English(EN) Governing Enterprise AI Traffic at Scale: Design and Operation of a Policy-Driven, Multi-Vendor LLM Gateway on Envoy

PayPal 详解 Cosmos LLM 网关，用于管理企业AI流量

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-27 04:00

一篇研究论文详细介绍了Cosmos LLM网关的设计与运行。该系统由PayPal开发，用于管理企业范围内对多个大型语言模型的访问。该网关基于Envoy构建，将控制决策与数据处理分离，以最大限度地减少延迟。它使用由Open Policy Agent评估的声明式策略进行治理，处理身份验证、凭证代理和预算评估。截至2026年9月，该系统已服务超过12,000名开发人员，支持数百个业务用例，日均处理约2000亿个token，覆盖超过100个模型，并具有高可用性。 AI

影响为管理多样化的LLM集成和大规模治理企业AI使用提供了蓝图。

排序理由该集群包含一篇详细介绍技术系统的研究论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 12 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍技术系统的研究论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: infra, product; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Same-day; Cluster formed today. Ranking reflects the current source set at time of score.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.AI TIER_1 English(EN) · Mithun Sasidharan, Srinivasan Manoharan, Rupesh Thakur, Abhinav TK, Ashwin Hegde, Anand Gautam · 2026-10-27 04:00

大规模企业人工智能流量治理：基于策略、多厂商的 Envoy LLM 网关设计与运营

arXiv:2610.35509v1 Announce Type: cross Abstract: Large organizations consume large language models (LLMs) from several vendors at once, each with its own API dialect, credential model, streaming protocol and pricing. Integrated directly, this heterogeneity yields credential spra…

本文如何被排名

报道来源 [1]

大规模企业人工智能流量治理：基于策略、多厂商的 Envoy LLM 网关设计与运营

相关实体

相关话题