站建设计价通C1_nouns.json

@@ -79,11 +79,6 @@
     "synonymous": [],
     "description": "施工机具使用费的价格依据"
   },
-  {
-    "name": "博微储能计价通C1",
-    "synonymous": [],
-    "description": "用于锂离子电池储能工程编制的软件"
-  },
   {
     "name": "试用账号",
     "synonymous": [],
@@ -108,7 +103,8 @@
     "name": "博微储能计价通C1软件",
     "synonymous": [
       "储能C1软件",
-      "储能软件"
+      "储能软件",
+      "博微储能计价通C1"
     ],
     "description": "用于锂离子电池储能工程编制的软件"
   },

data/wiki_extracted_nouns/西藏造价软件_nouns.json

@@ -235,12 +235,12 @@
     "description": "查看定额升级的具体规则"
   },
   {
-    "name": "西藏电力工程造价2016 V2.7.1.0",
+    "name": "西藏电力工程造价2016",
     "synonymous": [],
     "description": "西藏电力工程造价的旧版定额标准"
   },
   {
-    "name": "西藏配网造价2017 V2.12.3.0",
+    "name": "西藏配网造价2017",
     "synonymous": [],
     "description": "西藏配网造价的旧版定额标准"
   },
@@ -4207,11 +4207,6 @@
     "synonymous": [],
     "description": "用于批量设置或修改多个工程的模板、参数和物料信息"
   },
-  {
-    "name": "参数",
-    "synonymous": [],
-    "description": "工程设置中的各项参数"
-  },
   {
     "name": "地区范围",
     "synonymous": [],

rag2_0/demo/merge_nouns_with_llm.py

@@ -71,16 +71,16 @@ class TermMerger:
                     logging.warning(f"读取{file}失败: {e}")
         
         # 加载suffix_keywords.json文件
-        suffix_keywords_path = os.path.join(os.path.dirname(os.path.dirname(self.EXTRACTED_NOUNS_DIR)), 'data', 'nouns', 'suffix_keywords.json')
-        if os.path.exists(suffix_keywords_path):
-            try:
-                with open(suffix_keywords_path, 'r', encoding='utf-8') as f:
-                    suffix_terms = json.load(f)
-                    suffix_terms = [{"name": term["name"].upper(), "synonymous": "", "description": ""} for term in suffix_terms]
-                    all_terms.extend(suffix_terms)
-                    logging.info(f"加载{suffix_keywords_path}，共{len(suffix_terms)}条")
-            except Exception as e:
-                logging.warning(f"读取{suffix_keywords_path}失败: {e}")
+        # suffix_keywords_path = os.path.join(os.path.dirname(os.path.dirname(self.EXTRACTED_NOUNS_DIR)), 'data', 'nouns', 'suffix_keywords.json')
+        # if os.path.exists(suffix_keywords_path):
+        #     try:
+        #         with open(suffix_keywords_path, 'r', encoding='utf-8') as f:
+        #             suffix_terms = json.load(f)
+        #             suffix_terms = [{"name": term["name"].upper(), "synonymous": "", "description": ""} for term in suffix_terms]
+        #             all_terms.extend(suffix_terms)
+        #             logging.info(f"加载{suffix_keywords_path}，共{len(suffix_terms)}条")
+        #     except Exception as e:
+        #         logging.warning(f"读取{suffix_keywords_path}失败: {e}")
         
         return all_terms
 
@@ -154,6 +154,9 @@ class TermMerger:
                 for result in tqdm(executor.map(self.process_term, items_to_process), total=len(items_to_process)):
                     merged_terms.append(result)
 
+        # 4. 去重
+        merged_terms = self.deduplicate_synonymous_name(merged_terms)
+
         # 4. 保存合并结果
         os.makedirs(os.path.dirname(self.OUTPUT_PATH), exist_ok=True)
         with open(self.OUTPUT_PATH, 'w', encoding='utf-8') as f:
@@ -162,6 +165,26 @@ class TermMerger:
         
