springbootAI 2.3.10__tar.gz → 2.3.11__tar.gz

{springbootai-2.3.10 → springbootai-2.3.11}/CHANGELOG.md

@@ -1,5 +1,15 @@
 # Changelog
 
+## [2.3.11] - 2026-09-01
+
+- 修复 PyMyBatis 访问控制仅覆盖查询、行级条件直接拼接及 SQL 尾部条件位置错误，写操作和字段权限现在失败关闭，动态租户值必须使用绑定参数。
+- 修复事务提交前监听器异常被吞掉及 `REQUIRES_NEW`/`NOT_SUPPORTED` 同步上下文错误，提交否决会触发回滚并恢复外层上下文。
+- 修复 Spring Cloud Bus RabbitMQ API 调用错误和缺失消费者，增加 Broker 确认、广播消费者、去重、消息大小限制及明确的失败状态。
+- 加固 MCP/AI 工具边界：限制 HTTP 响应、队列、集合和 Schema 深度，拒绝压缩炸弹及高风险正则表达式。
+- 修复 Prometheus Summary 创建失败、Tracer 已结束 Span 无界增长以及 Actuator/启动日志暴露原始异常的问题。
+- 修复 CI 在 AI 测试切片后提前执行总覆盖率门槛的问题，覆盖率阈值现在只在全部测试切片完成后统一校验。
+- 修复 LangChain 完整演示未声明文档加载根目录的问题，并为临时文件自动清理及 Windows 重定向控制台增加 UTF-8 输出兼容。
+
 ## [2.3.10] - 2026-08-29
 
 - 修复 AI 代理上下游消息被 SQL 注入检测误报：`#{...}` 绑定值始终按数据处理，不再扫描消息内容；`${...}` 原始替换仍默认拒绝，只有配置允许且值完整匹配严格白名单时才能使用。

{springbootai-2.3.10/springbootAI.egg-info → springbootai-2.3.11}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: springbootAI
-Version: 2.3.10
+Version: 2.3.11
 Summary: Spring-style Python framework with Web, ORM, Cloud and AI integrations
 Author-email: YuConggen <1516933915@qq.com>
 License-Expression: MIT
@@ -19,7 +19,7 @@ Classifier: Programming Language :: Python :: 3.12
 Classifier: Topic :: Software Development :: Libraries :: Application Frameworks
 Classifier: Topic :: Database
 Classifier: Topic :: Internet :: WWW/HTTP :: Dynamic Content
-Requires-Python: >=3.10
+Requires-Python: <3.13,>=3.10
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: fastapi==0.141.1
@@ -120,6 +120,7 @@ Requires-Dist: psutil==6.1.1; extra == "all"
 Provides-Extra: dev
 Requires-Dist: pytest==9.1.1; extra == "dev"
 Requires-Dist: pytest-cov==7.1.0; extra == "dev"
+Requires-Dist: mypy==1.18.2; extra == "dev"
 Requires-Dist: redis==8.1.0; extra == "dev"
 Requires-Dist: sqlglot==27.28.1; extra == "dev"
 Provides-Extra: full
@@ -218,7 +219,7 @@ SpringBootAI 是一个采用 Spring 风格注解和分层结构的 Python 应用
 | AI 与编排 | 模型调用、Tools、RAG、Chain、状态图、MCP client/server | LangChain / LangGraph / 官方 MCP SDK |
 | 生产治理 | 健康检查、Prometheus、限流熔断、追踪、Swagger | prometheus-client / OpenTelemetry / OpenAPI |
 
