Releases
0.6.0
Compare
Sorry, something went wrong.
No results found
titusz
released this
22 Feb 10:29
Added
dirty write counter property on NphdIndex and ShardedIndex for caller-driven flush policies
(#16 )
"For Coding Agents" reference page in documentation with architecture map, decision dispatch tables,
constraints catalog, and task recipes
Fixed
Accept bytes and bytearray in pad_vectors() per-element loop (#14 )
Only count existing keys in dirty counter on remove() (#16 )
Generate API reference markdown from source for llms-full.txt instead of raw mkdocstrings directives
Changed
Reduce logging verbosity: default timer level from INFO to DEBUG, reserve INFO for aggregate
summaries like shard load/save counts (#15 )
Bump usearch-iscc minimum version to 2.24.2
Move docs overrides directory into docs/overrides for cleaner project root
Make shard_size, connectivity, expansion_add, expansion_search explicit parameters on
ShardedNphdIndex.__init__ (fixes griffe doc generation warnings)
Resolve all ty type checker warnings caused by upstream type stubs and type narrowing gaps
Match NphdIndex.search() signature to parent Index.search() (fixes LSP violation)
You can’t perform that action at this time.