🦊GitLab Blog•最新收集於 19h
GitLab 將代理式交付帶入 Dedicated 雲端環境

💡在不離開核准區域的情況下執行程式碼審查、Pipeline 修復與安全代理。
⚡ 30-Second TL;DR
有什麼變化
AI Gateway 會在 GitLab Dedicated 單租戶環境中運作,支援資料落地與組織隔離需求。
為什麼重要
企業可採用代理式軟體交付,而不必將推論資料移出 GitLab Dedicated 已核准的租戶與區域。對具有嚴格資料分類規則或既有區域模型承諾的團隊而言,這能降低合規與治理阻力。
下一步行動
請 GitLab Dedicated 管理員將 AI Gateway 連接至核准的 Amazon Bedrock 模型,並在具資料分類的儲存庫中試行 Duo Agent Platform。
誰應關注:Enterprise & Security Teams
關鍵要點
- •AI Gateway 會在 GitLab Dedicated 單租戶環境中運作,支援資料落地與組織隔離需求。
- •客戶可連接自行控管的模型,包括部署在選定 AWS 區域中 Amazon Bedrock 的模型。
- •Duo Agent Platform 的自訂程式碼審查、失敗 Pipeline 修復與安全分析等功能,現在可在 Dedicated 部署模式下使用。
- •GitLab Dedicated 持續提供由 GitLab 管理的營運、客戶自帶加密金鑰、災難復原,以及核心服務每月 99.9% 可用性的目標。
🧠 深度解析
背景與延伸:來自公開資料,非原文內容。引用 20 個來源。
🔑 增強重點摘要
- •The AI Gateway is a standalone service designed to provide access to AI-native GitLab Duo features across GitLab.com, Self-Managed, and Dedicated instances, handling authentication and model selection for various LLM providers.
- •GitLab Duo Agent Platform functions as an AI orchestration layer, enabling asynchronous collaboration between developers and specialized AI agents with full context across the entire Software Development Lifecycle (SDLC), including code, issues, epics, merge requests, CI/CD pipelines, wikis, analytics, and security scans.
- •The platform supports multi-agent flows where numerous AI agents can collaborate in parallel on complex tasks, leveraging comprehensive SDLC context to automate workflows like custom code reviews, failed-pipeline fixes, and security analysis.
- •For self-managed deployments, customers have the option to host their own AI Gateway instance and language models, providing greater control over their AI infrastructure and model providers.
- •GitLab Dedicated for Government was launched in June 2024, offering a FedRAMP-compliant DevSecOps solution through a secure, single-tenant SaaS model, further addressing stringent compliance needs.
🛠️ 技術深入
- The AI Gateway automatically routes requests using Cloudflare and Google Cloud Platform (GCP) load balancers to the nearest available deployment, prioritizing low latency.
- It exposes proxy endpoints that forward requests directly to upstream LLM providers such as Anthropic, OpenAI, and Vertex AI, with access controllable via environment variables.
- GitLab Dedicated instances are deployed in a fully isolated tenant environment within a customer's chosen AWS region, residing in its own Virtual Private Cloud (VPC) with dedicated resource quotas.
- The architecture leverages GitLab Geo for disaster recovery, maintaining a replicated secondary site, and deploys Gitaly in a sharded setup across multiple virtual machines for repository management.
- For private connectivity, AWS PrivateLink is available, allowing secure connections between the customer's network and their GitLab Dedicated instance.
- The AI Gateway can be deployed as a sub-chart alongside a GitLab instance on Kubernetes, enabling GitLab Duo Self-Hosted and the GitLab Duo Agent Platform in self-managed environments.
🔮 前景展望AI analysis grounded in cited sources
Highly regulated industries will increasingly adopt single-tenant AI-powered DevSecOps platforms.
GitLab Dedicated's emphasis on data residency, isolation, and compliance (e.g., FedRAMP for government) directly addresses critical requirements for these sectors, making AI adoption safer and more feasible.
The market will see a consolidation of AI capabilities within comprehensive DevSecOps platforms rather than relying on disparate AI tools.
GitLab Duo Agent Platform's orchestration of multiple AI agents across the entire SDLC within a single platform reduces toolchain complexity and improves contextual understanding, driving a preference for integrated solutions.
AI-driven security and compliance automation will become a standard expectation in enterprise DevSecOps.
The Duo Agent Platform's capabilities for security analysis, vulnerability resolution, and false positive detection, combined with the secure environment of GitLab Dedicated, will raise the bar for automated security and compliance throughout the development pipeline.
⏳ 時間線
2011
GitLab started as an open-source project.
2014
GitLab Inc. was incorporated.
2022-11
GitLab Dedicated announced its limited availability as a single-tenant SaaS offering.
2024-04
GitLab Duo Chat reached General Availability, integrating AI capabilities into a natural language chat interface.
2024-08
GitLab Duo Enterprise add-on became generally available, extending AI capabilities across the SDLC for Ultimate customers.
2026-01
GitLab Duo Agent Platform achieved General Availability, introducing an AI orchestration layer for agentic AI across the software lifecycle.
📎 來源 (20)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
📰
AI 週報
閱讀本週精選 AI 大事摘要 →
👉相關動態
AI 策展新聞聚合。所有內容版權歸原始發布者所有。
原始來源: GitLab Blog ↗
每週 AI 簡報
每週一封,可隨時退訂。