Search

Darryl Ruggles
Darryl Ruggles @RDarrylR ·
When AI agents stop and restart, they typically lose everything they wrote to disk. Files, installed packages, build artifacts, all gone. AWS now has a managed session storage feature for Bedrock #AgentCore Runtime that solves this by persisting the #filesystem across sessions. You specify a mount path, your agent uses the filesystem normally, and the runtime handles replication to durable storage in the background. Resume a session and a new microVM picks up the exact state your agent left behind. No custom checkpoint code required. lckhd.eu/gcyUkC
231
たくど
たくど @d01tkum ·
連載第二回が公開されました。 AgentCore Observability のゼロコード計装の仕組みを、CDK for TypeScript での実機検証とともに深掘りしています。 いつも技術評論社さんには丁寧にご対応いただき感謝です。 #AWS #AmazonBedrock #AgentCore #OpenTelemetry #Observability #gihyo #技術評論社
gihyo.jp gihyo.jp @gihyojp ·
AgentCoreを運用する技術 ── 本番品質を支える3つの視点「CDK for TypeScriptで検証するAgentCore Observabilityゼロコード計装の仕組み」 by @d01tkum 公開 gihyo.jp/article/2026/0…
CDK for TypeScriptで検証するAgentCore Observabilityゼロコード計装の仕組み | gihyo.jp

連載第二回である今回は、そのアーキテクチャを支える技術基盤のひとつであるAgentCore Observabilityに焦点を当てます。

From gihyo.jp
3
1.9K
Darryl Ruggles
Darryl Ruggles @RDarrylR ·
Building voice agents that feel conversational depends heavily on transport latency. WebSocket has worked well for streaming, but real-time audio and video in browsers and mobile apps benefit from something closer to peer-to-peer and UDP-based delivery. Amazon Bedrock #AgentCore Runtime now supports #WebRTC alongside WebSocket for bidirectional streaming. It gives developers a low-latency option for voice agents with flexibility around TURN relay setup, whether managed through Kinesis Video Streams or self-hosted. lckhd.eu/dhXeGN
2
124
Darryl Ruggles
Darryl Ruggles @RDarrylR ·
I've started building more projects with Amazon Bedrock #AgentCore lately. As AI agents start handling more real-world tasks, the question of governance gets harder to ignore. AWS recently made #Policy in Amazon Bedrock AgentCore generally available. This is a big step in the right direction. How do you let agents use tools freely enough to be useful while keeping them within boundaries that security and compliance teams can actually trust? Having policies like this lets teams define agent-tool access rules in natural language, which then converts to Cedar policies enforced at a gateway layer. The key detail is that it works outside the agent code, so #governance stays independent from development. Separating policy from code is a small architectural choice that could save teams a lot of headaches down the road. lckhd.eu/7h0ypN
242
たけたか | AWS全冠×上流工程SE
たけたか | AWS全冠×上流工程SE @taketakablog ·
AIエージェントのツール呼び出しを「コード外」から制御するAmazon Bedrock AgentCore Policyが正式リリース。自然言語でポリシーを書くとCedarに自動変換され、エージェントのコードを一切変えずにガードレールを適用できます。 zenn.dev/taketaka/artic… #AWS #Bedrock #AgentCore #Cedar
Amazon Bedrock AgentCore Policyが正式リリース——AIエージェントに「コード外ガードレール」を適用する From zenn.dev
100