revise v1.3.0 release notes

Author: Issac-Newton

docs/i18n/zh-Hans/docusaurus-plugin-content-docs/version-1.3.x/Release Notes/v1.3.0.md

@@ -1,107 +1,68 @@
 # v1.3.0
 
 ## 发布日期
-2026 年 2 月 3 日
-
+2026 年 3 月 4 日
 
 ---
 
 ## SDK
 
 ### 新功能
 
-#### 运行时环境（Runtime Environment）
-**[运行时环境参考文档](../References/Python%20SDK%20References/runtime-env.md)**
-
-- **新增**：通过 `extra_symlink_dir` 和 `extra_symlink_executables` 支持可执行文件符号链接，可将运行时可执行文件暴露到系统路径
-
-#### Model Service（实验性）
-**[Model Service 参考文档](../References/Python%20SDK%20References/model-service.md)**
-
-处理 Agent 与 LLM 推理服务之间的模型调用通信：
-- Local 模式：Agent 与 Roll 运行时之间的文件通信，基于文件的通信协议，实现请求-响应机制
-- Proxy 模式：支持请求转发到外部 LLM 服务，具备路由和重试功能
+#### Sandbox 客户端
+- **新增**: Sandbox 客户端中的 `close_session` 方法，用于显式管理会话生命周期 ([#451](https://github.com/alibaba/ROCK/pull/451))
 
-- SDK 支持 `anti_call_llm`，用于代理 Agent 请求
-- **新增**：完整的 CLI start 命令参数（`--config-file`、`--host`、`--port`、`--proxy-base-url`、`--retryable-status-codes`、`--request-timeout`）
-- **新增**：轨迹（traj）日志记录 - 将 LLM 请求/响应记录到 JSONL 文件（`ROCK_MODEL_SERVICE_DATA_DIR`、`ROCK_MODEL_SERVICE_TRAJ_APPEND_MODE`）
-- **新增**：服务端 `ModelServiceConfig` 配置（`host`、`port`、`proxy_base_url`、`proxy_rules`、`retryable_status_codes`、`request_timeout`）
+#### 评测示例
+- **swe_bench/**: 集成 iFlow 的 SWE-bench Verified 评测示例 ([#508](https://github.com/alibaba/ROCK/pull/508))
 
-#### Agent 示例
-**[示例目录](https://github.com/Modelized/ROCK/tree/master/examples/agents)**
+## Admin
 
-演示 ROCK Agent 与各种 AI 框架集成的实践示例：
+### 新功能
 
-- **claude_code/**: Claude Code Agent 集成示例（使用 Node.js 运行时）
-  - 命令：`claude -p ${prompt}`
-  - 依赖：`npm install -g @anthropic-ai/claude-code`
-  - 通过 `rock_agent_config.yaml` 配置
+#### Kubernetes Operator
+- **新增**: 支持 K8s Operator，用于基于 Kubernetes 的沙箱编排 ([#423](https://github.com/alibaba/ROCK/pull/423))
+- 将 `extensions` 重命名为 `extended_params` 以保持一致性 ([#470](https://github.com/alibaba/ROCK/pull/470))
 
-- **iflow_cli/**: iFlow CLI Agent 集成示例（使用 Node.js 运行时）
-  - 命令：`iflow -p ${prompt} --yolo`
-  - 依赖：`npm i -g @iflow-ai/iflow-cli@latest`
+#### Docker 镜像仓库登录
+- **新增**: 支持 Docker 镜像仓库登录，以便从私有镜像仓库拉取镜像 ([#537](https://github.com/alibaba/ROCK/pull/537))
 
-- **swe_agent/**: SWE Agent 集成示例
-  - 演示标准 ROCK Agent 设置模式
+#### Sandbox
 
-- **iflow_cli/integration_with_model_service/**: Model Service 集成示例
-  - **local/**: Local 模式示例，包含自定义 LLM 后端
-    - 展示 `anti_call_llm` 用法和服务循环模式
-  - **proxy/**: Proxy 模式示例，配合 iFlow CLI 使用
-    - 演示 `rock model-service start --type proxy --proxy-base-url` 用法
+- **新增**: 沙箱镜像元数据中新增 `image_os` 字段 ([#510](https://github.com/alibaba/ROCK/pull/510))
+- **新增**: 支持 Kata 容器运行时，增强沙箱隔离性 ([#474](https://github.com/alibaba/ROCK/pull/474))
+- **新增**: 命名空间级别的指标采集 ([#427](https://github.com/alibaba/ROCK/pull/427))
+- **新增**: 支持指定端点的监控 ([#469](https://github.com/alibaba/ROCK/pull/469))
+- **新增**: 沙箱配置中新增自动清理秒数参数 ([#458](https://github.com/alibaba/ROCK/pull/458))
+- **新增**: `get_status` 遇到超时或失败阶段时上报指标 ([#552](https://github.com/alibaba/ROCK/pull/552))
 