         return merged_terms
 
+    def deduplicate_synonymous_name(self, terms):
+        # 1. 删除name字段重复的条目
+        unique_names = set()
+        unique_data = []
+
+        for item in terms:
+            if item["name"] not in unique_names:
+                unique_names.add(item["name"])
+                unique_data.append(item)
+            # 如果重复，则跳过该条目
+
+        # 2. 如果A条目的某一个synonymou字段是B条目的name，则删除A条目中的对应的synonymou
+        name_set = {item["name"] for item in unique_data}
+
+        for item in unique_data:
+            # 过滤掉synonymous中与其他条目name重复的部分
+            filtered_synonymous = [syn for syn in item["synonymous"] if syn not in name_set]
+            item["synonymous"] = filtered_synonymous
+
+        return unique_data
 
 def main():
     """主函数，创建TermMerger实例并执行合并"""

rag2_0/dify/intent_recognition_api.py

@@ -63,7 +63,7 @@ def intent_recognize():
             for term in keywords["terms"]:
                 term_info = {
                     "名称": term["name"],
-                    # "同义词": ";".join(term["synonymous"]) if term["synonymous"] else "",
+                    # "同义词": ";".join(term["synonymous"]) if term["synonymous"] else [],
                     # "描述": term["description"]
                 }
                 term_details.append(term_info)

rag2_0/intent_recognition/DataModels.py

@@ -139,7 +139,7 @@ class SlotBase(BaseModel):
             if v in SOFTWARE_NAME_ALIAS_MAP:
                 return SOFTWARE_NAME_ALIAS_MAP[v].value
             
-            # 如果无法匹配，返回原值用于错误提示
+            # 如果无法匹配，返回原值
             return v
         
         return v
@@ -204,7 +204,7 @@ class ProfessionalConsultingSlots(SlotBase):
 
 # 2.2 数据问题
 class DataProblemSlots(SlotBase):
-    expense_type: str = Field(default="", description="费用类型")
+    expense_type: str = Field(default="", description="费用(数据)项、费用类型")
     operation_purpose: str = Field(default="", description="操作目的")
     software_name: Optional[str] = Field(default="", description="软件名称")
     project_type: Optional[str] = Field(default="",  description="工程类型")

rag2_0/intent_recognition/IntentRecognition.py

@@ -184,7 +184,7 @@ class IntentRecognizer:
         except Exception as e:
             raise RuntimeError(f"无法解析LLM关键词提取响应: {e}") from e
     
-    def _rerank_matched_terms(self, query_key: str, matched_terms: set, top_k: int = 2) -> List[Term]:
+    def _rerank_matched_terms(self, query_key: str, matched_terms: set, top_k: int = 2, rerank_score:float = 0.6) -> List[Term]:
         """
         对召回的专业术语进行重排序，按与用户查询的相关性排序
         
@@ -198,10 +198,14 @@ class IntentRecognizer:
         """
         if not matched_terms:
             return []
-            
+        
+        if len(matched_terms) <= top_k:
+            return list(matched_terms)
+
         try:
             # 将每个术语转换为可用于重排序的文本表示
-            term_texts = ["名称:" + term.name + "|" + "同义词:" + ";".join(term.synonymous) + "|" + "描述:" + term.description for term in matched_terms]
+            # term_texts = ["名称:" + term.name + "|" + "同义词:" + ";".join(term.synonymous) + "|" + "描述:" + term.description for term in matched_terms]
+            term_texts = ["名称:" + term.name + "|" + "同义词:" + ";".join(term.synonymous) for term in matched_terms]
             
             # 使用重排序模型
             xinference_reranker = SiliconFlowReRankerModel()
@@ -211,7 +215,7 @@ class IntentRecognizer:
             matched_terms_list = list(matched_terms)
             
             # 根据重排序结果获取排序后的术语列表
-            reranked_terms = [matched_terms_list[result["index"]] for result in rerank_results if result["score"] >= 0.6]
+            reranked_terms = [matched_terms_list[result["index"]] for result in rerank_results if result["score"] >= rerank_score]
             
             return reranked_terms
             
@@ -266,19 +270,21 @@ class IntentRecognizer:
         term_list = TermList(terms=list(matched_terms))
         return term_list, query_keys
     
-    def _rewrite_query(self, query: str, keywords: TermList, chat_history: List[Dict[str, str]] = None, context: str = "") -> QueryRewrite:
+    def _rewrite_query(self, query: str, keywords: TermList, query_keys:List[str], chat_history: List[Dict[str, str]] = None, context: str = "") -> QueryRewrite:
         """
         对用户问题进行改写
         
