Highlights
This release introduces hybrid search and importance scoring, giving you more powerful ways to find
and prioritize memories.
New Features
Hybrid Search (memory_hybrid_search)
Combines keyword (full-text) and semantic (vector) search using Reciprocal Rank Fusion (RRF) for better
results than either method alone.
memory_hybrid_search(
query="authentication flow",
semantic_weight=0.6, # 60% semantic, 40% keyword
top_k=10
)
Importance Scoring System
Memories now have dynamic importance scores calculated from:
- Base importance (default 1.0, manually adjustable)
- Recency decay (exponential, 30-day half-life)
- Access frequency (logarithmic boost from usage)
New database columns: importance, last_accessed, access_count
Memory Boost (memory_boost)
Manually increase a memory's importance to make it rank higher:
memory_boost(memory_id=42, boost_amount=0.5) # small=0.25, medium=0.5, large=1.0
Sort by Importance
memory_list now accepts sort_by_importance=True to order results by calculated importance score instead of
date.
Improvements
- Hierarchy paths now included in memory results (hierarchy_path field)
- Cloud DB examples added to README (S3/R2/GCS/Azure configuration)
- Portable paths - configuration examples now use $HOME instead of hardcoded paths
Upgrading
The schema migration is automatic. On first run, new columns (importance, last_accessed, access_count) will
be added to existing databases with sensible defaults.
---
Full Changelog: https://github.com/spokV/memory-mcp/compare/0.1.1...v0.2.0