UiPath Documentation
test-cloud
latest
false
Test Cloud 发行说明
  • Test Cloud
    • 2026
      • 2026 年 11 月
      • 2026 年 10 月
      • 2026 年 9 月
      • 2026 年 8 月
      • 2026 年 7 月
      • 2026 年 6 月
      • 2026 年 5 月
      • 2026 年 4 月
      • 2026 年 3 月
      • 2026 年 2 月
      • 2026 年 1 月
  • Test Cloud 公共部门版
  • Test Cloud 专用版
重要 :
新发布内容的本地化可能需要 1-2 周的时间才能完成。

2026 年 10 月

Test Cloud 2026 年 10 月发布说明。

2026 年 10 月 30 日​

Production Azure Front Door endpoint added to Automation Ops firewall requirements​

A production Azure Front Door endpoint is now included in the Open Automation Ops row of the Cloud firewall pages. If your organization restricts outbound traffic, allow the following domain so Automation Ops loads correctly:

  • https://stdadmprdcdn-e8fvhfgggwbbabbj.a01.azurefd.net

October 29, 2026​

Improvements to the tenant-level Consumables tab​

Added November 3, 2026: This feature was released on October 29, 2026 but was not included in the release notes at the time.

The tenant-level Consumables tab now also lists unit types with no allocation to the tenant, which show No units allocated. The Overview panel on each unit card also shows four summary cards: Consumed from allocation, Consumed from organization, Total consumed, and Available for Allocation.

Upcoming removal: gemini-3.6-flash​

Google plans to retire gemini-3.6-flash on 2026-11-19. gemini-3.8-flash is the recommended replacement.

有关模型当前在 UiPath Platform 上的可用性的信息,请参阅AI 功能和模型路由。有关更多信息,请参阅LLM 模型弃用时间线。

Upcoming change: low-code and coded agents share the same governance policies and LLM configurations​

Starting December 2, 2026, coded agents use the LLM configurations defined for low-code agents, and the agents settings in AI Trust Layer policies govern both low-code and coded agents. The change applies to the LLM calls that coded agents make through the AI Trust Layer. Existing coded agents need no code changes or redeployment.

On that date, the coded agents settings will be hidden from AI Trust Layer policies, LLM configurations, and Model hub, and will no longer be enforced.

警告:

This can be a breaking change for coded agents. Where the agents and coded agents settings in an AI Trust Layer policy differ, coded agents switch to the agents settings on December 2, 2026. Depending on the difference, coded agents can lose access to models through the AI Trust Layer, gain access to more models, lose guardrails and personally identifiable information (PII) masking, or fail because stricter guardrails apply to them.

Wherever the two sets of settings differ, coded agents change behavior:

  • Enable Agents controls both kinds of agents. In a policy where Enable Agents is disabled and Enable Coded Agents is enabled, coded agents can no longer call models through the AI Trust Layer, and coded agents that depend on those calls fail.
  • The agents guardrails apply to coded agents. Guardrails and PII masking configured only for coded agents no longer apply, and where the agents guardrails are stricter, coded agents can fail.
  • The Agents model controls on the Models tab decide which models coded agents can use, including the embedding models that only coded agents call. The Agent Builder in Studio Web does not list embedding models.
  • Coded agents use the agents LLM configurations for Design, Evaluate & Deploy. A configuration that exists only under coded agents, including a Bring Your Own Model (BYOM) configuration, is no longer used. If the LLM configurations for coded agents are already identical to those for low-code agents, the behavior of coded agents does not change.

Before December 2, 2026, check that the agents settings in each AI Trust Layer policy are how you want coded agents to behave: compare the product toggles, guardrails, and allowed models with the coded agents settings. To keep coded agents on their current settings:

  • Create an AI Trust Layer policy whose agents settings match your current coded agents settings, and deploy it to the users, groups, or tenants that use coded agents.
  • Re-create under agents any LLM configuration that exists only under coded agents.