+---
 
-## Admin
+### Enhancements
 
-### 新功能
+#### 性能优化
+- 将 `get_status` 从 rocklet 委托至 Ray operator，改善关注点分离 ([#487](https://github.com/alibaba/ROCK/pull/487))
 
-#### 任务调度器（Task Scheduler）
-一个灵活的任务调度系统，用于管理 Ray Worker 节点上的周期性维护任务。
-
-**内置任务：**
-- **ImageCleanupTask**：使用 [docuum](https://github.com/stepchowfun/docuum) 进行 Docker 镜像清理，支持可配置的磁盘阈值
-
-**配置示例：**
-```yaml
-scheduler:
-  enabled: true
-  worker_cache_ttl: 3600
-  tasks:
-    - task_class: "rock.admin.scheduler.tasks.image_cleanup_task.ImageCleanupTask"
-      enabled: true
-      interval_seconds: 3600
-      params:
-        threshold: "1T"
-```
-
-**可扩展性：**
-通过继承 `BaseTask` 并实现 `run_action(runtime: RemoteSandboxRuntime)` 方法来创建自定义任务。
-
-#### Entrypoints
-
-- **新增**: 批量查询沙箱状态 API (`POST /sandboxes/batch`)
-  - 在单个请求中高效检索多个沙箱的状态信息
-  - 接受包含 `sandbox_ids` 列表的 `BatchSandboxStatusRequest` 请求
-  - 返回包含所有请求沙箱状态详情的 `BatchSandboxStatusResponse` 响应
-  - 支持可配置的最大数量限制（默认值来自 `batch_get_status_max_count` 配置）
-
-- **新增**: 沙箱列表和过滤 API (`GET /sandboxes`)
-  - 使用灵活的查询参数查询和过滤沙箱
-  - 通过 `page` 和 `page_size` 参数支持分页
-  - 返回包含条目列表、总数和 `has_more` 标识的 `SandboxListResponse` 响应
-  - 支持按沙箱属性过滤（例如：部署类型、状态、自定义元数据等）
-  - 最大页面大小可通过 `batch_get_status_max_count` 配置
+#### 架构重构
+- 将沙箱生命周期管理操作（启动/停止/销毁）委托至 operator 层 ([#423](https://github.com/alibaba/ROCK/pull/423))
+- 将非管理类方法移回 proxy 服务，使服务边界更清晰 ([#433](https://github.com/alibaba/ROCK/pull/433))
+- 将调度器从独立进程重构为独立线程，降低资源开销 ([#507](https://github.com/alibaba/ROCK/pull/507))
 
-#### Sandbox
+#### 部署
+- 部署格式化方法中支持 `${}` 和 `<<>>` 模板语法 ([#485](https://github.com/alibaba/ROCK/pull/485))
+- 在不需要时跳过对运行命令的包装 ([#483](https://github.com/alibaba/ROCK/pull/483))
+- 沙箱代理 HTTP 支持 ([#453](https://github.com/alibaba/ROCK/pull/453))
 
-- 修改 Sandbox 日志格式：时间戳使用 ISO 8601 格式，默认时区为 Asia/Shanghai（亚洲/上海）
-- 丰富 SandboxInfo：添加 create_time、start_time、stop_time
-- 添加计费日志：当 sandbox 关闭时记录计费日志
+#### 配置
+- 将 `/tmp` 作为 `ROCK_SERVICE_STATUS_DIR` 的默认值 ([#522](https://github.com/alibaba/ROCK/pull/522))
+- 支持阿里云 MSE 的 `server_addresses` 以集成 Nacos ([#521](https://github.com/alibaba/ROCK/pull/521))
 
