v0.3.0

• Source Tracking & Versioning:
  • Introduced Documentrix::Utils::Digests for SHA256 hashing of strings
    and files, featuring an mtime-based cache.
  • Implemented source-based document management including
    normalize_source, source_exist?, source_modified?, source_update,
    and source_remove.
  • Added support for digest in Documentrix::Documents#add and
    source_update to enable version tracking.
• Text Splitting:
  • Added Documentrix::Documents::Splitters::Common to implement
    force_split behavior.
  • Integrated force splitting into Character, RecursiveCharacter, and
    Semantic splitters to ensure chunks strictly respect chunk_size.
• Cache Enhancements:
  • Added each_source to Documentrix::Documents::Cache::Common and an
    optimized SELECT DISTINCT source implementation in SQLiteCache.
  • Implemented a SQLite trigger delete_embedding_after_record to
    automatically clean the embeddings table when records are deleted.
  • Made object_class a required keyword argument in
    Documentrix::Documents::RedisCache#initialize.
• Search & Retrieval:
  • Added min_similarity parameter to Documentrix::Documents#find and
    associated cache retrieval methods.
  • Optimized SQLiteCache#find_records by moving similarity calculations
    and filtering into the SQL query using 1 - vec_distance_cosine.
  • Streamlined Documentrix::Documents#find_where logic and improved
    utilization of max_records.
• Bug Fixes:
  • Fixed an issue in Documentrix::Documents#find where max_records was
    hardcoded to nil when calling the cache.