Skip to content

v0.6.2 -- algorithms and core 0.6.1: README and npm text as on main, declarations built by tsup 9

Latest

Choose a tag to compare

@alexanderbering alexanderbering released this 15 Nov 11:14
· 5 commits to main since this release
e82267d

Only @zensation/algorithms and @zensation/core change version. Compared with their 0.6.0 tarballs, src, every
JavaScript file and every source map are byte-identical. What changes is the README, the npm description of
algorithms, and the type declarations: tsup 9 (#146) builds them differently, and every entry point still exports the
same names. The other packages stay at the versions 0.6.0 and 0.6.1 published, and the publish job skips them.

Package From To Why
@zensation/algorithms 0.6.0 0.6.1 README and npm description as on main: FSRS-style, the architecture counted as in paper version 9, the Node 22 note
@zensation/core 0.6.0 0.6.1 README as on main: FSRS-style, the Node 22 note, a Quick Start query that matches without an embedding provider

Changed

  • Type declarations of @zensation/algorithms and @zensation/core are built by tsup 9 (#146) instead of tsup 8. A
    sub-path's .d.ts now imports shared types from ./index.js instead of a hashed chunk, type-only exports are written
    export type { … }, and core's shared chunk has a new name. The exported names are the same at all 42 type entry
    points of algorithms and all 18 of core, and a consumer project with skipLibCheck: false compiles against the old
    and the new declarations, as ESM and as CommonJS. JavaScript and source maps are byte-identical to 0.6.0.
  • Package READMEs (#165): @zensation/core and @zensation/algorithms say that they need Node.js 22 or newer, and what
    npm installs on Node 20 instead (core@0.2.2 with algorithms@0.3.4). The core Quick Start recalls with a question
    that matches by wording, because it passes no embedding provider (the old query scored 0). The algorithms README counts
    the architecture as paper version 9 does and names ZenAI, the AI platform we ran in our own operation, as the origin of
    the modules. Both name the paper with its full title and its concept DOI.
  • README and docs/FAQ.md (#164, #166): the architecture is counted as paper version 9 counts it (seven memory layers, a
    MemoryCoordinator and nine foundational algorithms, plus six PMA components), the comparison sentence follows the
    paper, and the statements about production name our own operation, without claims about users or load.
  • Docs follow paper version 9 (Zenodo 10.5281/zenodo.23832288, 31 October 2026): the README, the package READMEs,
    llms.txt and docs/benchmarks.md give the input tokens in the answer step as 1/109.7 (version 8: 109.6), and describe the
    six head-to-head comparisons against Letta and Mem0 in version 9's words, as showing no significant difference.
    Where the text means this library's scheduler, it now says "FSRS-style": the scheduler follows FSRS in the form of its
    updates, with constants that are not fitted and an exponential forgetting curve. docs/benchmarks.md drops the LoCoMo
    row "multi-layer routing vs. flat single-layer baseline": version 9 withdraws that routing lead (item K-38), because it
    came from a run with placeholder embeddings. The npm description of @zensation/algorithms changes with its next release.
  • README: the benchmark section opens with what was measured (the Python-port arm, input tokens in the answer step only,
    building the memory not counted) and the versioned source, version 9 on Zenodo and arXiv:2604.23878v4; the paper block
    links arXiv v4. New near the top: versions at a glance (release, packages, research records) and three ways in
    (in the browser, across Claude Code sessions, re-running the numbers).

Added

  • Plugin 0.1.4: PRIVACY.md and the plugin README name both downloads from the npm registry, Claude Code's install of
    the plugin's lockfile and npx's fetch on first start; until now they named only the second, which the lockfile of
    0.1.3 made incomplete. The README's troubleshooting says what to do when /plugin reports that the packages the
    plugin lists are not installed (measured with Claude Code 2.1.308: the server still starts through npx). No
    change to what the plugin runs.
  • Plugin 0.1.3 runs @zensation/mcp@0.3.1, whose tools carry their titles in the annotations and which falls back to
    ~/.zenbrain/memory.db when ${user_config.db_path} reaches it unfilled. plugin.json gains termsOfServiceUrl,
    pointing at the Apache-2.0 licence that governs the use of the plugin and the server. The plugin folder now carries a
    package.json and package-lock.json for @zensation/mcp@0.3.1, every package with its npm source and sha512 hash,
    as Anthropic's plugin directory asks for its Verified badge ("Launcher lock missing"). The new CI step
    scripts/verify-plugin-pin.mjs fails when the launcher, the lock, package.json, the README and PRIVACY.md name
    different versions, when the launcher carries any npx option besides -y, when the pinned version is not on npm, or
    when the lock's hash differs from the tarball npm serves. PRIVACY.md writes the company name in running text as
    "ZenSation AI"; the structured author.name and owner.name fields keep "Zensation AI".