doobidoo/mcp-memory-service v8.28.0 on GitHub

What's New in v8.28.0

Cloudflare Tag Filtering - AND/OR operations for tag searches with unified API contracts (#228)
- Added search_by_tags(tags, operation, time_start, time_end) to the storage base class and implemented it across SQLite, Cloudflare, Hybrid, and HTTP client backends
- Normalized Cloudflare SQL to use GROUP BY + HAVING COUNT(DISTINCT ...) for AND semantics while supporting optional time ranges
- Introduced get_all_tags_with_counts() for Cloudflare to power analytics dashboards without extra queries
Hybrid Storage Sync Performance Optimization - Dramatic initial sync speed improvement (3-5x faster)
- Performance Metrics:
  - Before: ~5.5 memories/second (8 minutes for 2,619 memories)
  - After: ~15-30 memories/second (1.5-3 minutes for 2,619 memories)
  - 3-5x faster initial sync from Cloudflare to local SQLite
- Optimizations:
  - Bulk Existence Check: get_all_content_hashes() method eliminates 2,619 individual DB queries
  - Parallel Processing: asyncio.gather() with Semaphore(15) for concurrent memory processing
  - Larger Batch Sizes: Increased from 100 to 500 memories per Cloudflare API call (5x fewer requests)

Unified Tag Search API - All backends now support consistent search_by_tags() interface with time filtering
Cloudflare Analytics - get_all_tags_with_counts() replaces multiple queries for dashboard tag statistics

No breaking changes - all updates are backward compatible. Existing tag search code continues to work.

Full Changelog: https://github.com/doobidoo/mcp-memory-service/blob/main/CHANGELOG.md#8280---2025-11-21