 ---
 
-### 增强
+### 缺陷修复
+
+- 修复自定义安装命令未设置工作目录的问题 ([#505](https://github.com/alibaba/ROCK/pull/505))
+- 修复模型服务 CLI 命令中 Python 警告未被抑制的问题 ([#503](https://github.com/alibaba/ROCK/pull/503))
+- 修复本地 admin 启动失败的问题 ([#494](https://github.com/alibaba/ROCK/pull/494))
+- 修复调度器任务文件检测问题 ([#465](https://github.com/alibaba/ROCK/pull/465))
+- 修复抛出异常时 `count.total` 指标未上报的问题 ([#527](https://github.com/alibaba/ROCK/pull/527))
 
-#### 性能优化
-- 实现 get_status 接口逻辑与 Ray 的解耦：将 API 响应耗时从 1s 以上显著降低至 100ms 级别；支持新旧逻辑的动态在线切换。
-- 将ray相关操作从sandbox manager中下沉至ray service。

docs/versioned_docs/version-1.3.x/Release Notes/v1.3.0.md

@@ -1,7 +1,7 @@
 # v1.3.0
 
 ## Release Date
-February 3, 2026
+Mar 4, 2026
 
 
 ---
@@ -10,54 +10,12 @@ February 3, 2026
 
 ### New Features
 
-#### Runtime Environment
-**[Runtime Environment References Documentation](../References/Python%20SDK%20References/runtime-env.md)**
-
-- **NEW**: Executable symlink support via `extra_symlink_dir` and `extra_symlink_executables` for exposing runtime executables to system paths ([#384](https://github.com/alibaba/ROCK/pull/384))
-
-#### Model Service (Experimental)
-**[Model Service References Documentation](../References/Python%20SDK%20References/model-service.md)**
-
-Handles AI model call communications between agents and LLM inference services:
-- Local mode: File-based IPC between Agent and Roll runtime, implementing request-response via file protocol
-- Proxy mode: Request forwarding to external LLM services with routing and retry support
-
-- SDK supports `anti_call_llm` for proxying Agent requests ([#300](https://github.com/alibaba/ROCK/pull/300))
-- **NEW**: Complete CLI start command parameters (`--config-file`, `--host`, `--port`, `--proxy-base-url`, `--retryable-status-codes`, `--request-timeout`) ([#386](https://github.com/alibaba/ROCK/pull/386), [#388](https://github.com/alibaba/ROCK/pull/388))
-- **NEW**: Trajectory (traj) logging - records LLM request/response to JSONL files (`ROCK_MODEL_SERVICE_DATA_DIR`, `ROCK_MODEL_SERVICE_TRAJ_APPEND_MODE`) ([#394](https://github.com/alibaba/ROCK/pull/394), [#396](https://github.com/alibaba/ROCK/pull/396))
-- **NEW**: Server-side `ModelServiceConfig` configuration (`host`, `port`, `proxy_base_url`, `proxy_rules`, `retryable_status_codes`, `request_timeout`) ([#386](https://github.com/alibaba/ROCK/pull/386))
-
 #### Sandbox Client
 - **NEW**: `close_session` method in Sandbox client for explicit session lifecycle management ([#451](https://github.com/alibaba/ROCK/pull/451))
 
-#### Agent Examples
-**[Examples Directory](https://github.com/alibaba/ROCK/tree/master/examples/agents)**
-
-Practical examples demonstrating ROCK Agent integration with various AI frameworks:
-
-- **claude_code/**: Claude Code Agent integration example using Node.js runtime ([#356](https://github.com/alibaba/ROCK/pull/356))
-  - Command: `claude -p ${prompt}`
-  - Requires: `npm install -g @anthropic-ai/claude-code`
-  - Configuration via `rock_agent_config.yaml`
-
-- **iflow_cli/**: iFlow CLI Agent integration example using Node.js runtime ([#356](https://github.com/alibaba/ROCK/pull/356))
-  - Command: `iflow -p ${prompt} --yolo`
-  - Requires: `npm i -g @iflow-ai/iflow-cli@latest`
-
-- **swe_agent/**: SWE Agent integration example ([#143](https://github.com/alibaba/ROCK/pull/143))
-  - Demonstrates standard ROCK Agent setup pattern
-
-- **qwen_code/cursor_cli/**: Qwen Code and Cursor CLI Agent integration examples ([#426](https://github.com/alibaba/ROCK/pull/426))
-
 #### Evaluation Examples
 - **swe_bench/**: SWE-bench Verified evaluation demo with iFlow integration ([#508](https://github.com/alibaba/ROCK/pull/508))
 
