v0.28.0 — MIT license, MCP registry manifest, PyPI publishing
First release published to PyPI: uv tool install turbo-quant-memory
See CHANGELOG.md for the full entry.
Added
- MIT license. The repo previously shipped with none, so GitHub reported
"license": null— legally all-rights-reserved, which blocked corporate users and every directory listing. server.jsonfor the official MCP registry, validated against the 2025-12-11 schema.- PyPI publishing via Trusted Publishing (OIDC, no stored token), guarded against tag/version drift.
Changed
- The distribution is now
turbo-quant-memory(wasturbo-memory-mcp), matching the repository, the README title and the registry entry. Theturbo-memory-mcpcommand is unchanged, so existing MCP client configs keep working; aturbo-quant-memoryalias was added souvx turbo-quant-memory serveresolves. - READMEs rewritten, 349 → ~140 lines in all three languages; long-form sections moved to
docs/.
Fixed
- An overstated claim: the advertised "zero network code" proof command actually returned a hit. It is now anchored to import statements, and fastembed's one-time model download is named as the real exception.
- The
[torch]fallback told you topip install, which lands in the active venv rather than the uv-tool environment running the server.