扩展 Gemini 企业智能体平台选择：推出 Parallel Web Search 验证功能
英文标题：Expanding Choice in Gemini Enterprise Agent Platform: Introducing Grounding with Parallel Web Search
Google Cloud 与 Parallel Web Systems 合作，将 Parallel 的搜索基础设施作为原生网络验证提供商集成到 Gemini 企业智能体平台。该集成使开发者能够将 AI 智能体锚定于可验证的实时网络结果，大幅提升事实准确性，并可通过 Gemini API、Agent Studio 和 Google Cloud Marketplace 便捷使用。开发者还可将搜索结果传递给其他大语言模型，支持目录充实、自主智能体和多智能体编排等高级工作流，对金融、法律等领域的生产级智能体部署具有重要意义。
11.7 KB
138.4 KB
18.4 KB
224.8 KB
80.6 KB
中文
扩展验证提供商选择：与 Parallel Web Systems 合作
随着生成式 AI 从简单的聊天界面演进到能够执行自动化“了解你的客户”（KYC）检查、产品目录数据充实、实时新闻分析和企业尽职调查等复杂任务的智能体，对事实准确性的需求前所未有地高。
为了让开发者更灵活地将 Gemini 模型锚定到可验证的实时信息上，今天我们兴奋地宣布，Parallel Web Systems 现已成为 Gemini 企业智能体平台的原生网络验证提供商。
通过 Parallel Web Search 进行验证将 Gemini 模型锚定在高质量的实时网络结果中。通过提供原始来源的精确引用，它确保你的智能体依据准确且可验证的信息行动。此功能现已在整个智能体平台中可用：可在 Gemini API 中调用、在 Agent Studio 中选择，并可通过 Google Cloud Marketplace 订阅，用量按你现有的 Google Cloud 发票计费。
“企业希望在一个开放的平台上使用最佳工具构建 AI 智能体，以交付真正的价值。通过将 Parallel 添加为 Gemini 企业智能体平台上的验证选项，客户现在可以使用他们首选的实时网络数据流让 Gemini 更加智能。这有助于他们在现有 Google Cloud 环境中，平稳地将智能体从原型推向部署。” – 马特·伦纳，Google Cloud 总裁兼首席营收官。
“AI 智能体很快将比人类以往任何时候都更多地使用网络，需要专门为其运行模式构建的搜索基础设施。我们创建 Parallel 正是为了提供这种基础设施，而与 Google Cloud 的这次合作，将其原生地带到企业正在构建生产级智能体的地方。” – 帕拉格·阿格拉瓦尔，Parallel 创始人兼首席执行官。
关于 Parallel Web Systems
Parallel Web Systems 构建了专门为智能体设计的搜索基础设施。Parallel 的 Search API 提供结构化、专为 LLM 优化的结果，由其专为智能体工作负载构建的私有网络索引驱动。通过与金融服务、法律和科技等行业的企业合作，Parallel 帮助大规模驱动生产级自主智能体工作流程。
将 Gemini 与 Parallel Web Search 连接
通过智能体平台上的 Parallel Web Search 进行验证，将 Gemini 模型连接到由 Parallel Web Systems 的搜索 API 提供的公共网络数据。这项服务将 Parallel Web Systems 的网络索引与 Gemini 原生能力相结合，能够理解并分解复杂的用户提示，从最相关的搜索结果中提炼洞察，并生成高度准确的答案，同时附有精确的引用注释。由于 Parallel Web Search 运行在 Google Cloud 上，你便可在自己现有的云环境中享受简化且安全的集成，并可为敏感工作负载选择零数据保留（ZDR）选项。
以更大的灵活性解锁复杂架构
将 Parallel 加入我们的生态系统，为客户提供了更灵活的许可与架构选择。通过 Parallel Web Search 进行验证的一个关键差异化优势在于其对新型用例的扩展支持方式。对于开发者而言，这意味着可以大规模执行编程调用、提取并缓存网络数据以丰富内部数据集，以及使用其他大语言模型（LLM）对 Parallel Web Search 的验证结果进行后处理。这种灵活性支持复杂架构，使企业能够自信地在生产环境中运行智能体，并将 Gemini 的网络验证能力集成到其现有技术栈中。
这种扩展的灵活性让开发者能够构建高度专业化的智能体工作流，例如：
- 产品目录与数据库充实： 提取并永久存储可验证的网络数据，以填补庞大内部系统中的空白。例如，管理大型产品库存的组织可以提取并永久存储网络数据，持续丰富目录属性；或利用搜索数据自动补充和更新庞大的供应商和联系人数据库。
- 自主智能体： 构建能持续导航系统、检索实时信息并执行复杂编程工作流的智能体。例如，有严格监管要求的组织可以构建自动将内部文档与公开域名进行交叉引用的智能体，或者在无需人工干预的情况下，基于实时网络数据以编程方式运行全球合规和风险评估检查。
- 多智能体编排： 构建复杂的多智能体系统，其中央编排器在多个模型间分配任务。由于通过 Parallel Web Search 进行验证允许开发者将搜索结果传递给其他 LLM，他们可以创建能够将上下文和查询无缝传递给由不同 LLM 驱动的各种任务特定子智能体的助手，从而交付高度综合的洞察。
立即开始
集成 Parallel Web Search 验证非常简单，并直接构建在你的智能体平台工作流程中。要开始使用，请通过 Google Cloud Marketplace 订阅 Parallel 服务，接受服务条款并查看定价。购买将计入你符合条件的商城福利，账单合并计入你的 Google Cloud 发票。
订阅后，导航至 Agent Studio。在导航菜单中，点击 + New 并选择 Chat。展开 Model settings 面板，点击 Grounding with Partners 开关以打开 Customize partner grounding 菜单。选择 Parallel Web Search 作为你的验证来源，应用该配置，你的 Gemini 模型便会立即准备好生成锚定于可验证实时网络信息的响应。
English / 原文
Expanding Grounding Provider Choice: Partnership with Parallel Web Systems
As generative AI evolves from simple chat interfaces to agents executing complex tasks like automated Know Your Customer (KYC) checks, catalog data enrichment, real-time news analysis, and corporate due diligence, the need for factual accuracy has never been higher.
