feat(agent_runtime): align SDK model with agentrun-20250910 & support workspace_name

[Sodawyx]

Summary

• 对齐官方模型 (agentrun-20250910)：补齐 AgentRuntime / Endpoint / List 入参字段差异，下线已被 system_tags 覆盖的 tags 字段；新增 NASConfig / OSSMountConfig / ScalingConfig / RegistryConfig / ProtocolSettings 等辅助模型一比一对齐。
• 支持 workspace_name：创建 / 查询 Agent Runtime 时可直接传 workspace 名称，SDK 自动调 ListWorkspaces 解析为 workspace_id（带 (ak, region, name) 缓存），无需手查 ID；同时传 workspace_id + workspace_name 时抛 ValueError，找不到抛 ResourceNotExistError，重名抛 ValueError。
• 示例 examples/quickstart_runtime.py 演示通过镜像部署 AgentRuntime 并以 workspace_name 选择工作空间。

主要字段变更

• AgentRuntimeMutableProps: + disk_size / enable_session_isolation / nas_config / oss_mount_config; − tags
• AgentRuntimeListInput: + status / workspace_ids / workspace_name / workspace_names; − tags
• AgentRuntimeEndpointMutableProps: + disable_public_network_access / scaling_config; − tags
• AgentRuntimeEndpointUpdateInput: + delete_scaling_config
• AgentRuntimeContainer: + acr_instance_id / image_registry_type / port / registry_config
• AgentRuntimeProtocolConfig: + protocol_settings
• AgentRuntimeEndpointRoutingWeight.weight: int → float
• AgentRuntimeImmutableProps: + workspace_name（流入 CreateInput）

Test plan

• 全量 3469 单测通过
• agentrun.agent_runtime 总覆盖率 99%，_workspace.py 95%，model.py 100% 行/分支覆盖
• mypy --config-file mypy.ini agentrun/agent_runtime/ 无 issue
• 新增 25 用例覆盖 workspace 解析（精确匹配 / 缓存 / 空名 / 找不到 / 重名 / Tea 异常透传 / sync+async 互斥校验）

[Copilot]

Pull request overview

This PR updates the agentrun.agent_runtime SDK to align with the official agentrun-20250910 model/schema while adding a convenience feature that lets callers specify workspace_name (resolved internally to workspace_id) when creating and listing Agent Runtimes.

Changes:

• Added workspace name → ID resolution helper (_workspace.py) with in-memory caching and integrated it into AgentRuntimeClient create/list flows.
• Aligned Agent Runtime models/inputs with agentrun-20250910 (new config sub-models, new fields, tags removal in favor of system_tags, routing weight type change).
• Updated unit tests to cover new models and workspace resolution behavior.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
agentrun/agent_runtime/_workspace.py	New resolver module to map workspace_name/workspace_names to IDs with caching and Tea-exception mapping.
agentrun/agent_runtime/client.py	Integrates workspace resolution and adds mutual-exclusion validation for workspace fields in create/list.
agentrun/agent_runtime/model.py	Adds/updates models and fields to match agentrun-20250910 (registry/protocol/NAS/OSS/scaling, tags→system_tags, weight float, workspace_name field).
agentrun/agent_runtime/runtime.py	Updates high-level AgentRuntime.list_all* filter parameters to match new model fields (system_tags, status, workspace filters).
agentrun/agent_runtime/__client_async_template.py	Updates codegen template to include workspace resolution logic for async client methods.
agentrun/agent_runtime/__runtime_async_template.py	Updates codegen template for list_all_async signature/args to match new filters.
tests/unittests/agent_runtime/test_workspace.py	New unit tests covering resolver behavior plus client create/list integration paths.
tests/unittests/agent_runtime/test_model.py	Adds coverage for new/updated models and serialization aliases.
tests/unittests/agent_runtime/test_runtime.py	Updates list_all filter test to use system_tags and new filters.

Comments suppressed due to low confidence (4)

agentrun/agent_runtime/client.py:163

• Same issue as async create: workspace_name resolution ignores self.config when config arg is None (can resolve with wrong credentials/region), and the truthy check skips empty-string validation/clearing. Merge configs before resolving and treat empty strings as invalid (or normalize/strip then validate).

        if input.workspace_id and input.workspace_name:
            raise ValueError(
                "workspace_id and workspace_name are mutually exclusive; please"
                " only set one of them."
            )
        if input.workspace_name:
            input.workspace_id = resolve_workspace_id_by_name(
                input.workspace_name, config
            )
            input.workspace_name = None

agentrun/agent_runtime/client.py:405

• list_async workspace_name(s) resolution ignores self.config when config arg is None, which can resolve workspace IDs using the wrong account/region compared to the subsequent API call (which uses the client’s configured control_api). Also, if input.workspace_name: / if input.workspace_names: skips empty-string validation/normalization. Merge configs before resolving and validate non-empty strings.

