Search “AWS”

AWS explains prompt caching patterns for using OpenAI models on Amazon Bedrock

AWS has published a tutorial explaining how to use prompt caching to reduce the cost of agents repeatedly reading context when calling OpenAI models on Amazon Bedrock, distinguishing between two approaches—implicit caching and explicit prompt_cache_breakpoint—and recommending their combined use. The article provides code examples for tool and instruction caching, rolling breakpoints across multi-turn history, adding tools mid-session, and shared prefixes across sub-agents, and states that in a six-turn workload on GPT-6 Astra, implicit caching reduces input costs by about 62%, while explicit rolling breakpoints reduce them by about 61%. Cacheable prefixes must be at least 1024 tokens, and cache writes are billed at 1.25 times the normal input price.

AWS team open-sources multi-agent system Deep Insight, breaking down five major challenges in production environments

The AWS Solutions Architecture team built and open-sourced the multi-agent system Deep Insight, which can transform CSV data and natural language prompts into cited .docx analysis reports. The system consists of eight agents—Coordinator, Planner, Supervisor, Coder, Validator, Reporter, Tracker, and others—collaborating in layers, controlling execution flow based on the Graph and Agents-as-Tools patterns of the Strands Agents SDK, and selecting Claude Haiku, Opus, and Sonnet by role. Production deployment relies on the Amazon Bedrock AgentCore runtime; LLM-generated code is executed in isolation within per-session ephemeral containers on AWS Fargate, accompanied by a Web UI and an Ops dashboard.

i14 Trains and Deploys the Laplace 3B Linear Attention Model on AWS Trainium2 Using Native PyTorch

i14 partnered with AWS to train and deploy Laplace 3B, based on Gated Linear Attention, on AWS Trainium2 using TorchNeuron Native. A single trn2.48xlarge achieves a training throughput of about 120,000 tokens/s. The model's recurrent state per sequence is fixed at 29.5 MB and does not grow with context length; inference reaches 11,544 output tokens/s at 1,280 concurrent sequences. The team also reworked Muon's all_to_all communication to cut per-GPU communication volume by 24x and raise training throughput by 1.63x.

AWS Publishes Agent Registry Runtime Tool Discovery Practice

AWS introduces the Agent Registry runtime tool discovery approach: agents use hybrid semantic and keyword search to find approved MCP servers and A2A agents, then instantly build clients from descriptors to execute requests. The sample code is based on Amazon Bedrock AgentCore Gateway and runtime, integrates an order management MCP server and two A2A agents for pricing and customer service, and demonstrates a return and refund scenario using pure semantic matching. The registry returns only connection metadata; invocation credentials must be obtained separately through Cognito OAuth or IAM, and new capabilities require no changes to orchestrator code.

Pydantic AI Compares Nine Python Agent Frameworks in 2026

Pydantic AI officially released a 2026 comparison of Python AI agent frameworks, evaluating nine frameworks item by item against criteria such as model switchability, durable execution after crashes, human approval, frontend and voice integration, MCP, and observability, and giving the applicable scenarios for each. The article states that Pydantic AI supports 30+ model providers, 8 durable execution engines (Temporal, DBOS, Prefect, Restate, AWS Lambda, Absurd, pgtask, Apache Airflow), 5 speech-to-speech providers, and 10 sandboxes, while also noting that it supports only Python, does not offer a hosted platform, and has fewer retrieval-related integrations than LangChain. The comparison also covers LangGraph/LangChain, OpenAI Agents SDK, Google ADK, Microsoft Agent Framework, Strands Agents, CrewAI, Claude Agent SDK, and Agno, and separately lists the two voice frameworks LiveKit Agents and Pipecat.

Building Efficient Agents with Knowledge Distillation on Amazon SageMaker AI

An AWS official blog post explains how to fine-tune a shopping agent on Amazon SageMaker AI using sequence-level knowledge distillation (SeqKD): using GLM-5.3 on Amazon Bedrock as the teacher model to generate trajectories, using the ShoppingBench scorer to retain only trajectories that solve the task, and then performing serverless SFT on the Qwen3.5-4B student model. On 324 held-out test questions, the distilled 4B student's average absolute success rate increased from 0.297 to 0.541, essentially on par with the teacher model and mid-sized frontier models; when the teacher's reasoning process was removed and only tool calls were learned, the average success rate dropped to 0.415.

Databricks uses Lakebase branching to create an independent database per PR: a CI/CD practice

Databricks describes Snorkel AI's CI/CD solution that uses Lakebase branching to create an independent read-write database for each open PR, reducing branch readiness time from over 30 minutes to under 60 seconds. Branches use copy-on-write; on a production dataset of about 2.5 TB, 100 concurrent PR branches store only 5 to 10 GB of data each for their own changes, and TTL automatically cleans them up after the PR ends. Production data is continuously synchronized to the main branch via AWS DMS; developers log in with Okta SSO, and the CI pipeline authenticates with a service principal and short-lived OAuth tokens.

Amazon SageMaker AI Multi-Turn Reinforcement Learning Practical Guide: Training gpt-oss-20b with a Wordle Example

AWS has published a practical guide for multi-turn reinforcement learning (MTRL) on Amazon SageMaker AI. Using Wordle as an example, it deploys the agent and reward function to Amazon Bedrock AgentCore, and the training job updates LoRA adapters via GRPO. In the example, gpt-oss-20b at default reasoning effort improved its solve rate on held-out words from 61% to 84%, at a training token cost of about $339 and an AgentCore runtime cost of about $79. The article also details failures such as two inverted reward design errors, token invalidation under concurrency, and tool calls degrading into plain text, and recommends evaluating under the decoding settings used in final serving.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.