AWS Open-Sources ExtendDB for DynamoDB-Compatible Local Development
AWS has released ExtendDB as an open-source project — a DynamoDB-compatible adapter that sits on top of alternative storage backends, exposing the DynamoDB API and data model without requiring a live AWS connection. The primary use case is local development and testing: teams working with DynamoDB can run integration tests against ExtendDB using the same SDK calls, query patterns, and conditional expressions as production, and then deploy unchanged to DynamoDB in AWS. This closes a longstanding friction point where DynamoDB Local (the official Docker-based emulator) lagged behind production feature parity, particularly for newer expression types and stream configurations.
ExtendDB is especially relevant for CI pipelines that previously had to either provision real DynamoDB tables per test run or accept the emulator's feature gaps. The open-source release also means teams can contribute backend adapters — the architecture is designed to allow DynamoDB semantics on top of storage engines like SQLite, Postgres, or custom backends, which has potential value for embedded and edge scenarios where pulling in a full AWS dependency is undesirable.
AWS SAM CLI Now Supports CloudFormation Language Extensions Locally
The AWS SAM CLI has added local support for AWS CloudFormation Language Extensions, closing a significant gap between local development workflows and production deployments. Previously, dynamic references (like SSM Parameter Store resolve references), CloudFormation transforms, and macro-based constructs would either fail silently or require a live stack deployment to validate their behavior. With this update, developers can test these features locally with SAM CLI using the same semantics they'd see in AWS.
This matters particularly for teams that lean heavily on CloudFormation macros or use Secrets Manager and SSM Parameter Store dynamic references throughout their serverless templates. Being able to run sam local invoke or sam local start-api against templates that use these extensions without deploying removes a whole class of "works locally, breaks in CI" problems that have been a persistent complaint in the serverless community.
Kiro Web: AI-Assisted Development from Any Browser
AWS has launched Kiro Web, a browser-based interface for Kiro — its AI-powered development environment built around spec-driven development workflows. Kiro Web exposes the core Kiro capabilities (AI chat, spec generation, agent-driven implementation) without requiring installation of the desktop IDE. This makes Kiro accessible from shared machines, remote developer environments, or lightweight setups where a full desktop install isn't practical.
The spec-driven approach that distinguishes Kiro from chat-first tools is preserved in the web interface: developers write a requirements spec, Kiro generates a design document and implementation plan, and an agent executes the plan with visibility into each step. The web interface is a logical complement to Kiro's existing VS Code extension and desktop app, and positions it for scenarios like code review, quick edits, and onboarding workflows where spinning up the full IDE is overhead.
SageMaker Adds OpenAI-Compatible APIs for Zero-Code Migration
Amazon SageMaker AI now supports OpenAI-compatible APIs on inference endpoints, enabling teams to migrate workloads from OpenAI to SageMaker without changing SDK calls. By exposing the same request and response shapes as the OpenAI Chat Completions API, SageMaker inference endpoints become a drop-in replacement at the HTTP level — changing the base URL and credentials is all that's required for most integrations.
The practical impact is significant for organizations that have built internal tooling or integrated third-party libraries (LangChain, LlamaIndex, AutoGen) against the OpenAI interface and want to run foundation models on their own infrastructure for cost, latency, or data residency reasons. SageMaker's model catalog includes models from Anthropic, Meta, Mistral, and others, giving teams a range of options behind a stable API surface. Combined with the earlier addition of Bedrock-compatible APIs in OpenAI Codex, the industry is converging on the OpenAI API shape as the interoperability layer for LLM providers.