            if input.workspace_id and input.workspace_name:
                raise ValueError(
                    "workspace_id and workspace_name are mutually exclusive;"
                    " please only set one of them."
                )
            if input.workspace_ids and input.workspace_names:
                raise ValueError(
                    "workspace_ids and workspace_names are mutually exclusive;"
                    " please only set one of them."
                )
            if input.workspace_name:
                input.workspace_id = await resolve_workspace_id_by_name_async(
                    input.workspace_name, config
                )
                input.workspace_name = None
            if input.workspace_names:
                input.workspace_ids = (
                    await resolve_workspace_ids_by_names_async(
                        input.workspace_names, config
                    )
                )
                input.workspace_names = None

agentrun/agent_runtime/client.py:463

• Same issue as list_async: workspace_name(s) resolution should use the effective config (merge self.config with the override config) so lookup runs under the same credentials/region as the API call, and it should not silently accept empty strings due to truthy checks.

            if input.workspace_id and input.workspace_name:
                raise ValueError(
                    "workspace_id and workspace_name are mutually exclusive;"
                    " please only set one of them."
                )
            if input.workspace_ids and input.workspace_names:
                raise ValueError(
                    "workspace_ids and workspace_names are mutually exclusive;"
                    " please only set one of them."
                )
            if input.workspace_name:
                input.workspace_id = resolve_workspace_id_by_name(
                    input.workspace_name, config
                )
                input.workspace_name = None
            if input.workspace_names:
                input.workspace_ids = resolve_workspace_ids_by_names(
                    input.workspace_names, config
                )
                input.workspace_names = None

agentrun/agent_runtime/__client_async_template.py:256

• Template: list_async workspace_name(s) resolution ignores self.config when config is None, and truthy checks skip empty-string validation. Merge configs before resolving and validate/normalize non-empty workspace_name(s) so the resolver runs under the same account/region as the API call.

        try:
            if input is None:
                input = AgentRuntimeListInput()

            if input.workspace_id and input.workspace_name:
                raise ValueError(
                    "workspace_id and workspace_name are mutually exclusive;"
                    " please only set one of them."
                )
            if input.workspace_ids and input.workspace_names:
                raise ValueError(
                    "workspace_ids and workspace_names are mutually exclusive;"
                    " please only set one of them."
                )
            if input.workspace_name:
                input.workspace_id = await resolve_workspace_id_by_name_async(
                    input.workspace_name, config
                )
                input.workspace_name = None
            if input.workspace_names:
                input.workspace_ids = (
                    await resolve_workspace_ids_by_names_async(
                        input.workspace_names, config
                    )
                )
                input.workspace_names = None

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[OhYee]

Review: PR #102 — SDK Model Alignment & workspace_name Support

整体质量不错，模型对齐清晰，测试覆盖充分（99%），CI 全绿。以下几点建议：

1. 🔴 Workspace 分页硬编码 page_size=50

_workspace.py 中 ListWorkspacesRequest(name=name, page_size="50") — 如果用户有超过 50 个 workspace，resolver 可能找不到目标 workspace（API 返回的是 name 模糊匹配还是精确匹配？如果是模糊匹配，50 条可能不够）。建议至少做翻页兜底，或明确文档说明限制。

2. 🟡 resolve_workspace_id_by_name 不继承 client 的 config

Copilot 已经指出了这个问题。client.py:103 和 __client_async_template.py:93 中 workspace 解析只用了方法级 config，没有合并 self.config。如果调用者不传 config 但 client 构造时传了，resolver 会用默认凭证，可能解析到错误的 workspace。

3. 🟡 create() 中 mutate 了 input 对象

input.workspace_name = None  # client.py

调用者传入的 input 对象被修改了，这可能出乎意料。建议做浅拷贝或只在内部变量上操作。

4. 🟡 Module-level cache 无 TTL / 大小限制

_RESOLVE_CACHE 是全局 dict，无过期无上限。长期运行的服务中如果 workspace 被重命名或删除，缓存会返回过时的 ID。建议考虑使用 functools.lru_cache 或加 TTL。

5. ℹ️ tags → system_tags 是 breaking change

PR description 已说明，确认这是有意为之。建议在 changelog 中明确标注 BREAKING，方便用户升级。

6. ℹ️ resolve_workspace_ids_by_names_async 串行解析

多个 workspace name 用 for 循环逐个 await，可以用 asyncio.gather 并行。非阻塞建议。

总体：模型对齐部分扎实，workspace_name 是好的 DX 改进。建议至少处理 #1 (分页) 和 #3 (input mutation)，其余可 follow-up。