Releases: myscale/MyScaleDB
Releases · myscale/MyScaleDB
MyScaleDB v1.8.0
Features & Improvements
- Add support for multiple
distance()
functions in a single query. - Resolve score calculation error in distributed hybrid search.
- Improve filterd vector search performance by intrudcing a setting for parallel reading in
performPrefilter()
. - Enhance
TextSearch()
andHybridSearch()
to support multi-column FTS indexing.
Docker Image
docker pull myscale/myscaledb:1.8.0
MyScaleDB v1.7.1
Features & Improvements
- Add support for full-text search across multiple text columns.
- Optimize queries per second (QPS) for TextSearch during inserts.
- Correct BM25 calculation error in distributed text search.
- Implement parallel reading in performPrefilter().
- Optimize filter search for large datasets.
Fix
- Resolve error when performing hybrid search on distributed tables.
- Fix segmentation fault when executing parallel text search selects with FINAL.
- Address various bugs in the full-text search function.
- Fix a critical issue with removing the FTS index cache directory.
- Resolve the "failed to load Tantivy index file" error when deleting parts.
- Resolve the tantivy index loading error.
Docker Image
docker pull myscale/myscaledb:1.7.1
MyScaleDB v1.6.4
Features & Improvements
- Hybrid Search Refactor.
- Support basic paging operations for hybrid, vector, and text searches.
- Add parameters enable_nlq and operator to TextSearch() and HybridSearch().
- Add a query setting, enbale_fts_index_for_string_functions, to disable the FTS index for string functions by default.
- Add the icu multilingual tokenizer for the FTS index.
- Improve SELECT performance when PREWHERE includes the partition key.
- Optimize the logic for loading FTS indexes.
- Upgrade the runtime to Ubuntu 22.04.
- Enhance SELECT FINAL performance with two-stage vector search enabled.
- Fix TextSearch performance degradation after lightweight delete.
- Fix the vector search performance issue related to low filter ratios.
Fix
- Resolve the issue where ReplacingMergetree destroys the row IDs map data of the decoupled index
Docker Image
docker pull myscale/myscaledb:1.6.4
MyScaleDB v1.6
Features & Improvements
- Improve
SELECT
performance whenPREWHERE
includes the partition key
Fix
- Resolve the problem of mismatched checksums in FTS index replicas
- Resolve the problem of insufficient results returned by ScaNN index
- Fix the bug that the write speed limit does not work when writing data to s3
Docker Image
docker pull myscale/myscaledb:1.6
MyScaleDB v1.5
Features & Improvements
- Full-text search with BM25 scoring implemented with the Tantivy library
- Conduct an in-depth refactoring of the vector index
- Add default vector index type
Docker Image
docker pull myscale/myscaledb:1.5
MyScaleDB v1.4
Features & Improvements
- Add support for multiple column vector indexes.
- Refactor vector index manager.
Fixes
- Resolve an issue where canceling a vector index build causing the index to freeze.
- Address an issue where the vector index might not update properly when using the ReplicatedMergeTree engine.
Docker Image
docker pull myscale/myscaledb:1.4