1.6.2

Fixed API mismatch with minAppVersion causing a semantic tokenizer to never fire

1.6.1

v1.6.1: address community-plugin review (lint, license, release, warn…

1.6.0

v1.6.0: address Obsidian community-plugin review (lint, license, rele…

1.5.10

v1.5.10: scoped Restore-all + Forget learned preferences

1.5.9

Merge branch 'master' of https://github.com/MrDanRave/auto-linker

Merging online README edits

1.5.5

v1.5.5: three-way enable modal + rewritten README

- Semantic enable modal now offers Cancel / Use local model / Download, with an
  inline path field — no need to set the model path before flipping the toggle.
  Download enables with the default; Use local model enables from the given path.
- Rewrote README: what it does, why it's different (scored confidence vs binary),
  signals, features, privacy, the semantic tier (download vs local), install,
  usage, build. Added demo.gif placeholder at the top.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

1.4.4

v1.4.4 (Phase 6): rebalance weights, tighten fuzzy, skip fenced code

- Skip suggestions inside ```fenced code blocks``` (FENCE_PATTERN, checked
  before inline backticks; handles unclosed/EOF).
- Fuzzy tightened: min token length 5; 2 edits only for tokens >= 8 chars,
  else 1; never fuzzy-match into a common function word. Kills the absurd
  "renorm"->Record (d=2 on 6 chars) and "note"->NOT (fuzz into stop-word),
  while databse->Database and kuberentes->Kubernetes still match.
- Weights rebalanced; semantics demoted to a low refiner:
  lex .40, sig .25, case .15, graph .10, sem .05, accept .05.

Verified: 4 absurd cases rejected, 8 good matches survive, fenced-block skip,
at default sensitivity 55 (no need to crank sensitivity down).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

1.3.1

v1.3.1: skip suggestions inside backtick-delimited spans

- scoreRegion: add BACKTICK_PATTERN to skip ranges — a backtick to the next
  backtick, EOL, or EOF (inline code). Verified across closed/unclosed/multiline.
- SEMANTIC_PLAN: add Phase 7 "learned aliases" (positive mirror of the reject
  list, sourced from observed [[target|surface]] links; injects an index edge so
  e.g. value->valuable works with no stemmer change). Add design decision #12
  (self-link skip is an intentional, pedagogical property).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

v1.3.0 (Phase 4): three-bucket tokenizer + weighted-DAG index

Rework the matcher per SEMANTIC_PLAN v2:
- nlp.ts: bucket tokenizer (intra/phrase/anchor), atoms w/ anchor gaps,
  splitUnits, extractBase (digit-variant father). 802.1q stays one atom.
- TitleIndex: unit-based weighted bipartite index (byToken edges + byPhrase),
  base/child digit-strip edges, IDF weights. Replaces the flat inverted index.
- scoreRegion: n-gram + single-token candidates from the index (no prefix scan,
  no COVERAGE_THRESHOLD gate, no mid-word). lexScore = IDF-weighted coverage +
  exact-token bonus; significance folds frequency-list + IDF; case penalty -1.5.
- Dedup policies: merge corroborating adjacent anchors ("hub dumbo" -> one),
  overlap -> highest confidence (suppresses contained sub-spans).
- scanRegion deleted.
- Settings: per-vault tokenizer bucket editor + restore-defaults; rebuild on change.
- SCORING weights add `accept` (Phase 7); renormalize over available signals.

Verified against 13 acceptance cases incl. the original Hub1/"hub" bug,
kruger->Dunning-Kruger, 802.1q, Story of Love (phrase), user/root/mnt
(IDF self-suppression), and the "hub dumbo" merge.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

v1.2.4: remove eye-button tooltip, restore peek positioning

- Drop the eye button's aria-label ("Peek note") — Obsidian was rendering it
  as a dark hover tooltip that animated in behind the peek, which looked like
  a second window
- Restore the real peek positioning (below by default, flip above only when
  it won't fit) after the diagnostic build

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>