Embedding Dimension Cache Fix
Fixed a critical bug where embedding dimensions were not being restored from the model cache, causing dimension mismatches (384 vs 768) across multiple storage instances.
Fixed
- Embedding Dimension Cache (#412): Fixed embedding dimension not being restored from model cache, causing dimension mismatches (384 vs 768) across multiple storage instances
- Added
_DIMENSION_CACHEdictionary alongside_MODEL_CACHEto track embedding dimensions - Store dimension when caching models (external API, ONNX, SentenceTransformer)
- Restore dimension when retrieving cached models
- Result: Consistent embedding dimensions across all storage instances
- Added
Installation
pip install mcp-memory-service==10.10.5Or upgrade from previous version:
pip install --upgrade mcp-memory-serviceFull Changelog
See CHANGELOG.md for complete version history.
Automatic PyPI Publication
This release will be automatically published to PyPI via GitHub Actions workflow "Publish and Test (Tags)".