mm-qa-mcp 0.2.0__py3-none-any.whl

minimax_qa_mcp/src/get_weaviate_info/__init__.py

@@ -0,0 +1,6 @@
+"""
+coding:utf-8
+@Software: PyCharm
+@Time: 2025/4/27 18:18
+@Author: xingyun
+"""

minimax_qa_mcp/src/get_weaviate_info/get_weaviate_info.py

@@ -0,0 +1,298 @@
+"""
+coding:utf-8
+@Software: PyCharm
+@Time: 2025/4/27 18:19
+@Author: xingyun
+"""
+import json
+import logging
+import requests
+import os
+
+from minimax_qa_mcp.utils.utils import Utils
+
+# 设置日志
+logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
+logger = logging.getLogger('知识库检索')
+
+# ===== Weaviate参数配置 =====
+os.environ["WEAVIATE_GRPC_ENABLED"] = "False"  # 禁用gRPC，使用HTTP协议
+HTTP_HOST = Utils.get_conf("weaviate_url", "url")
+HTTP_PORT = Utils.get_conf("weaviate_url", "port")
+COLLECTION_NAME = "BusinessDocs_Auto"
+
+
+class GetWeaviateInfo:
+    """
+    获取Weaviate信息的类
+    """
+
+    def __init__(self, input_question, is_need_module: bool = False):
+        """
+        初始化Weaviate信息获取器
+        :param input_question: 用户输入
+        :param is_need_module: 是否调用模型分析
+        """
+        self.input_question = input_question
+        self.is_need_module = is_need_module
+        # 模型API配置
+        self.api_url = Utils.get_conf("generator_case_conf", "module_api_url")
+        self.timeout = 120  # 设置超时时间
+
+        # 懒加载Weaviate客户端
+        self.client = None
+        self._init_weaviate_client()
+
+    def _init_weaviate_client(self):
+        """
+        初始化Weaviate客户端，延迟导入以避免循环导入问题
+        使用Weaviate 3.x版本的API
+        """
+        try:
+            # 在方法内部延迟导入weaviate和其他需要的模块
+            import weaviate
+            from urllib.parse import urlparse
+
+            # 获取主机地址，确保不重复http://
+            http_host_value = HTTP_HOST
+            if http_host_value and (http_host_value.startswith('http://') or http_host_value.startswith('https://')):
+                # 如果配置已经包含协议，则提取主机部分
+                parsed_url = urlparse(http_host_value)
+                http_host_value = parsed_url.netloc
+                logger.info(f"从URL '{HTTP_HOST}'中提取主机部分: '{http_host_value}'")
+
+            logger.info(f"尝试连接到Weaviate服务器: {http_host_value}:{HTTP_PORT}")
+
+            # 使用Weaviate 3.x的客户端连接方式
+            self.client = weaviate.Client(f"http://{http_host_value}:{HTTP_PORT}")
+
+            # 检查连接
+            self.client.is_ready()
+            logger.info("Weaviate客户端初始化成功")
+        except Exception as e:
+            logger.error(f"Weaviate客户端初始化失败: {e}")
+            raise
+
+    def get_info(self):
+        """
+        获取Weaviate信息
+        :return: Weaviate信息字典
+        """
+        try:
+            # 使用3.x版本的方式获取元数据
+            meta = self.client.get_meta()
+
+            info = {
+                "version": meta.get("version", "未知"),
+                "schema": self.client.schema.get(),
+                "status": "已连接" if self.client.is_ready() else "未连接"
+            }
+            return info
+        except Exception as e:
+            logger.error(f"获取Weaviate信息失败: {e}")
+            return {
+                "version": "未知",
+                "schema": {},
+                "status": "错误",
+                "error": str(e)
+            }
+
+    def search_documents(self, limit=5):
+        """
+        基于相似度的语义搜索
+        
+        Args:
+            limit: 返回结果数量
+            
+        Returns:
+            包含查询结果的字典列表
+        """
+        try:
+            # 使用Weaviate 3.x版本的查询方式
+            properties = [
+                "title", "summary", "content", "category", "doc_id", "file_path", "doc_type",
+                "submitter", "business_tags"  # 添加其他可能的属性字段，但不包含时间字段
+            ]
+
+            # 创建查询构建器
+            query_builder = (
+                self.client.query
+                .get(COLLECTION_NAME, properties)
+                .with_limit(limit)
+                .with_near_text({"concepts": [self.input_question]})
+                .with_additional(["certainty"])
+            )
+
+            # 执行查询
+            results = query_builder.do()
+
+            # 处理结果
+            data = results.get("data", {}).get("Get", {}).get(COLLECTION_NAME, [])
+            logger.info(f"搜索到 {len(data)} 个结果")
+
+            processed_results = []
+            for item in data:
+                result = dict(item)
+
+                # 添加相似度分数
+                if "_additional" in item and "certainty" in item["_additional"]:
+                    result["similarity"] = item["_additional"]["certainty"]
+                    del result["_additional"]
+
+                processed_results.append(result)
+
+            return processed_results
+
+        except Exception as e:
+            logger.error(f"搜索失败: {e}")
+            return []
+
+    def call_model(self, content_list):
+        """调用模型API，添加了重试机制