-当前版本是 `2.3.10`；支持 Python 3.10、3.11 和 3.12，许可证为 MIT。项目仍标记为 Beta。用于公网高并发、合规敏感或支付/订单/库存等核心系统前，必须完成目标数据库、流量模型、故障恢复和安全基线验证。内嵌 Gateway 适合内部路由，不替代公网 Nginx/Kong/WAF；Seata `distributed` 对接官方 TC + TCC 回调；`at` 是无 TC 全局锁的单进程受限补偿模式，不应当作跨服务强一致方案。
+当前版本是 `2.3.11`；支持 Python 3.10、3.11 和 3.12，许可证为 MIT。项目仍标记为 Beta。用于公网高并发、合规敏感或支付/订单/库存等核心系统前，必须完成目标数据库、流量模型、故障恢复和安全基线验证。内嵌 Gateway 适合内部路由，不替代公网 Nginx/Kong/WAF；Seata `distributed` 对接官方 TC + TCC 回调；`at` 是无 TC 全局锁的单进程受限补偿模式，不应当作跨服务强一致方案。
 
 [新手指南](https://github.com/YUCONGGEN/springbootAI/blob/master/doc/BEGINNER_GUIDE.md) | [全部文档](https://github.com/YUCONGGEN/springbootAI/tree/master/doc) | [变更日志](https://github.com/YUCONGGEN/springbootAI/blob/master/CHANGELOG.md) | [安全报告](https://github.com/YUCONGGEN/springbootAI/blob/master/SECURITY.md) | [发布检查](https://github.com/YUCONGGEN/springbootAI/blob/master/doc/RELEASE_CHECKLIST.md)
 
@@ -349,10 +350,10 @@ SpringBootAI 是一个 **Python Web 框架**。它把 Java Spring Boot 的"注
 
 | 组件 | 当前版本 |
 |------|----------|
-| `springbootai` 框架 API | 2.3.10 |
-| `springbootai.orm.pymybatis` | 2.3.10 |
-| `springbootai.ai` AI 模块 | 2.3.10 |
-| `springbootai.langchain` LangChain 模块 | 2.3.10 |
+| `springbootai` 框架 API | 2.3.11 |
+| `springbootai.orm.pymybatis` | 2.3.11 |
+| `springbootai.ai` AI 模块 | 2.3.11 |
+| `springbootai.langchain` LangChain 模块 | 2.3.11 |
 | Python | 3.10+ |
 
 ### 1.4 适合什么场景

{springbootai-2.3.10 → springbootai-2.3.11}/README.md

@@ -56,7 +56,7 @@ SpringBootAI 是一个采用 Spring 风格注解和分层结构的 Python 应用
 | AI 与编排 | 模型调用、Tools、RAG、Chain、状态图、MCP client/server | LangChain / LangGraph / 官方 MCP SDK |
 | 生产治理 | 健康检查、Prometheus、限流熔断、追踪、Swagger | prometheus-client / OpenTelemetry / OpenAPI |
 
-当前版本是 `2.3.10`；支持 Python 3.10、3.11 和 3.12，许可证为 MIT。项目仍标记为 Beta。用于公网高并发、合规敏感或支付/订单/库存等核心系统前，必须完成目标数据库、流量模型、故障恢复和安全基线验证。内嵌 Gateway 适合内部路由，不替代公网 Nginx/Kong/WAF；Seata `distributed` 对接官方 TC + TCC 回调；`at` 是无 TC 全局锁的单进程受限补偿模式，不应当作跨服务强一致方案。
+当前版本是 `2.3.11`；支持 Python 3.10、3.11 和 3.12，许可证为 MIT。项目仍标记为 Beta。用于公网高并发、合规敏感或支付/订单/库存等核心系统前，必须完成目标数据库、流量模型、故障恢复和安全基线验证。内嵌 Gateway 适合内部路由，不替代公网 Nginx/Kong/WAF；Seata `distributed` 对接官方 TC + TCC 回调；`at` 是无 TC 全局锁的单进程受限补偿模式，不应当作跨服务强一致方案。
 
