Skip to content

v1.1.0 — Operator-grade governed evolution for OpenClaw and Atlas

Latest
Latest

Choose a tag to compare

View all tags
@dddabtc dddabtc released this 13 Mar 21:30
· 2 commits to main since this release
v1.1.0
031f82d

Atlas Evolution v1.1.0

Atlas Evolution is a governed evolution layer for OpenClaw, Atlas Memory, and similar agent runtimes.

v1.1.0 is the first version that feels like an operator-grade local evolution system rather than a generic self-evolving-agent demo shell.

What this release adds

Runtime ingest and audit

  • typed runtime event schema
  • CLI ingest from file or stdin
  • local HTTP ingest endpoint
  • raw inbound envelope ledger
  • projected feedback ledger
  • inspect / report surfaces over the audit path

OpenClaw / Atlas integration

  • formal OpenClaw/Atlas contract
  • openclaw-import for realistic OpenClaw operator session artifacts
  • OpenClaw handoff context surfaced in reports
  • replayable handoff bundles for restart / interruption / handoff scenarios

Governance and operator control

  • governance metadata for readiness, risk, and rollback context
  • review queue with ready / risky / rollback-sensitive / blocked proposal buckets
  • reviewable promotion artifacts and dry-run support
  • restart-safe workflow state and resume

Why this release matters

Most projects in this space emphasize automatic skill evolution.
Atlas Evolution v1.1.0 goes deeper on the things serious operators actually need:

  • auditability
  • rollback-aware promotion
  • recovery after interruption
  • replayable handoff across people or machines
  • explicit operator review

Scope boundary

This release is still intentionally conservative:

  • local-first
  • no RL
  • no OPD
  • no cloud training
  • no blind self-modification

Docs

  • README.md
  • README_zh.md
  • CHANGELOG.md
  • docs/v1_1_milestone.md
  • docs/v1_2_productization_roadmap.md

Validation

  • python3 -m unittest discover -s tests -v
  • 19/19 tests passing locally at release time

中文说明

Atlas Evolution 是一个面向 OpenClaw、Atlas Memory 及同类 Agent runtime 的「受治理进化层」。

v1.1.0 是第一个真正开始具备 operator-grade 本地演化系统 气质的版本，而不是一个“会自动自进化”的演示壳。

这个版本新增了什么

Runtime ingest 与审计

  • typed runtime event schema
  • CLI ingest（文件 / stdin）
  • 本地 HTTP ingest endpoint
  • raw inbound envelope ledger
  • projected feedback ledger
  • inspect / report 审计面

OpenClaw / Atlas 接线面

  • formal OpenClaw/Atlas contract
  • openclaw-import：导入更真实的 OpenClaw operator session artifact
  • report 可展示 OpenClaw handoff context
  • replayable handoff bundles，支持重启 / 中断 / 接手后的重放

治理与 operator control

  • readiness / risk / rollback context 的治理元数据
  • ready / risky / rollback-sensitive / blocked 的 review queue
  • reviewable promotion artifacts + dry-run
  • restart-safe workflow state + resume

为什么这个版本重要

很多同类项目重点在“自动长 skill”。
Atlas Evolution v1.1.0 往更硬的方向走了一步：

  • 可审计
  • 可回滚
  • 中断后可恢复
  • 可跨人/跨机器 handoff
  • 有明确的人类 review 面

边界

这个版本仍然是保守设计：

  • local-first
  • 不做 RL
  • 不做 OPD
  • 不做云训练
  • 不做盲目自我修改
Assets 2
Loading