✦ Features 13
- breakpersist subagent and swarm interrupts on the parent in TypeScriptmultiagenthil
- accept DecisionModel as HumanInTheLoop risk classifierhilinterventions
- support request-wide limits on Graph and Swarmmultiagentmodel
- add System One providersmodel
- add S3 Vectors search strategy for S3Storagepersistence
- add Model.info with capabilities and limits (TypeScript)model
- port edit tool to TypeScript, deprecate fileEditortool
- port write tool to TypeScripttool
- port read tool to TypeScripttool
- add graph tool in TypeScriptmultiagenttool
- use agent.cancel() to stop the event loop (TypeScript)languageagent
- add structured output attempt limitsstructured-outputagent
- default Bedrock maxTokens to the model info output limit (Python + TypeScript)model
⊘ Fixes 24
- add gpt-oss context windows and integ coverage (py + ts)contextmodel
- add Gemini 3.8 Flash context window and integ smoke tests (py + ts)contextmodel
- reject blank prompts before they enter agent historydevxagent
- add Gemini 4 Argon context window and integ smoke tests (py + ts)contextmodel
- add Bedrock Mistral context windows and integ smoke tests (py + ts)contextmodel
- add OpenAI chat-latest context window and integ smoke tests (py + ts)contextmodel
- add Bedrock Gemma 4 context windows and integ smoke tests (py + ts)contextmodel
- add Bedrock Grok 4.x context windows and integ smoke tests (py + ts)contextmodel
- add Bedrock OpenAI GPT 5.4/5.5/5.6/6/6.1 context windows (py + ts)contextmodel
- let providers enforce structured output when a model cannot be forcedmodelstructured-output
- report aborted tools as cancelledasynctool
- grant the integ role the new smoke-test Bedrock modelsmodel
- make integ test metrics and dashboards trustworthy
- grant the integ role the global Claude Haiku 4.5 profilemodel
- tighten URL scheme filtering in web fetch and diagram comment strippingtool
- run executeCode tests with node instead of pythonserver
- stop skips and shared TS names skewing flaky-test metrics
- preserve structured output before history offloadingcontextstructured-output
- never stash or serve offload previews as originals (py + ts)contextpersistence
- ignore pending tool execution when interrupt state is not activatedhil
- gate trusted tools on enableTrust and let live answers winhil
- recognize Bedrock OpenAI Daybreak model ids (py + ts)contextmodel
- add GPT-6.1 Terra context window (py + ts + cli)contextmodel
- keep Bedrock auto caching on for Claude ids the model info does not list (Python + TypeScript)model
⋯ Other 28
- bump re-actors/alls-green from 1.2.2 to 1.3.0
- bump hono from 4.13.8 to 4.13.13
- apply area-catalog label to PRs touching catalog entries
- bump actions/upload-artifact from 7.0.1 to 7.0.2
- bump actions/download-artifact from 8.0.1 to 8.0.2
- bump the production-minor group across 1 directory with 2 updates
- bump astral-sh/setup-uv from 10.0.1 to 10.3.0
- bump qs from 6.15.2 to 6.16.1
- run harness tests against the local SDK on SDK changes
- enforce TypeScript formatting
- one parametrized model smoke table per SDK for the 2026-10 catalog batchmodel
- close superseded release announcements
- add red teaming launch post for Strands Evals
- align tool deprecation guidance with strands-py warn rules
- stop referencing strands-agents-tools
- cover harness, CLI, and new SDK subsystems in issue classifier
- bump @aws-sdk/client-bedrock-runtime from 3.1152.0 to 3.1154.0 in the production-minor group across 1 directory
- bump actions/setup-python from 7.0.0 to 7.1.0
- bump astral-sh/setup-uv from 10.3.0 to 10.3.2
- paginate issue events and comments in auto-close
- bump zod from 4.6.5 to 4.6.7
- bump fast-uri from 3.1.8 to 3.1.9
- bump @modelcontextprotocol/sdk from 1.30.0 to 1.32.0
- read Node from .node-version and raise it to 22.23.4
- warn that onError proceed fails openinterventions
- pick Node for previews independently of the PR branch
- raise the TypeScript integ metrics test-name cap to 2000
- update dev dependencies and move tests to vitest 5