generated: '2026-07-17'
method: searched
source: https://github.com/ArenaX-Labs/rlmesh/blob/main/CHANGELOG.md
docs: https://docs.rlmesh.dev/en/v0.1.0-rc.2/changelog/
description: >-
  Changelog for the RLMesh Python package (PyPI). Follows Keep a Changelog 2.0.0
  and Semantic Versioning 2.0.0. The Rust crates are internal implementation
  detail with no separate user stability promise.
scheme: semver
changelog_format: keepachangelog-2.0.0
current_version: 0.1.0-rc.3
entries:
  - version: '0.1.0'
    date: Unreleased
    breaking: false
    highlights:
      - >-
        First release: RLMesh connects models to environments across process,
        dependency, and machine boundaries with a Gymnasium-style API.
      - Freezes the `rlmesh-wire-v1` protocol generation and seals the `2026.06` workflow edition.
      - >-
        Serve Gymnasium-style environments over local or remote gRPC
        (EnvServer, RemoteEnv, RemoteVectorEnv, `python -m rlmesh.serve`).
      - Evaluate models locally, remote, or in a sandbox (Model, RemoteModel, SandboxModel).
      - Tag-driven IO adapters, action chunking / batched prediction, PyTorch and JAX end to end.
      - Isolated sandboxes, live debug viewer, eval recording to a portable bundle, client timeouts.
  - version: '0.1.0-rc.3'
    date: '2026-07-06'
    breaking: false
    additions:
      - Grouped predicts fuse across environments when the model exposes a batched corner.
      - "`Model.allow_fusion` (default True) opts out of fused batched forwards."
      - Adapter advisories carry a severity tier (`info` vs `caution`).
    fixes:
      - Batched and chunked prediction now works in local `run()` evals, not only on the served wire path.