 [新手指南](https://github.com/YUCONGGEN/springbootAI/blob/master/doc/BEGINNER_GUIDE.md) | [全部文档](https://github.com/YUCONGGEN/springbootAI/tree/master/doc) | [变更日志](https://github.com/YUCONGGEN/springbootAI/blob/master/CHANGELOG.md) | [安全报告](https://github.com/YUCONGGEN/springbootAI/blob/master/SECURITY.md) | [发布检查](https://github.com/YUCONGGEN/springbootAI/blob/master/doc/RELEASE_CHECKLIST.md)
 
@@ -187,10 +187,10 @@ SpringBootAI 是一个 **Python Web 框架**。它把 Java Spring Boot 的"注
 
 | 组件 | 当前版本 |
 |------|----------|
-| `springbootai` 框架 API | 2.3.10 |
-| `springbootai.orm.pymybatis` | 2.3.10 |
-| `springbootai.ai` AI 模块 | 2.3.10 |
-| `springbootai.langchain` LangChain 模块 | 2.3.10 |
+| `springbootai` 框架 API | 2.3.11 |
+| `springbootai.orm.pymybatis` | 2.3.11 |
+| `springbootai.ai` AI 模块 | 2.3.11 |
+| `springbootai.langchain` LangChain 模块 | 2.3.11 |
 | Python | 3.10+ |
 
 ### 1.4 适合什么场景

{springbootai-2.3.10 → springbootai-2.3.11}/pyproject.toml

@@ -12,7 +12,7 @@ build-backend = "setuptools.build_meta"
 [project]
 # Distribution name: springbootAI (pip install springbootAI); import package: springbootai.
 name = "springbootAI"
-version = "2.3.10"
+version = "2.3.11"
 description = "Spring-style Python framework with Web, ORM, Cloud and AI integrations"
 readme = "README.md"
 license = "MIT"
@@ -32,7 +32,7 @@ classifiers = [
     "Topic :: Database",
     "Topic :: Internet :: WWW/HTTP :: Dynamic Content",
 ]
-requires-python = ">=3.10"
+requires-python = ">=3.10,<3.13"
 
 # ========================================
 # 核心依赖——类似 Spring Boot Starter Web/Core
@@ -199,6 +199,7 @@ all = [
 dev = [
     "pytest==9.1.1",
     "pytest-cov==7.1.0",
+    "mypy==1.18.2",
     "redis==8.1.0",
     "sqlglot==27.28.1",
 ]
@@ -278,6 +279,43 @@ markers = [
     "seata_integration: requires a compatible Seata SDK and real Seata Server",
 ]
 
+[tool.coverage.run]
+branch = true
+source = ["springbootai"]
+
+[tool.coverage.report]
+fail_under = 68
+show_missing = true
+
+[tool.mypy]
+python_version = "3.10"
+files = [
+    "springbootai/ai/annotation_runtime.py",
+    "springbootai/ai/advisors.py",
+    "springbootai/ai/vectorstore.py",
+    "springbootai/cloud/gateway.py",
+    "springbootai/websocket/router.py",
+    "springbootai/messaging/kafka.py",
+    "springbootai/messaging/rabbitmq.py",
+    "springbootai/data/repository.py",
+    "springbootai/retry/retry_decorator.py",
+    "springbootai/ai/tools.py",
+    "springbootai/cloud/bus.py",
+    "springbootai/cloud/tracer.py",
+    "springbootai/mcp/client.py",
+    "springbootai/mcp/config.py",
+    "springbootai/monitoring/prometheus.py",
+    "springbootai/orm/pymybatis/core/sql_session.py",
+    "springbootai/orm/pymybatis/security/access_control.py",
+    "springbootai/tx/synchronization.py",
+]
+ignore_missing_imports = true
+follow_imports = "skip"
+allow_untyped_defs = true
+allow_incomplete_defs = true
+check_untyped_defs = true
+show_error_codes = true
+
 [project.urls]
 Homepage = "https://github.com/YUCONGGEN/springbootAI"
 Repository = "https://github.com/YUCONGGEN/springbootAI.git"
@@ -296,4 +334,4 @@ include = ["springbootai*"]
 namespaces = false
 