+        
+        Args:
+            content_list: 待总结的内容列表
+            
+        Returns:
+            模型返回的结果
+        """
+        # 构建给模型的输入
+        prompt = f"请对以下文档内容进行总结，提取关键信息：\n\n"
+        for i, doc in enumerate(content_list):
+            prompt += f"文档{i + 1}：{doc.get('title', 'N/A')}\n"
+            prompt += f"内容：{doc.get('content', '')[:1000]}...\n\n"
+
+        prompt += "请提供一个简洁的总结，包含这些文档的核心要点。"
+
+        # 使用更简单的字符串替换处理
+        clean_params = prompt.replace('\\"', "'")  # 替换嵌套双引号为单引号
+        clean_params = clean_params.replace("\n", " ").strip()
+
+        payload = {
+            "scene": "qa_agent",
+            "params": {
+                "user_content": clean_params
+            }
+        }
+
+        # 使用线程锁保护日志
+        logger.info(f"==== 发送请求调用模型 ======")
+
+        try:
+            # 添加timeout参数，增加请求超时控制
+            response = requests.post(
+                self.api_url,
+                json=payload,
+                headers={'Content-Type': 'application/json'},
+                verify=False,
+                timeout=self.timeout
+            )
+
+            logger.info(f"API响应状态码: {response.status_code}")
+            logger.info(f"API响应内容: {response.text}")
+
+            # 检查响应状态
+            if response.status_code != 200:
+                logger.error(f"API请求失败，状态码: {response.status_code}")
+                return None
+
+            # 尝试解析JSON响应
+            try:
+                resp_json = response.json()
+
+                if 'response' in resp_json:
+                    # 解析二层JSON
+                    try:
+                        model_response = json.loads(resp_json['response'])
+
+                        # 从content中提取文本
+                        if 'content' in model_response and isinstance(model_response['content'], list):
+                            text_content = ""
+                            for item in model_response['content']:
+                                if item.get('type') == 'text':
+                                    text_content += item.get('text', '')
+                            return text_content
+                        return str(model_response)
+                    except Exception as e:
+                        logger.error(f"解析二层JSON失败: {e}")
+                        return resp_json['response']
+                return response.text
+            except Exception as e:
+                logger.error(f"解析JSON失败: {e}")
+                return response.text
+
+        except requests.RequestException as e:
+            logger.error(f"请求异常: {e}")
+            return None
+
+    def get_knowledge(self, limit=5):
+        """
+        获取知识库信息
+        
+        Args:
+            limit: 检索结果数量限制
+            
+        Returns:
+            检索结果或模型总结的JSON
+        """
+        # 首先检索文档
+        search_results = self.search_documents(limit=limit)
+
+        # 构建基本返回结果
+        result = {
+            "query": self.input_question,
+            "result_count": len(search_results),
+            "results": search_results
+        }
+
+        # 根据is_need_module决定是否调用模型总结
+        if self.is_need_module and search_results:
+            try:
+                summary = self.call_model(search_results)
+                if summary:
+                    result["model_summary"] = summary
+            except Exception as e:
+                logger.error(f"调用模型总结失败: {e}")
+                result["model_summary_error"] = str(e)
+
+        return result
+
+    def __del__(self):
+        """析构函数，确保在对象销毁时关闭连接"""
+        if hasattr(self, 'client') and self.client is not None:
+            try:
+                # 在3.x版本中可能没有显式的close方法
+                pass
+            except:
+                pass
+
+
+if __name__ == "__main__":
+    print("开始测试GetWeaviateInfo...")
+
+    try:
+        # 确保连接成功
+        print("实例化GetWeaviateInfo...")
+        get_weaviate_info = GetWeaviateInfo("海螺视频 图生视频 测试case", is_need_module=True)
+        print("已成功实例化GetWeaviateInfo")
+
+        # 测试获取基本信息
+        print("测试获取Weaviate基本信息...")
+        basic_info = get_weaviate_info.get_info()
+        print(f"Weaviate版本: {basic_info.get('version')}")
+        print(f"Weaviate状态: {basic_info.get('status')}")
+
+        # 测试知识库搜索
+        print("\n测试知识库搜索...")
+        search_results = get_weaviate_info.get_knowledge(limit=3)
+        print(f"查询: {search_results.get('query')}")
+        print(f"结果数量: {search_results.get('result_count')}")
+
+        # 打印每个结果的标题和相似度分数
+        for i, result in enumerate(search_results.get('results', [])):
+            print(f"结果 {i + 1}: {result.get('title')} (相似度: {result.get('similarity', 0):.3f})")
+
+        print("\n完整结果:")
+        print(search_results)
+    except Exception as e:
+        print(f"测试过程中发生错误: {e}")

