Skip to content

v3.53.0 — Positioning rewrite + Honest comparison + AGPL

Choose a tag to compare

@Patdolitse Patdolitse released this 08 Jun 07:51
· 140 commits to main since this release

What's new

Governance-first positioning

README first screen now leads with the governance story — "Local-first AI memory you can see, edit, and override" — rather than just continuity. The cross-tool identity layer becomes the supporting pillar, not the only pitch. This matches how piia-engram is actually different from competitors.

Honest head-to-head comparison (docs/honest-comparison.md · docs/honest-comparison.zh-CN.md)

Bilingual (EN + ZH) narrative comparison vs mem0, Basic Memory, and ByteRover. Every competitor claim is dated and footnoted to their own public docs (snapshot 2026-06-08). Includes:

  • "Where they win" (honest acknowledgement of competitor strengths)
  • "Where we are weaker" (install friction, no GUI, semantic recall)
  • Honest decision guide ("when to use which")

Published metadata honesty correction

Removed overclaim phrases ("user-approved", "you decide what becomes permanent") from all four published metadata surfaces — .mcp/server.json, pyproject.toml, .claude-plugin/plugin.json, .cursor-plugin/plugin.json. Replaced with the honest capability frame: "AI proposes; high-risk items wait for your review; everything stays visible and reversible."

Guard test hardening

tests/test_public_positioning.py now explicitly bans the overclaim phrases and asserts the honest wording. 3045/3045 tests pass.

License: Apache-2.0 → AGPL-3.0-or-later

The open-source core is now AGPL-3.0-or-later. You can still use, modify, and self-host freely. Network or hosted redistribution must make the corresponding source available under the same license (AGPL §13). Releases ≤ 3.52.0 remain Apache-2.0.


本版新增

治理优先的定位重写

README 第一屏现在以治理故事开头——"全本地 AI 记忆,看得见、可改可删、可覆盖"——而不只是跨工具连续性。跨工具身份层成为底座支柱,不再是唯一卖点。这与 piia-engram 真正的差异化方向一致。

诚实头对头对比文(docs/honest-comparison.md · .zh-CN.md

中英双语叙事版对比 vs mem0、Basic Memory、ByteRover。每条竞品声明标日期、脚注到竞品自己的公开文档(快照 2026-06-08)。含"对方赢在哪"、"我们更弱的地方"和诚实选型指南。

已发布元数据诚实纠偏

从四个已发布元数据面移除过度承诺用语,改为诚实能力框架:"AI 建议;高风险先等你审查;全部可见可逆"。

许可证变更:Apache-2.0 → AGPL-3.0-or-later

开源核心改为 AGPL-3.0-or-later。可自由使用、修改、自部署;联网或托管形式的再分发须以相同许可证开放源代码。3.52.0 及更早版本仍为 Apache-2.0。