 [tool.setuptools.package-data]
-springbootai = ["*.py", "**/*.py", "**/*.xml", "**/*.yml", "**/*.yaml"]
+springbootai = ["*.py", "py.typed", "**/*.py", "**/*.xml", "**/*.yml", "**/*.yaml"]

springbootai-2.3.11/requirements-ai.txt

@@ -0,0 +1,44 @@
+# ========================================
+# SpringBootAI optional AI/LangChain dependencies (exact versions)
+# ========================================
+# The AI module uses optional dependencies with a built-in fallback:
+#   - Without this file, native HTTP providers and FakeChatModel remain usable.
+#   - Installing it enables LangChain adapters detected by ai/providers.py.
+#
+# Install examples:
+#   pip install -r requirements-ai.txt
+#   pip install langchain-openai==1.4.2
+#
+# Recommended Python: 3.10+ (CI covers 3.10, 3.11 and 3.12).
+# Every pinned version is covered by tests and dependency auditing.
+# Keep this file ASCII-only: pip-audit may use the Windows system encoding.
+# ========================================
+
+# LangChain core (BaseChatModel, Embeddings and Runnable APIs).
+langchain-core==1.5.4
+
+# Classic APIs including LLMChain, ConversationChain and initialize_agent.
+langchain-classic==1.0.8
+
+# OpenAI-compatible providers (OpenAI, Azure, DeepSeek and Moonshot).
+langchain-openai==1.4.2
+
+# RecursiveCharacterTextSplitter and CharacterTextSplitter.
+langchain-text-splitters==1.1.2
+
+# Community integrations including FAISS, Chroma, DuckDuckGo and Wikipedia.
+langchain-community==0.4.2
+
+# Vector similarity calculations; 2.2.6 retains Python 3.10 compatibility.
+numpy==2.2.6
+
+# Pydantic v2 for validation and structured output parsing.
+pydantic==2.13.4
+
+# ========================================
+# Additional optional packages (install only when needed)
+# ========================================
+
+# Vector database: faiss-cpu
+# PDF document loading: pypdf
+# HTML parsing: beautifulsoup4

{springbootai-2.3.10 → springbootai-2.3.11}/requirements.txt

@@ -55,9 +55,4 @@ pydantic==2.13.4
 # 测试依赖
 pytest==9.1.1
 pytest-cov==7.1.0
-openpyxl==3.1.5
-
-bcrypt
-springbootai
-springbootai
-springbootai
+openpyxl==3.1.5

{springbootai-2.3.10 → springbootai-2.3.11/springbootAI.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: springbootAI
-Version: 2.3.10
+Version: 2.3.11
 Summary: Spring-style Python framework with Web, ORM, Cloud and AI integrations
 Author-email: YuConggen <1516933915@qq.com>
 License-Expression: MIT
@@ -19,7 +19,7 @@ Classifier: Programming Language :: Python :: 3.12
 Classifier: Topic :: Software Development :: Libraries :: Application Frameworks
 Classifier: Topic :: Database
 Classifier: Topic :: Internet :: WWW/HTTP :: Dynamic Content
-Requires-Python: >=3.10
+Requires-Python: <3.13,>=3.10
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: fastapi==0.141.1
@@ -120,6 +120,7 @@ Requires-Dist: psutil==6.1.1; extra == "all"
 Provides-Extra: dev
 Requires-Dist: pytest==9.1.1; extra == "dev"
 Requires-Dist: pytest-cov==7.1.0; extra == "dev"
+Requires-Dist: mypy==1.18.2; extra == "dev"
 Requires-Dist: redis==8.1.0; extra == "dev"
 Requires-Dist: sqlglot==27.28.1; extra == "dev"
 Provides-Extra: full
@@ -218,7 +219,7 @@ SpringBootAI 是一个采用 Spring 风格注解和分层结构的 Python 应用
 | AI 与编排 | 模型调用、Tools、RAG、Chain、状态图、MCP client/server | LangChain / LangGraph / 官方 MCP SDK |
 | 生产治理 | 健康检查、Prometheus、限流熔断、追踪、Swagger | prometheus-client / OpenTelemetry / OpenAPI |
 