minimax_qa_mcp/src/grafana/service.py

@@ -0,0 +1,104 @@
+import requests
+from datetime import datetime, timedelta, timezone
+from minimax_qa_mcp.utils.logger import logger
+from minimax_qa_mcp.utils.utils import Utils
+
+
+class GetFromGrafana:
+    def __init__(self, scene, psm="", from_time=None, to_time=None):
+
+        self.cluster = Utils.get_conf(f"{scene}_business_info", "grafana_cluster")
+        self.name_space = Utils.get_conf(f"{scene}_business_info", "grafana_name_space")
+        self.psm = psm.replace('.', '-')
+        # 拉取一天前的日志
+        self.url = Utils.get_conf('common', 'grafana_url')
+        if to_time is None and from_time is None:
+            to_formatted_time = datetime.now(timezone(timedelta(hours=8))).isoformat()
+            # 获取一天前的时间
+            # 格式化为 ISO 8601 格式的字符串，包含微秒和时区信息
+            from_formatted_time = (datetime.now(timezone(timedelta(days=1))) - timedelta(hours=1)).isoformat()
+
+            self.to_time = str(to_formatted_time)
+            self.from_time = str(from_formatted_time)
+        else:
+            self.to_time = to_time
+            self.from_time = from_time
+
+    def post_grafana(self, msgs: list):
+        query = f"_namespace_:\"{self.name_space}\" "
+        if self.psm:
+            query += f"and app:\"{self.psm}\" "
+        if len(msgs) > 0:
+            for msg in msgs:
+                query += f"and msg:\"{msg}\" "
+        data = {
+            "from": self.from_time,
+            "to": self.to_time,
+            "query": query,
+            "limit": 20,
+            "topic_name": f"_mlogs_{self.cluster}/{self.name_space}"
+        }
+        logger.info(f"grafana的入参为：{data}")
+        try:
+            grafana_resp = requests.post(self.url, json=data)
+            if grafana_resp.status_code == 200:
+                return grafana_resp.json()['data']['items']
+        except Exception as e:
+            logger.error(f'get grafana resp error, psm is:{self.psm}, method is: {msgs}, error is: {e}')
+            return []
+
+
+class GetApiFromGrafana:
+
+    def __init__(self, scene, psm):
+        self.psm = psm
+        self.scene = scene
+        self.url = Utils.get_conf('common', 'swing_url')
+        logger.info(f"GetApiFromGrafana init psm:{psm},scene:{scene}")
+
+    def get_method_list(self):
+        try:
+            res = requests.get(url=self.url + "/swing/api/fetch_api_by_psm?psm=" + str(self.psm.replace("-", ".")))
+            if res.status_code == 200:
+                return res.json()["data"]["apis"]
+        except Exception as e:
+            logger.error(f"get_method_list error: {e}")
+            return [e]
+
+    def get_top_qps(self):
+        try:
+            res = requests.get(url=self.url + "/swing/api/get_top_qps?scene=" + str(self.scene))
+            if self.psm is None or len(self.psm) <= 0:
+                res_data = res.json()["data"]
+                return {key: value for psm in res_data for key, value in res_data[psm].items()}
+            else:
+                return res.json()["data"][str(self.psm)]
+        except Exception as e:
+            logger.error(f"get_top_qps error: {e}")
+            return [e]
+
+    def get_need_method(self):
+        try:
+            qps_method_list = self.get_top_qps()
+            # 如果是明确增加某个psm的rpc方法，则关注psm idl，并返回接口定义
+            if "rpc" in self.scene and self.psm is not None and len(self.psm) > 0:
+                res_list = []
+                psm_method_list = self.get_method_list()
+                for method in psm_method_list:
+                    if method["method"] in list(qps_method_list.keys()):
+                        method["qps"] = qps_method_list[method["method"]]
+                        res_list.append(method)
+                return res_list
+            else:
+                return list([{"method": key, "qps": value} for key, value in qps_method_list.items()])
+
+        except Exception as e:
+            logger.error(f"get_top_qps error: {e}")
+            return [e]
+
+
+if __name__ == '__main__':
+    print("test")
+    # print(GetFromGrafana("xingye_test").post_grafana())
+    # print(GetApiFromGrafana("hailuo_video_us_http","").get_need_method())
+    print(GetApiFromGrafana("xingye_prod", "weaver-account-account").get_need_method())

minimax_qa_mcp/src/query_segments/__init__.py

@@ -0,0 +1,6 @@
+"""
+coding:utf-8
+@Software: PyCharm
+@Time: 2025/3/20 17:33
+@Author: xingyun
+"""