To give developers more flexibility in how they ground Gemini models to verifiable, real-time information, today, we are excited to announce that Parallel Web Systems is now a natively integrated web grounding provider on Gemini Enterprise Agent Platform.
Grounding with Parallel Web Search anchors Gemini models in high-quality, real-time web results. By providing exact citations to original sources, it ensures your agents act on information that's accurate and verifiable. This capability is now available across Agent Platform: callable in the Gemini API, selectable in Agent Studio, and available to subscribe through Google Cloud Marketplace, with usage metered on your existing Google Cloud invoice.
"Enterprises want to build AI agents on an open platform with the best tools to deliver real value. By adding Parallel as a grounding option on our Gemini Enterprise Agent Platform, customers can now use their preferred live-web data stream to make Gemini smarter. This helps them smoothly take agents from prototype to deployment, all within their existing Google Cloud environment." – Matt Renner, President and Chief Revenue Officer, Google Cloud.
"AI agents will soon use the web far more than humans ever have, and need search infrastructure purpose-built for how they operate. We built Parallel to provide that infrastructure, and this partnership with Google Cloud brings it natively to where enterprises are building production agents." – Parag Agrawal, Founder and CEO, Parallel.
About Parallel Web Systems
Parallel Web Systems builds search infrastructure purpose-built for agents. Parallel’s Search API delivers structured, LLM-optimized results powered by a proprietary web index built specifically for agentic workloads. By working with enterprises across financial services, legal, and technology, Parallel helps power production-grade autonomous agent workflows at scale.
Connecting Gemini with Parallel Web Search
Grounding with Parallel Web Search on Agent Platform connects Gemini models to public web data provided by Parallel Web Systems' search API. This service combines Parallel Web System’s web index with Gemini's native ability to understand and decompose complex user prompts, distill insights from the most relevant search results, and generate highly accurate answers complete with precise citation annotations. Because Parallel Web Search runs on Google Cloud, you benefit from a streamlined, secure integration directly within your existing cloud environment, with an option for zero data retention (ZDR) available for sensitive workloads.
Unlocking complex architectures with expanded flexibility
Adding Parallel to our ecosystem empowers customers with expanded licensing and architectural flexibility. A key differentiator of Grounding with Parallel Web Search is its expansive approach to new use cases. For developers, this means the freedom to execute programmatic calls at scale, extract and cache web data to enrich internal datasets, and post-process Grounding with Parallel Web Search results using other LLMs. This flexibility supports complex architectures, allowing enterprises to confidently run agents in production and integrate Gemini's web-grounding capabilities into their existing tech stacks.
This expanded flexibility enables developers to build highly specialized agentic workflows, such as:
- Catalog & database enrichment: Extract and permanently store verifiable web data to fill gaps in massive internal systems. For example, organizations managing large product inventories can extract and permanently store web data to continuously enrich catalog attributes, or leverage search data to automatically supplement and update vast vendor and contact databases.
- Autonomous Agents: Build agents that continuously navigate systems, retrieve real-time information, and execute complex programmatic workflows. For instance, organizations with strict regulatory requirements can build agents that automatically cross-reference internal documents against public domains, or programmatically run global compliance and risk-assessment checks against live web data without human intervention.
- Multi-Agent Orchestration: Build sophisticated multi-agent systems where a central orchestrator routes tasks across different models. Because Grounding with Parallel Web Search allows developers to pass search results to other LLMs, they can create assistants that seamlessly pass context and queries to diverse, task-specific sub-agents powered by different LLMs to deliver highly synthesized insights.
Get Started Today
Integrating Grounding with Parallel Web Search is simple and built directly into your Agent Platform workflow. To get started, subscribe to the Parallel service via the Google Cloud Marketplace, accept the Terms of Service, and review the pricing. Purchases count towards your eligible marketplace benefits, with consolidated billing on your Google Cloud invoice.
After subscribing, navigate to Agent Studio. From the navigation menu, click + New and select Chat. Expand the Model settings pane and click the Grounding with Partners toggle to open the Customize partner grounding menu. Select Parallel Web Search as your grounding source, apply the configuration, and your Gemini model will be instantly ready to generate responses anchored in verifiable, real-time web information.