-当前版本是 `2.3.10`；支持 Python 3.10、3.11 和 3.12，许可证为 MIT。项目仍标记为 Beta。用于公网高并发、合规敏感或支付/订单/库存等核心系统前，必须完成目标数据库、流量模型、故障恢复和安全基线验证。内嵌 Gateway 适合内部路由，不替代公网 Nginx/Kong/WAF；Seata `distributed` 对接官方 TC + TCC 回调；`at` 是无 TC 全局锁的单进程受限补偿模式，不应当作跨服务强一致方案。
+当前版本是 `2.3.11`；支持 Python 3.10、3.11 和 3.12，许可证为 MIT。项目仍标记为 Beta。用于公网高并发、合规敏感或支付/订单/库存等核心系统前，必须完成目标数据库、流量模型、故障恢复和安全基线验证。内嵌 Gateway 适合内部路由，不替代公网 Nginx/Kong/WAF；Seata `distributed` 对接官方 TC + TCC 回调；`at` 是无 TC 全局锁的单进程受限补偿模式，不应当作跨服务强一致方案。
 
 [新手指南](https://github.com/YUCONGGEN/springbootAI/blob/master/doc/BEGINNER_GUIDE.md) | [全部文档](https://github.com/YUCONGGEN/springbootAI/tree/master/doc) | [变更日志](https://github.com/YUCONGGEN/springbootAI/blob/master/CHANGELOG.md) | [安全报告](https://github.com/YUCONGGEN/springbootAI/blob/master/SECURITY.md) | [发布检查](https://github.com/YUCONGGEN/springbootAI/blob/master/doc/RELEASE_CHECKLIST.md)
 
@@ -349,10 +350,10 @@ SpringBootAI 是一个 **Python Web 框架**。它把 Java Spring Boot 的"注
 
 | 组件 | 当前版本 |
 |------|----------|
-| `springbootai` 框架 API | 2.3.10 |
-| `springbootai.orm.pymybatis` | 2.3.10 |
-| `springbootai.ai` AI 模块 | 2.3.10 |
-| `springbootai.langchain` LangChain 模块 | 2.3.10 |
+| `springbootai` 框架 API | 2.3.11 |
+| `springbootai.orm.pymybatis` | 2.3.11 |
+| `springbootai.ai` AI 模块 | 2.3.11 |
+| `springbootai.langchain` LangChain 模块 | 2.3.11 |
 | Python | 3.10+ |
 
 ### 1.4 适合什么场景

{springbootai-2.3.10 → springbootai-2.3.11}/springbootAI.egg-info/SOURCES.txt

@@ -18,6 +18,7 @@ springbootAI.egg-info/top_level.txt
 springbootai/__init__.py
 springbootai/devtools.py
 springbootai/main.py
+springbootai/py.typed
 springbootai/ai/__init__.py
 springbootai/ai/advisors.py
 springbootai/ai/annotation_runtime.py
@@ -306,6 +307,8 @@ tests/test_devtools_lifecycle_hardening.py
 tests/test_di_config_event_full.py
 tests/test_enterprise_safety_regressions.py
 tests/test_excel_module.py
+tests/test_final_enterprise_hardening.py
+tests/test_followup_enterprise_fixes.py
 tests/test_hateoas.py
 tests/test_i18n_module.py
 tests/test_io_http_logging_hardening.py
@@ -317,6 +320,7 @@ tests/test_langchain_ext.py
 tests/test_langchain_module.py
 tests/test_langgraph_comprehensive.py
 tests/test_langgraph_module.py
+tests/test_latest_enterprise_audit_fixes.py
 tests/test_logging_config.py
 tests/test_lombok_annotations.py
 tests/test_mcp_module.py
@@ -328,9 +332,11 @@ tests/test_new_features.py
 tests/test_oauth2.py
 tests/test_optional_component_timeouts.py
 tests/test_orm_pymybatis_full.py
+tests/test_post_audit_enterprise_fixes.py
 tests/test_production_readiness.py
 tests/test_pymybatis_configuration_hardening.py
 tests/test_pymybatis_contract.py
+tests/test_remaining_enterprise_hardening.py
 tests/test_request_metrics.py
 tests/test_retry.py
 tests/test_runtime_configuration_fixes.py