For example, if your coded agents settings allow only GPT-5 and your agents settings allow all models, coded agents can use all models after the change. To keep the restriction, create two AI Trust Layer policies and deploy them by group:

  • Policy A: the agents settings allow only GPT-5. Deploy it to the groups that use coded agents.
  • Policy B: the agents settings allow all models. Deploy it to the groups that use low-code agents.

Use the same approach for differences in the product toggles and guardrails.

If you take no action, your agents settings apply to coded agents as they are. You can update your policies and their deployments at any time, including after December 2. After the change takes effect, check that your coded agents behave as expected. Pricing, licensing, and stock keeping units (SKUs) do not change.

October 28, 2026​

Insights now available in the European Union delayed region​

Insights is now available in the European Union delayed hosting region. For more information about regional availability, see Global cloud regions.

2026 年 10 月 23 日​

External applications in another organization now redirect you to sign in there​

Previously, if you were signed in to one organization and opened an external application registered in a different organization, and the application did not specify which organization to sign in to, the sign-in failed with an invalid_request error. You are now signed out of the current organization and redirected to the owning organization's sign-in page instead. After you sign in, you are returned to the external application. For more information about external applications, see External applications.

Organization members can now read external applications through the API​

Organization members without administrator rights can now read the external applications registered in their organization through the External Applications API, including application names, redirect URIs, scopes, and secret metadata such as creation and expiration dates. Secret values remain hidden. Creating, updating, and deleting external applications, managing their secrets, and listing, viewing, creating, updating, or deleting federated credentials remain restricted to organization administrators. For more information, see the External client API.

2026 年 10 月 20 日​

解决方案现已可在印度使用​

2026 年 10 月 21 日添加:此功能于 2026 年 10 月 20 日发布,但当时的发布说明中未包含此功能。

Solutions 现已在印度地区提供完整数据驻留功能。有关区域可用性的更多信息,请参阅全球云区域。

2026 年 10 月 16 日​

在新的 AI Trust Layer 策略中,用于审核的用于保存输入/输出的功能默认关闭​

默认情况下,新 AI Trust Layer 策略的审核日志中不再保存提示词输入和 LLM 生成的输出。创建策略时,系统将禁用“输入/输出保存”功能,因此,在您启用该策略之前,由该策略管理的 LLM 调用将显示在审核日志中,其输入和输出显示为“已策略阻止” 。现有策略保留其当前设置。

有关详细信息,请参阅查看审核日志和AI Trust Layer 策略设置。

2026 年 10 月 14 日​

已删除:Claude Sonnet 4​

UiPath Platform 上不再提供Claude Sonnet 4 ( anthropic.claude-sonnet-4-20250514-v1:0 )。在 Amazon Bedrock 停用该功能后,UiPath 已于 2026 年 10 月 13 日将其删除。Claude Sonnet 4.6 ( anthropic.claude-sonnet-4-6 ) 即可替换。

UiPath 托管模型会自动迁移到Claude Sonnet 4.6 ,您无需执行任何操作。将您自己的模型引入 AI Trust Layer 并引用Claude Sonnet 4 的配置将失效。如果您未将它们更新为Claude Sonnet 4.6 ,则它们将回退到 UiPath 托管模型。

有关模型当前在 UiPath Platform 上的可用性的信息,请参阅AI 功能和模型路由。有关更多信息,请参阅LLM 模型弃用时间线。

模型中心针对自带 (BYO) 配置的改进​

模型中心现在显示自带 (BYO) 配置的状态,使您可以检查已启用配置的状态,并允许您启用已禁用的配置。您还可以选择在删除 BYO 配置时显示文件夹列并查看更多上下文。有关详细信息,请参阅在模型中心中“管理模型”和“配置 LLM” 。

2026 年 10 月 13 日​

Snowflake 和 SQL Server 连接的中继支持​

Snowflake 连接器和 Database Hub 连接器中的 SQL Server 的私有 (Relay) 连接现在通过 Relay 内部部署执行程序运行。JDBC 驱动程序内置在执行程序中,因此不需要连接器库或依赖项目录。需要26.4.4或更高版本的中继客户端。Relay 仍不支持其他 Database Hub 数据库类型。