         Args:
             query: 用户原始问题
             keywords: 匹配到的关键词列表
-            
+            query_keys: 用户查询中提取的关键词列表
+
         Returns:
             改写结果
         """
         # 准备问题改写提示
-        terms_dict = [term.model_dump(exclude={"description"}) for term in keywords.terms] 
+        # terms_dict = [term.model_dump(exclude={"description"}) for term in keywords.terms] 
+        terms_dict = [term.model_dump() for term in keywords.terms] 
         keywords_str = json.dumps(terms_dict, ensure_ascii=False)
         query_rewrite_parser = PydanticOutputParser(pydantic_object=QueryRewrite)
         # formatted_prompt = query_rewrite_prompt.format(query=query, 
@@ -362,12 +368,13 @@ class IntentRecognizer:
         rewrite = self._rewrite_query(
             query=query, 
             keywords=keywords_terms,
+            query_keys=query_keys,
             chat_history=chat_history,
             context=conversation_context
         )
         
         # 步骤3: 进行意图识别和槽位填充
-        result = self._process_intent_and_slot(query, conversation_context, chat_history, previous_slots)
+        result = self._process_intent_and_slot(rewrite.rewrite, conversation_context, chat_history, previous_slots)
         result.update({"keywords": keywords_terms.model_dump(),
                               "rewrite": rewrite.model_dump(),
                               "query_keys": query_keys})

rag2_0/intent_recognition/Multi_PromptTemplates.py

@@ -159,7 +159,7 @@ graph TD
 
 ### 三、重构优先级
 1. **背景补充**
-   - 历史对话中确定的软件/地区必须继承（例："这软件"→"【配网工程D3】"）
+   - 历史对话中确定的背景信息需要保留（例："这软件"→"【配网工程D3】"）
  
 2. **术语处理**
    - 同义词转标准词 → 批量设置定额
@@ -190,9 +190,7 @@ graph TD
 
 
 intent_and_slot_prompt = """
-# 电力造价软件意图分类与槽位填充统一提示词
-
-你是一个专业的电力造价领域智能助手，负责对用户输入进行意图分类识别和关键信息槽位填充。
+# 你是一个专业的电力造价领域智能助手，负责对用户输入进行意图分类识别和关键信息槽位填充。
 
 {classification_info}
 
@@ -206,6 +204,7 @@ intent_and_slot_prompt = """
 - **技改检修工程计价通T1软件**：别名包括技改T1、T1软件、技改检修软件等
 - **技改检修清单计价通T1软件**：别名包括技改清单T1、T1清单软件、技改检修清单软件等
 - **主网电力建设计价通软件**：别名包括主网软件、电力建设软件、主网建设软件、博微电力建设计价通等
+不在上述软件之列的，使用用户输入中的软件名称
 
 ## 【任务要求】
 

rag2_0/intent_recognition/ProfessionalNounVector.py

@@ -127,7 +127,9 @@ class ProfessionalNounVectorizer:
             
             # 准备数据
             texts, metadatas = self._prepare_terms_for_faiss(deduplicated_terms)
-            
+            suffix_text,suffix_metadatas = self._updata_suffix_item()
+            texts.extend(suffix_text)
+            metadatas.extend(suffix_metadatas)
             # 创建索引
             faiss_index = self._create_index(texts, metadatas)
             
@@ -140,6 +142,30 @@ class ProfessionalNounVectorizer:
             logging.error(f"多文件向量化处理失败: {e}")
             return False
     
+    def _updata_suffix_item(self)->Tuple[List[str], List[Dict]] :
+        """
+        更新suffix_keywords.json文件
+        
+        Returns:
+            更新后的术语列表
+        """
+        # 加载suffix_keywords.json文件
+        text=[]
+        meta_info=[]
+
+        suffix_keywords_path = os.path.join(".", 'data', 'nouns', 'suffix_keywords.json')
+        if os.path.exists(suffix_keywords_path):
+            try:
+                with open(suffix_keywords_path, 'r', encoding='utf-8') as f:
+                    suffix_terms = json.load(f)
+                    suffix_terms = [{"name": term["name"].upper(), "synonymous": [], "description": ""} for term in suffix_terms]
+                    for cur_suffix in suffix_terms:
+                        text.append(cur_suffix["name"].upper())
+                        meta_info.append(cur_suffix)
+                    logging.info(f"加载{suffix_keywords_path}，共{len(suffix_terms)}条")
+            except Exception as e:
+                logging.warning(f"读取{suffix_keywords_path}失败: {e}")
+        return text,meta_info
     