{springbootai-2.3.10 → springbootai-2.3.11}/springbootAI.egg-info/requires.txt

@@ -50,6 +50,7 @@ httpx==0.28.1
 [dev]
 pytest==9.1.1
 pytest-cov==7.1.0
+mypy==1.18.2
 redis==8.1.0
 sqlglot==27.28.1
 

{springbootai-2.3.10 → springbootai-2.3.11}/springbootai/__init__.py

@@ -19,7 +19,7 @@ from .config import *
 from .utils import *
 from .main import create_app, run, SpringApplication, run_cli
 
-__version__ = "2.3.10"
+__version__ = "2.3.11"
 __author__ = "yuconggen"
 __license__ = "MIT"
 

{springbootai-2.3.10 → springbootai-2.3.11}/springbootai/ai/__init__.py

@@ -16,7 +16,8 @@ SpringBootAI AI 模块 - 对齐 Spring AI 的 ChatClient/Advisor/ETL 抽象，
 from springbootai.ai.core import (
     Advisor, AdvisorRequest, ChatClient, ChatClientBuilder, ChatModel,
     ChatResponse, EmbeddingModel, Generation, Message, MessageType,
-    PromptSpec, TokenBudgetExceededError, ToolLoopLimitExceededError,
+    PromptSpec, AIConcurrencyLimitError, TokenBudgetExceededError,
+    ToolLoopLimitExceededError,
 )
 from springbootai.ai.annotations import (
     Agent, AiAdvisor, AiCache, AiClient, AiMemory, AiRetry, ContentModeration,
@@ -57,14 +58,14 @@ from springbootai.ai.observability import AIMetrics, ai_metrics
 from springbootai.ai.autoconfig import AIProperties, bind_ai_config, configure_ai
 from springbootai.ai.annotation_runtime import ContentModerationError
 
-__version__ = "2.3.10"
+__version__ = "2.3.11"
 
 __all__ = [
     # core
     "Advisor", "AdvisorRequest", "ChatClient", "ChatClientBuilder",
     "ChatModel", "ChatResponse", "EmbeddingModel", "Generation", "Message",
     "MessageType", "PromptSpec", "TokenBudgetExceededError",
-    "ToolLoopLimitExceededError",
+    "ToolLoopLimitExceededError", "AIConcurrencyLimitError",
     # annotations
     "AiAdvisor", "AiClient", "AiMemory", "Tool", "Prompt", "RAG",
     "StructuredOutput", "Agent", "Embedding", "VectorStore", "AiRetry",

{springbootai-2.3.10 → springbootai-2.3.11}/springbootai/ai/advisors.py

@@ -2,6 +2,7 @@
 Advisor 实现 - QuestionAnswerAdvisor（RAG）与 MessageChatMemoryAdvisor（会话记忆）。
 """
 import inspect
+import hashlib
 from urllib.parse import quote
 from typing import Any, Dict, List, Optional, Sequence
 
@@ -27,8 +28,14 @@ class MessageChatMemoryAdvisor(Advisor):
 
