Skip to content

Releases: emergent-wisdom/meaning-model

Meaning Model v0.2.1

Choose a tag to compare

@github-actions github-actions released this 12 Sep 02:49

Application-defined modeling

  • Model-facing MCP guidance and documentation now explain how applications choose their own processes, categories, depth, and optional layers.
  • Structural starters are distinguished from experimental Story/Decision models. Existing template defaults are unchanged.
  • A runnable category-revision example demonstrates developing an application-specific vocabulary while preserving the model's validation and revision rules.
  • The revised paper and grammar clarify general-purpose use, continuity of understanding, and linked world, understanding, and artifact records. Storytelling remains the first worked application; learning benefits remain untested.

Installation and compatibility

Install or build the matching 0.2.1 engine when upgrading the MCP package. Relative to 0.2.0, engine modeling behavior and stored-data formats are unchanged; no data migration is required. npm users upgrading from 0.1.x should also read the 0.2.0 compatibility notes.

Installation guide · Full changelog · Revised paper · Grammar appendix

Native engines are provided for Apple Silicon and Intel macOS, Linux x64, and Windows x64, with adjacent SHA-256 files. Each platform must pass native Rust and MCP smoke checks before release.

Meaning Model v0.2.0

Choose a tag to compare

@github-actions github-actions released this 06 Sep 14:05

Meaning Model 0.2.0

This release makes progressive world construction easier to revise and provides verified native engines for macOS, Linux, and Windows.

Changes

  • Refine or explicitly revise a world after accepted events, with head checks and immutable before-and-after receipts.
  • Check that temporal detail recomposes its parent, including whether a partial expansion leaves a feasible remainder.
  • Install a version-matched native Rust engine with SHA-256 verification using --install-engine; ordinary startup does not download or build anything.
  • Reject invalid narrative uncertainty and attempts to overwrite reserved MCP command fields.
  • Include normalized Cuts, context roots, and temporal contracts in the relevant queries and estimation checks.
  • Add a changelog and a runnable progressive-authoring example.

Install

npx -y @emergent-wisdom/meaning-model-mcp@0.2.0 --install-engine

Then configure your MCP client to run npx with arguments -y and @emergent-wisdom/meaning-model-mcp@0.2.0.

Prebuilt engines support macOS 14+ (Apple Silicon and Intel), Linux x64 with glibc 2.35+, and Windows x64. Source builds remain available with --build-engine.

Compatibility

Use the matching 0.2.0 engine when upgrading. Existing models without the new optional temporal contracts retain their hashes; no automatic data conversion occurs. New processes require explicit current values, and revisions cannot change an existing process's shape or unit.

Histories spanning world revisions remain available in SQLite, and narrative authoring and rendering continue to work. Portable project/checkpoint and accepted-history training exports across revision boundaries are not yet supported.

The Book and paper PDFs are unchanged by this release.