Skip to content

v0.2.0

Choose a tag to compare

View all tags
@sshanzel sshanzel released this 09 Jun 18:01
· 49 commits to main since this release
v0.2.0
98beb2c

First release since 0.1.0.

Ranking & learning

  • Principled-tuning engine: Beta-Bernoulli pitfall confidence, Salton association-strength co-change, personalized-PageRank blast radius, adaptive (anisotropy-aware) thresholds.
  • Blast enrichment + raw-feature persistence on the brain, and plex eval: an offline nDCG check with a READY / NOT YET / DEFAULTS-ALREADY-WIN verdict on whether a re-weight is worth it.

Review quality & UX

  • Findings presented as one Severity | Finding | Location table; no confidence or internal-thought display; report, don't decide.
  • The reviewer no longer narrates reconcile/outcome bookkeeping.
  • A rejected finding stops re-surfacing (reject now suppresses like waive/acknowledge).
  • One-line "embeddings are off" nudge so the learning layer is discoverable.

Setup & CLI

  • plex index/init are gated to a git repo; init no longer registers the MCP (the plugin provides it), so no duplicate server.
  • Install simplified to the plugin + npx @sshanzel/plex init; CLI docs moved to docs/cli.md.

Install: npm i -g @sshanzel/plex, or run ad-hoc with npx @sshanzel/plex <cmd>. The plugin tracks latest.

Assets 2
Loading