Skip to content

v0.5.1 - Natural Spaced Repetition

Choose a tag to compare

View all tags
@prefrontalsys prefrontalsys released this 20 Oct 22:07
· 182 commits to main since this release
Immutable release. Only release title and notes can be modified.
v0.5.1
793aa2f

🎯 Natural Spaced Repetition System

v0.5.1 introduces a conversation-based memory reinforcement system inspired by the "Maslow effect" - how concepts naturally strengthen when encountered across different contexts.

No flashcards. No explicit review sessions. Just natural conversation.

✨ New Features

observe_memory_usage Tool

Records when memories are actively used in conversation (not just retrieved), enabling:

  • Automatic reinforcement through usage
  • Cross-domain usage detection (tag Jaccard similarity <30%)
  • Strength boost for cross-domain usage (capped at 2.0)
  • Natural review priority calculation

Example:

observe_memory_usage(
    memory_ids=["jwt-pref-123"],
    context_tags=["api", "authentication", "backend"]
)
# Cross-domain detected (original tags: [security, jwt, preferences])
# Strength boosted: 1.0 → 1.1

Review Priority System

Memories in the "danger zone" (0.15-0.35 decay score) get highest priority using an inverted parabola curve:

  • Peak priority at midpoint (0.25 by default)
  • Targets memories most at risk of being forgotten
  • Automatically surfaces in search results

Blended Search Results

search_memory now automatically includes review candidates:

  • Default: 70% primary results, 30% review candidates
  • Review candidates filtered for relevance to query
  • Configurable via MNEMEX_REVIEW_BLEND_RATIO

Memory Model Extensions

Four new backward-compatible fields:

  • review_priority: 0.0-1.0 urgency score
  • last_review_at: Last reinforcement timestamp
  • review_count: Total reinforcement events
  • cross_domain_count: Cross-domain usage count

📚 Documentation Updates

  • README.md: Added Natural Spaced Repetition section
  • docs/architecture.md: 122-line deep dive on review system
  • docs/api.md: Complete observe_memory_usage documentation
  • docs/prompts/memory_system_prompt.md: Auto-Observe pattern examples
  • CLAUDE.md, AGENTS.md, GEMINI.md: Updated for v0.5.1

⚙️ Configuration

New environment variables:

MNEMEX_REVIEW_BLEND_RATIO=0.3           # 30% review candidates
MNEMEX_REVIEW_DANGER_ZONE_MIN=0.15      # Lower bound of danger zone
MNEMEX_REVIEW_DANGER_ZONE_MAX=0.35      # Upper bound of danger zone
MNEMEX_AUTO_REINFORCE=true              # Auto-reinforce on observe

🧪 Test Coverage

  • 24 new tests in tests/test_review.py
  • 99% coverage on src/mnemex/core/review.py
  • All 791 tests passing across entire codebase

📦 Installation

# Recommended: UV tool install
uv tool install git+https://github.com/simplemindedbot/mnemex.git@v0.5.1

# Or from PyPI (once published)
pip install mnemex==0.5.1

🔗 Related

  • PR #51: Natural Spaced Repetition System implementation
  • PR #52: Documentation updates for v0.5.1
  • Closes #2: Spaced Repetition System

📖 Learn More

Full Changelog: v0.5.0...v0.5.1

Assets 3
Loading