     @staticmethod
     def _build_namespace(context: dict) -> str:
-        tenant = quote(str(context.get("tenant_id", "")).strip()[:128], safe="-_.~")
-        user = quote(str(context.get("user_id", "")).strip()[:128], safe="-_.~")
+        def identity_part(value: Any) -> str:
+            raw = str(value or "").strip()
+            if len(raw) > 128:
+                raw = "sha256-" + hashlib.sha256(raw.encode("utf-8")).hexdigest()
+            return quote(raw, safe="-_.~")
+
+        tenant = identity_part(context.get("tenant_id", ""))
+        user = identity_part(context.get("user_id", ""))
         parts = [p for p in (tenant, user) if p]
         return ":".join(parts) if parts else ""
 
@@ -52,9 +59,17 @@ class MessageChatMemoryAdvisor(Advisor):
             kwargs["namespace"] = namespace
         return method(*args, **kwargs)
 
-    def __init__(self, memory: ChatMemory, max_messages: int = 20):
+    def __init__(self, memory: ChatMemory, max_messages: int = 20,
+                 allow_global_namespace: bool = False):
+        if isinstance(max_messages, bool) or not isinstance(max_messages, int):
+            raise TypeError("max_messages must be an integer")
+        if not 1 <= max_messages <= 100_000:
+            raise ValueError("max_messages must be in [1, 100000]")
+        if not isinstance(allow_global_namespace, bool):
+            raise TypeError("allow_global_namespace must be a boolean")
         self.memory = memory
         self.max_messages = max_messages
+        self.allow_global_namespace = allow_global_namespace
 
     def advise_request(self, request: AdvisorRequest) -> AdvisorRequest:
         conv_id = request.context.get("conversation_id")
@@ -69,6 +84,14 @@ class MessageChatMemoryAdvisor(Advisor):
         # Namespace is passed per operation. Never mutate a shared memory bean:
         # concurrent requests may belong to different tenants.
         ns = self._build_namespace(request.context)
+        if not ns and not self.allow_global_namespace:
+            request.context["memory_disabled"] = True
+            logger = __import__("logging").getLogger("Spring.AI")
+            logger.debug(
+                "MessageChatMemoryAdvisor: authenticated tenant/user identity "
+                "is missing; skipping global memory access")
+            return request
+        request.context.pop("memory_disabled", None)
         request.context["memory_namespace"] = ns
 
         history = self._call_memory(
@@ -82,7 +105,7 @@ class MessageChatMemoryAdvisor(Advisor):
     def advise_response(self, response: ChatResponse,
                         request: AdvisorRequest) -> ChatResponse:
         conv_id = request.context.get("conversation_id")
-        if not conv_id:
+        if not conv_id or request.context.get("memory_disabled"):
             return response
         ns = str(request.context.get("memory_namespace") or
                  self._build_namespace(request.context))
@@ -120,7 +143,20 @@ class QuestionAnswerAdvisor(Advisor):
                  embedding_model=None,
                  harden_injection: bool = True,
                  filter_metadata: Optional[Dict[str, Any]] = None,
-                 filter_context_keys: Sequence[str] = ("tenant_id",)):
+                 filter_context_keys: Sequence[str] = ("tenant_id",),
+                 max_context_chars: int = 100_000,
+                 max_document_chars: int = 25_000):
+        if isinstance(top_k, bool) or not isinstance(top_k, int):
+            raise TypeError("top_k must be an integer")
+        if not 1 <= top_k <= 1000:
+            raise ValueError("top_k must be in [1, 1000]")
+        for name, value in (
+                ("max_context_chars", max_context_chars),
+                ("max_document_chars", max_document_chars)):
+            if isinstance(value, bool) or not isinstance(value, int):
+                raise TypeError(f"{name} must be an integer")
+            if not 1 <= value <= 10_000_000:
+                raise ValueError(f"{name} must be in [1, 10000000]")
         self.vector_store = vector_store
         self.prompt_template = prompt_template or self.DEFAULT_PROMPT_TEMPLATE
         self.top_k = top_k
@@ -128,6 +164,8 @@ class QuestionAnswerAdvisor(Advisor):
         self.harden_injection = harden_injection
         self.filter_metadata = dict(filter_metadata or {})
         self.filter_context_keys = tuple(filter_context_keys)
+        self.max_context_chars = max_context_chars
+        self.max_document_chars = max_document_chars
 
