Skip to content

ROADMAP.md

Latest commit

 

History

History
170 lines (125 loc) · 5.7 KB

ROADMAP.md

File metadata and controls

170 lines (125 loc) · 5.7 KB
Raw

Mnemex Roadmap

This document outlines the development roadmap for Mnemex. For detailed implementation notes, see future_roadmap.md.

Version 1.0.0 (Released ✅)

Status: Production-ready, feature-complete

  • ✅ 11 MCP tools for memory management
  • ✅ Temporal decay with 3 models (power-law, exponential, two-component)
  • ✅ JSONL storage with in-memory indexing
  • ✅ Algorithmic memory consolidation
  • ✅ Unified search across STM and LTM
  • ✅ Git integration for backups
  • ✅ Obsidian vault integration
  • ✅ 7 CLI commands
  • ✅ Complete documentation suite
  • ✅ CI/CD with GitHub Actions

Version 1.1.0 (Planned - Q1 2026)

Focus: Stability, Testing, Security

High Priority

  • Security Hardening (#6)

    • Dependency scanning (Dependabot, safety, pip-audit)
    • Code security scanning (Bandit, Semgrep)
    • Supply chain verification (SBOM)
    • SECURITY.md policy

  • Fix mypy Type Checking (#1)

    • Fix 30+ type errors
    • Re-enable mypy in CI

  • Improve Test Coverage (#7)

    • Target: 80%+ coverage (currently 40%)
    • CLI tool tests
    • Integration tests
    • Error handling tests

  • Production Hardening (#8)

    • File corruption handling
    • Graceful degradation
    • File locking for concurrent access
    • Better logging
    • Configuration validation

Medium Priority

  • Platform Testing (#9)

    • Windows testing (community help needed)
    • Linux testing (community help needed)
    • Cross-platform bug fixes

  • Performance Optimizations (#4)

    • Benchmark suite
    • Tag/entity indexing
    • Embedding cache
    • Score caching

Recent Improvements (v0.6.6-dev)

Completed: 2025-11-14

Critical Bug Fixes

  • ✅ Fixed use_count=0 scoring bug causing new memories to be immediately GC-eligible
    • Changed formula from use_count^β to (use_count+1)^β
    • New memories now get grace period (baseline score ~1.0) instead of zero score

Search & Clustering Enhancements

  • ✅ Upgraded search.py with Jaccard similarity fallback
    • Matches clustering.py quality for consistent semantic search
    • Better results even without embeddings
  • ✅ Updated review candidate filtering to use text_similarity

Embeddings & Maintenance

  • ✅ Added backfill_embeddings MCP tool for batch embedding generation
  • ✅ Achieved 100% embedding coverage (171/171 memories)
  • ✅ Verified high-quality clustering with embeddings (10 clusters, cohesion 0.77-0.82)

Feature Planning

  • ✅ Created comprehensive auto-recall specification
  • ✅ Feature branch ready: feature/auto-recall-conversation

Version 0.7.0 (Planned - Q1 2026)

Focus: Natural Language Activation Phase 2

High Priority

  • Auto-Recall During Conversation (Spec created 2025-11-14)

    • Automatic memory search when discussing related topics
    • Silent reinforcement via observe_memory_usage
    • Contextual surfacing (subtle/interactive modes)
    • Cross-domain usage detection (Maslow effect)
    • Feature branch: feature/auto-recall-conversation
    • Spec: features/auto-recall-conversation.md
    • Implementation phases:
      1. Silent Reinforcement (MVP) - Background search + auto-reinforce
      2. Subtle Surfacing - Natural context injection
      3. Interactive Mode - User-controlled surfacing
      4. Cross-Domain Detection - Maslow effect tracking

  • Conversational Memory Review

    • Natural review prompts during conversation
    • "Memory check-in" mode for research topics
    • Batch reinforcement by project/tag

Version 1.2.0 (Planned - Q2 2026)

Focus: Advanced Features, User Experience

High Priority

  • Enhanced Spaced Repetition (#2)

    • ✅ Basic natural spaced repetition (v0.5.1 - DONE)
    • Review scheduling improvements
    • Review queue tool
    • Adaptive intervals (SM-2 inspired)

  • Adaptive Decay Parameters (#3)

    • Category-based decay profiles
    • Usage-pattern learning
    • Auto-detection from tags/content

Low Priority

  • LLM-Assisted Consolidation (#5)
    • Optional LLM-powered merge decisions
    • Semantic understanding for better merges
    • Opt-in feature

Version 2.0.0 (Future)

Focus: Advanced AI Features, Ecosystem Integration

  • Machine learning for decay parameter optimization
  • Multi-user support
  • API server mode
  • Plugins/extensions system
  • Integration with popular tools (Raycast, Alfred, etc.)
  • Mobile client support (iOS, Android)

Contributing

We welcome contributions! Priority areas:

  1. Platform Testing - Help test on Windows/Linux (#9)
  2. Security - Implement security hardening (#6)
  3. Testing - Increase coverage (#7)

See CONTRIBUTING.md for details.

Links

Last Updated: 2025-11-14 Current Version: 0.6.5 (Natural Language Activation + Spaced Repetition) Next Release: 0.7.0 (Q1 2026 - Auto-Recall & Conversational Review)