Merge branch 'feat/llm_stream_check_python' into 'integration_2025-05-22_911100755970'

feat: [development task] waf runtime (1244130)

See merge request iaasng/volcengine-python-sdk!618

Author: BitsAdmin

volcenginesdkwafruntime/__init__.py

@@ -0,0 +1,17 @@
+# coding: utf-8
+
+# flake8: noqa
+
+"""
+    waf
+
+    No description provided (generated by Swagger Codegen https://github.com/swagger-api/swagger-codegen)  # noqa: E501
+
+    OpenAPI spec version: common-version
+    
+    Generated by: https://github.com/swagger-api/swagger-codegen.git
+"""
+
+from volcenginesdkwaf import *
+from volcenginesdkwafruntime.api.waf_runtime_api import  WAFRuntimeApi
+from volcenginesdkwafruntime.models.llm_stream_session import LLMStreamSession

volcenginesdkwafruntime/api/waf_runtime_api.py

@@ -0,0 +1,89 @@
+
+from volcenginesdkwaf import WAFApi, CheckLLMResponseStreamRequest
+from volcenginesdkwafruntime.models.llm_stream_session import LLMStreamSession
+
+global_llm_send_len = 10
+
+class WAFRuntimeApi(WAFApi):
+    """继承自 WAFApi 并重写 check_llm_response_stream 方法"""
+    def check_llm_response_stream(
+            self,
+            body: CheckLLMResponseStreamRequest,
+            **kwargs
+    ):
+        session: LLMStreamSession = kwargs.pop('session', None)
+        if session is None:
+            raise ValueError("session parameter is required")
+        """
+        重写父类方法，增加 session 参数，实现流式内容的合并与处理
+
+        Args:
+            body: 请求体，包含 content、use_stream 和 msg_id 等信息
+            session: LLM 流会话对象，用于记录流式内容和状态
+            **kwargs: 其他参数，包括 async_req 等
+
+        Returns:
+            CheckLLMResponseStreamResponse: 处理结果
+        """
+        # 1. 拼接内容到 session.stream_buf
+        if body.content:
+            session.append_stream_buf(body.content)
+
+        # 2. 处理 use_stream 为 2 或 没有msgid （第一次调用） 的情况（直接发送，不累计长度）
+        if (body.use_stream == 2
+                or not session.msg_id):
+            # 准备请求体
+            body.content = session.get_stream_buf()
+            body.msg_id = session.get_msg_id()
+            # 调用原始 API
+            # 同步调用并处理结果
+            resp = self.check_llm_response_stream_with_http_info(body, **kwargs)
+            if isinstance(resp, tuple):
+                response = resp[0]  # 获取元组的第一个元素
+            else:
+                response = resp
+
+            # 首次调用时(没有msgid) ，保存 msg_id 到 session
+            if (not session.msg_id) and response.msg_id:
+                session.set_msg_id(response.msg_id)
+
+            # 保存默认响应
+            session.set_default_body(response)
+
+            # 重置流缓冲区和发送长度
+            session.set_stream_send_len(0)
+
+            return response
+
+        # 3. 处理 use_stream 为其他值的情况（累计长度，超过阈值才发送）
+        else:
+            # 如果未发送长度超过 10 个字符，调用 API
+            if session.get_stream_send_len() > global_llm_send_len:
+                # 准备请求体，使用 session 中的完整流内容
+                body.content = session.get_stream_buf()
+                body.msg_id = session.get_msg_id()
+                # 调用原始 API
+                if kwargs.get('async_req'):
+                    return self.check_llm_response_stream_with_http_info(body, **kwargs)
+
+                # 同步调用并处理结果
+                resp = self.check_llm_response_stream_with_http_info(body, **kwargs)
+                if isinstance(resp, tuple) and len(resp) > 0:
+                    response = resp[0]  # 获取元组的第一个元素
+                else:
+                    response = resp
+
+                # 保存默认响应
+                session.set_default_body(response)
+
+                # 重置发送长度，保留完整流内容（因为可能还有后续数据）
+                session.set_stream_send_len(0)
+                return response
+            # 如果未发送长度不足 10 个字符，返回上次的结果
+            else:
+                default_body = session.get_default_body()
+                if default_body:
+                    return default_body
+                else:
+                    # 如果没有默认结果，调用一次 API（这种情况理论上不会发生，因为首次调用 use_stream 应为 0）
+                    return self.check_llm_response_stream(body, **kwargs)

volcenginesdkwafruntime/models/llm_stream_session.py

@@ -0,0 +1,49 @@
+from typing import Optional
+
+from volcenginesdkwaf import CheckLLMResponseStreamResponse
+
+
+class LLMStreamSession:
+    """对应 Java 中的 LLMStreamSession 类"""
+
+    def __init__(
+            self,
+            stream_buf: str = "",
+            stream_send_len: int = 0,
+            msg_id: str = "",
+            default_body: Optional[CheckLLMResponseStreamResponse] = None
+    ):
+        self.stream_buf = stream_buf
+        self.stream_send_len = stream_send_len
+        self.msg_id = msg_id
+        self.default_body = default_body
+
+    def get_stream_buf(self) -> str:
+        return self.stream_buf
+
+    def set_stream_buf(self, stream_buf: str) -> None:
+        self.stream_buf = stream_buf
+
+    def get_stream_send_len(self) -> int:
+        return self.stream_send_len
+
+    def set_stream_send_len(self, stream_send_len: int) -> None:
+        self.stream_send_len = stream_send_len
+
+    def get_msg_id(self) -> str:
+        return self.msg_id
+
+    def set_msg_id(self, msg_id: str) -> None:
+        self.msg_id = msg_id
+
+    def get_default_body(self) -> Optional[CheckLLMResponseStreamResponse]:
+        return self.default_body
+
+    def set_default_body(self, default_body: Optional[CheckLLMResponseStreamResponse]) -> None:
+        self.default_body = default_body
+
+    def append_stream_buf(self, text: Optional[str]) -> None:
+        """追加文本到流缓冲区并更新发送长度"""
+        if text is not None:
+            self.stream_buf += text
+            self.stream_send_len += len(text)