     def advise_request(self, request: AdvisorRequest) -> AdvisorRequest:
         # 取最后一条用户消息作为查询
@@ -160,10 +198,49 @@ class QuestionAnswerAdvisor(Advisor):
             filter_metadata=filters or None,
         )
         docs = self.vector_store.similarity_search(search_req)
+        # Missing identity must never become an implicit "all tenants" scope.
+        # Public documents (without tenant/user metadata) remain available for
+        # single-tenant and public-knowledge use cases.
+        identity_keys = set(self.filter_context_keys) | {"user_id"}
+        identity_scoped = any(key in filters for key in identity_keys)
+        if not identity_scoped:
+            docs = [
+                doc for doc in docs
+                if isinstance(doc.metadata, dict)
+                and not any(
+                    doc.metadata.get(key) not in (None, "")
+                    for key in identity_keys
+                )
+            ]
         if not docs:
             return request
 
-        context = "\n---\n".join(d.content for d in docs)
+        selected_docs = []
+        context_parts: List[str] = []
+        remaining = self.max_context_chars
+        truncated = False
+        separator = "\n---\n"
+        for doc in docs:
+            content = str(doc.content or "")
+            if len(content) > self.max_document_chars:
+                content = content[:self.max_document_chars]
+                truncated = True
+            separator_cost = len(separator) if context_parts else 0
+            available = remaining - separator_cost
+            if available <= 0:
+                truncated = True
+                break
+            if len(content) > available:
+                content = content[:available]
+                truncated = True
+            if not content:
+                continue
+            context_parts.append(content)
+            selected_docs.append(doc)
+            remaining -= separator_cost + len(content)
+        if not context_parts:
+            return request
+        context = separator.join(context_parts)
         system_text = self.prompt_template.format(context=context)
         if self.harden_injection:
             # Prompt 注入加固：把外部文档与指令清晰隔离，并要求模型将上下文
@@ -179,8 +256,10 @@ class QuestionAnswerAdvisor(Advisor):
         request.messages = new_messages
         # 记录引用文档
         request.context["retrieved_documents"] = [
-            {"id": d.id, "content": d.content[:200]} for d in docs
+            {"id": d.id, "content": str(d.content)[:200]}
+            for d in selected_docs
         ]
+        request.context["retrieval_truncated"] = truncated
         return request
 
 
@@ -190,11 +269,24 @@ class SimpleLoggerAdvisor(Advisor):
     """
     order = 0
 
-    def __init__(self):
+    def __init__(self, max_events: int = 1000):
+        if isinstance(max_events, bool) or not isinstance(max_events, int) \
+                or max_events <= 0 or max_events > 100_000:
+            raise ValueError("max_events must be between 1 and 100000")
+        import threading
+        self.max_events = max_events
         self.events: List[Dict[str, Any]] = []
+        self._events_lock = threading.RLock()
+
+    def _record(self, event: Dict[str, Any]) -> None:
+        with self._events_lock:
+            self.events.append(event)
+            overflow = len(self.events) - self.max_events
+            if overflow > 0:
+                del self.events[:overflow]
 
     def advise_request(self, request: AdvisorRequest) -> AdvisorRequest:
-        self.events.append({
+        self._record({
             "phase": "request",
             "message_count": len(request.messages),
             "tools": len(request.tool_registry.names())
@@ -204,7 +296,7 @@ class SimpleLoggerAdvisor(Advisor):
 
     def advise_response(self, response: ChatResponse,
                         request: AdvisorRequest) -> ChatResponse:
-        self.events.append({
+        self._record({
             "phase": "response",
             "content_length": len(response.content()),
         })