v0.1.3 Release

v0.1.3

Features

• verl training backend — Support verl as a training backend, with GRPO config, dataset/loop manager, LLM server, trainer, and a GSM8K math-agent example (#70); plus setup and launch documentation (#73). Fire all tasks in the batch concurrently instead of capping by semaphore (#75)
• slime training backend — Add SlimeRunner as the Python entry point for training (#56), with rollout/reward/gateway/trace integration and SGLang token-ID capture patches; Sample.metadata is passed as the agent payload verbatim
• Structured JSON logging — Attach JSON logging with sessionId to the root logger (#63)

Bug Fixes

• Client — Use a unified rate limiter for all ACR API calls (#51)
• Examples — Fix env setup flag (#53); standardize app naming and fix reward-format incompatibility (#39, #41)

Examples

• AppWorld example — New strands_appworld_agent example for AppWorld API interaction tasks (#38); adapt the official baseline system prompt and few-shot examples for good performance (#48)
• tau-bench example — New strands_taubench_agent example, using OpenAIModel for the assistant

Documentation

• Introduce a Starlight documentation site (#65)
• Add a CloudWatch session-logs skill (#71)
• Migrate docs from vLLMModel to standard OpenAIModel (#40)
• Update rLLM integration status and add training-example links (#45)

Full Changelog: v0.1.2...v0.1.3

v0.1.2 Release

v0.1.2

Features

• Async client API — invoke_async(), result_async(), done_async(), and run_batch_async() for non-blocking usage in asyncio event loops (#28)
• Automatic session cancellation — ACR sessions are automatically cancelled after result fetch, timeout, or error — no manual cleanup needed (#32)
• Automatic event loop lock update — RolloutClient detects when the async event loop has changed and updates its internal lock accordingly (#33)
• Flexible rollout return values — @rollout_entrypoint now accepts any JSON-serializable dict (no required keys) (#29)
• OfficeBench example — New strands_officebench_agent example for office automation tasks (#34)
• Configurable connection pool — max_pool_connections parameter for tuning boto S3/ACR connection limits (#31)

Bug Fixes

• Catch BaseException in rollout_entrypoint to properly handle CancelledError (#25)
• Fix vLLMModel bug and update math agent Dockerfile (#27)
• Fix context explosion from spammy Maven output in migration agent (#26)
• Add max_pool_connections to boto config, enable passing sampling_params to evaluation scripts (#30)

Documentation

• Add async API section to README, OfficeBench to project structure, and session cancellation details to AGENTS.md (#35)

v0.1.1 Release

Changelog (v0.1.0 → v0.1.1)

Features

• client: Add RolloutClient with S3 HEAD polling (#13)
• client: Add timeout and session cancellation to batch runs (#15)
• client: Add input_id attribute to RolloutFuture (#23)
• app: Return S3 result location in rollout response (#12)
• examples: Add code migration agent (#14)
• examples: Add evaluation script and config for migration agent (#19)

Refactoring

• app: Remove abstract method requirement from AgentCoreRLApp (#11)
• app: Remove SQS dependency, use S3-only result delivery (#17)
• app: Rename _training to _rollout and TrainingConfig to RolloutConfig (#18)
• app: Derive session_id from ACR header instead of _rollout payload (#22)

Bug Fixes

• app: Validate rollout config before launching background task (#20)
• Fix: Specify main build context in docker build script (#9)

CI

• Add GitHub Actions workflows for unit tests and pre-commit (#10)
• Add PyPI publish workflow and bump version to 0.1.1

Documentation

• Add AGENTS.md (#3)
• Improve project README (#4, #5)
• Remove veRL integration references from README (#16)
• Rewrite docs for client-side APIs and simplify codebase
• Migrate docs + examples to AgentCoreRLApp with vLLMModel (#21)
• Rewrite intro and section structure for clarity
• Annotate architecture diagram with SDK component mappings