有关详细信息,请参阅支持的基于 TCP 的连接。

中继客户端 26.4.4​

中继客户端26.4.4添加以下内容:

  • relay restart --all和relay restart <id1>,<id2>一次重新启动多个 Relay 组,例如在提取新的 Relay 客户端存档文件后,并以已重新启动的组和失败的组的摘要结束。
  • relay describe显示中继客户端注册到的组织和租户名称,在启动失败后,以最后一个错误和恢复提示结束。--json输出会在现有last_error的基础上添加organization_name 、 tenant_name和hint 。

有关详细信息,请参阅命令参考。

2026 年 10 月 7 日​

针对自动获取 AI 和代理用量充值的监管策略​

组织管理员可以在 AI Trust Layer 中配置监管策略,允许用户自动获取 AI 和代理使用量充值,而无需每次都手动请求。此功能正在逐步推出,您的组织可能尚无法使用。

2026 年 10 月 5 日​

消耗品中心改进​

  • 在选定时段内无消耗的文件夹将不再显示在“文件夹消耗”列表中,从而在无需更改文件夹结构的情况下减少混乱。有关详细信息,请参阅了解消耗品中心。
  • 简化了“消耗品中心”中的状态指示器,移除了组织级别状态徽章、无分配或无消耗租户的租户状态徽章,并将消耗量达到或超过 70% 的租户的“关键”层级合并为“警告”。有关详细信息,请参阅了解消耗品中心。

任何 UiPath 服务可能需要 Cloudflare IP 范围​

任何 UiPath 服务可能都需要将 Cloudflare 104.16.0.0/13出站 IP 范围列入允许列表,而不仅仅是2026 年 7 月发布说明中宣布的 Document Understanding。如果您的防火墙限制出站流量,并且您的网络设备不支持基于 DNS 的允许列表,请允许此范围,以防止流量被阻止。建议使用基于 DNS 的允许列表,因为 Cloudflare IP 范围很宽并且可能会发生变化。

2026 年 10 月 1 日​

已移除:gpt-4o-2024-05-13​

gpt-4o-2024-05-13在 UiPath Platform 上不再可用。在 Azure OpenAI 停用该功能后,UiPath 已于 2026 年 9 月 30 日将其删除。替换项为gpt-5.1-2025-11-13 。

UiPath 托管模型已迁移到gpt-5.1-2025-11-13会自动生成,无需您执行任何操作。通过 AI Trust Layer 引入您自己的模型并引用gpt-4o-2024-05-13 的配置将失效。如果您尚未将其更新为gpt-5.1-2025-11-13 ,系统会回退使用 UiPath 托管模型。

有关模型当前在 UiPath Platform 上的可用性的信息,请参阅AI 功能和模型路由。有关更多信息,请参阅LLM 模型弃用时间线。

请自带防护机制订阅[预览]​

您现在可以使用组织自己的安全供应商订阅将自己的防护机制引入 AI Trust Layer。然后,您的提供商将根据您所在区域和供应商协议对智能体防护机制进行评估。在预览期间,您可以订阅 Azure AI 语言、Azure 内容安全或 Noma。

这可以帮助以下组织:无法向未经审核的提供程序发送提示内容;需要满足数据驻留要求;或者已为安全服务付费并希望在其智能体中重用该服务。您提供程序的护栏与 UiPath 托管的护栏相结合,并且以相同的方式工作,具有相同的作用域和操作:智能体开发者将它们应用于单个智能体,管理员在组织范围内强制执行。

每次评估都与运行评估的提供程序一起记录在智能体追踪记录中,因此您可以确定触发了哪些检查以及在何处触发。如果您的提供程序变得无法访问,每个护栏还可以回退到 UiPath 托管的等效护栏,因此其强制执行绝不取决于单个供应商保持可用。

有关详细信息,请参阅配置防护栏。

此页面有帮助吗?

连接

需要帮助? 支持(opens in new window)

想要了解详细内容? UiPath Academy(opens in new window)

有问题? UiPath 论坛(opens in new window)

保持更新