-- **iflow_cli/integration_with_model_service/**: Model Service integration examples ([#402](https://github.com/alibaba/ROCK/pull/402))
-  - **local/**: Local mode example with custom LLM backend
-    - Shows `anti_call_llm` usage and model service loop pattern
-  - **proxy/**: Proxy mode example with iFlow CLI
-    - Demonstrates `rock model-service start --type proxy --proxy-base-url` usage
-
 
 ## Admin
 
@@ -67,54 +25,12 @@ Practical examples demonstrating ROCK Agent integration with various AI framewor
 - **NEW**: K8s Operator support for Kubernetes-based sandbox orchestration ([#423](https://github.com/alibaba/ROCK/pull/423))
 - Rename `extensions` to `extended_params` for consistency ([#470](https://github.com/alibaba/ROCK/pull/470))
 
-#### Early Docker Validation
-- **NEW**: Docker availability validation in `SandboxManager.start_async()` — fails fast with clear error message before attempting sandbox creation ([#542](https://github.com/alibaba/ROCK/pull/542))
-
 #### Docker Registry Login
 - **NEW**: Support Docker registry login for pulling images from private registries ([#537](https://github.com/alibaba/ROCK/pull/537))
 
-#### Task Scheduler
-A flexible task scheduling system for managing periodic maintenance tasks across Ray workers.
-
-**Built-in Tasks:**
-- **ImageCleanupTask**: Docker image cleanup using [docuum](https://github.com/stepchowfun/docuum) with configurable disk threshold ([#382](https://github.com/alibaba/ROCK/pull/382))
-
-**Configuration Example:**
-```yaml
-scheduler:
-  enabled: true
-  worker_cache_ttl: 3600
-  tasks:
-    - task_class: "rock.admin.scheduler.tasks.image_cleanup_task.ImageCleanupTask"
-      enabled: true
-      interval_seconds: 3600
-      params:
-        threshold: "1T"
-```
-
-**Extensibility:**
-Create custom tasks by extending `BaseTask` and implementing the `run_action(runtime: RemoteSandboxRuntime)` method.
-
-#### Entrypoints
-
-- **NEW**: Batch sandbox status query API (`POST /sandboxes/batch`) ([#262](https://github.com/alibaba/ROCK/pull/262))
-  - Efficiently retrieve status information for multiple sandboxes in a single request
-  - Accepts `BatchSandboxStatusRequest` with `sandbox_ids` list
-  - Returns `BatchSandboxStatusResponse` containing status details for all requested sandboxes
-  - Supports up to configurable maximum count (default from `batch_get_status_max_count`)
-
-- **NEW**: Sandbox listing and filtering API (`GET /sandboxes`) ([#266](https://github.com/alibaba/ROCK/pull/266))
-  - Query and filter sandboxes with flexible query parameters
-  - Supports pagination via `page` and `page_size` parameters
-  - Returns `SandboxListResponse` with items, total count, and `has_more` indicator
-  - Enables filtering by sandbox attributes (e.g., deployment type, status, custom metadata)
-  - Maximum page size configurable via `batch_get_status_max_count`
 
 #### Sandbox
 
-- Modify Sandbox log format: use iso 8601 format for timestamp, default time zone is Asia/Shanghai ([#289](https://github.com/alibaba/ROCK/pull/289))
-- Enrich SandboxInfo: add create_time, start_time, stop_time for metrics ([#365](https://github.com/alibaba/ROCK/pull/365))
-- Add Billing log when sandbox is closed ([#365](https://github.com/alibaba/ROCK/pull/365))
 - **NEW**: `image_os` field for sandbox image metadata ([#510](https://github.com/alibaba/ROCK/pull/510))
 - **NEW**: Kata container runtime support for enhanced sandbox isolation ([#474](https://github.com/alibaba/ROCK/pull/474))
 - **NEW**: Namespace-level metrics collection ([#427](https://github.com/alibaba/ROCK/pull/427))
@@ -127,8 +43,6 @@ Create custom tasks by extending `BaseTask` and implementing the `run_action(run
 ### Enhancements
 
 #### Performance Optimizations
-- Decouple the `get_status` API logic from Ray, reducing API latency from 1s+ to 100ms+; support dynamically toggling between the new and legacy logic. ([#268](https://github.com/alibaba/ROCK/pull/268))
-- Sink ray operations from sandbox manager into ray service. ([#403](https://github.com/alibaba/ROCK/pull/403))
 - Delegate `get_status` from rocklet into Ray operator for improved separation of concerns ([#487](https://github.com/alibaba/ROCK/pull/487))
 
 #### Architecture Refactoring