                 Changes in version 2.0.2 (2026-10-01)                  

  - Removed the dependency on 'text2vec', which pulled in 'rsparse' and
    'float' (scheduled for archival on CRAN). similarXKCD() now uses
    latent semantic analysis (TF-IDF + truncated SVD) computed with base
    R and 'Matrix' instead of 'GloVe' embeddings. The function signature
    and return value are unchanged.
  - updateConfig() now stores word vectors in ~/.RXKCD/word_vectors.rds;
    run it once after upgrading to rebuild the embeddings.
  - Query and corpus tokenization are now shared and strip punctuation.
  - Fixed updateConfig() failing on every second run while rebuilding
    the full-text search index, which also left searchXKCD() without an
    index.

                 Changes in version 2.0.1 (2026-04-15)                  

  - Full-text search with BM25 ranking in searchXKCD() and semantic
    similarity search in similarXKCD(), backed by a local 'DuckDB'
    cache.