     def _prepare_terms_for_faiss(self, terms: List[Dict[str, Any]]) -> Tuple[List[str], List[Dict]]:
         """
@@ -156,15 +182,9 @@ class ProfessionalNounVectorizer:
         
         for term in terms:
             name = term["name"]
-            texts.append(name.strip())
             synonymous = term.get("synonymous", [])
             description = term.get("description", "")
             # 记录元数据
-            metadatas.append({
-                "name": name,
-                "synonymous": synonymous,
-                "description": description
-            })
 
             if len(synonymous) > 0:
                 for synonyms_str in synonymous:
@@ -175,13 +195,21 @@ class ProfessionalNounVectorizer:
                         "description": description
                     })
             
-            if len(description) > 0:
-                texts.append(description.strip())
-                metadatas.append({
-                    "name": name,
-                    "synonymous": synonymous,
-                    "description": description
-                })
+            # texts.append(name.strip())
+            # metadatas.append({
+            #     "name": name,
+            #     "synonymous": synonymous,
+            #     "description": description
+            # })
+
+            # 不检索描述字段
+            # if len(description) > 0:
+            #     texts.append(description.strip())
+            #     metadatas.append({
+            #         "name": name,
+            #         "synonymous": synonymous,
+            #         "description": description
+            #     })
         
         return texts, metadatas
     

rag2_0/intent_recognition/PromptTemplates.py

@@ -34,7 +34,7 @@ classification_info="""【垂直领域分类】：
 4. 其他 -- 指与软件或电力造价专业无关的日常对话、问候、感慨、情绪表达等。
 
 【软件问题包括以下两类】：
-1. 软件功能：询问软件功能的使用、操作、位置等
+1. 软件功能：询问软件功能的使用、功能操作(调整)、功能位置、如何设置、如何转换等
 2. 故障排查：软件运行异常、软件报错、软件显示错误等
 
 【业务问题包括以下两类】：

rag2_0/tool/APIKeyManager.py

@@ -18,6 +18,11 @@ import requests
 # sk-dvbaktabkdwdpjgxyoozlwnejosjyhdgqwllfeborqahndxs
 
 API_KEY_LIST=[
+"sk-kvgfuqeqvpmfsccykyoohheshclcrtvjlnewratvrjpkpbkc",
+"sk-zhnbqnpuumuuvegnvbgoggxafpukbzchpgrugpkobiwkzsar",
+"sk-kzhxlqvqcxlnbdgnpalqnzumkmspepkttkgbophnkqanainw",
+"sk-bzttugqtlskrvguvhckwamdssvgmgnrqpsialpdbskfsyyak",
+"sk-tovmogiablsoeabwgqyvevpcfichyjpuzqdymmvksspdrtqt",
 "sk-wqdpapdkisovziexgcyxvumpwzbjnhqbxvcqcspzctjhyhjk",
 "sk-bbntrnifrtdzhhgrtlrhvwbnaysuszviemshdakxonnnymnb",
 "sk-vmpnwjxersrwybmfhfxgsvbmhsmpjldxseiyxovnysrlbuzi",
@@ -91,12 +96,6 @@ API_KEY_LIST=[
 "sk-jrdzerhmvrtvzawkksowbgkggkubwfquplmrxbdhespqgtis",
 "sk-jjbpnkbeupsxyclcivbhizcfpfjrppddunbqynyjkqhtmpwu",
 "sk-oqehupcveovkjqqtxypqyifidcdissuyehwrkdwgruoyjkpq",
-"sk-orhfntzrbpmpavybcjyylofxncdvufdmvlznofmhxmnjymjl",
-"sk-kvgfuqeqvpmfsccykyoohheshclcrtvjlnewratvrjpkpbkc",
-"sk-zhnbqnpuumuuvegnvbgoggxafpukbzchpgrugpkobiwkzsar",
-"sk-kzhxlqvqcxlnbdgnpalqnzumkmspepkttkgbophnkqanainw",
-"sk-bzttugqtlskrvguvhckwamdssvgmgnrqpsialpdbskfsyyak",
-"sk-tovmogiablsoeabwgqyvevpcfichyjpuzqdymmvksspdrtqt",
 ]
 
 class APIKeyManager: