SuperLocalMemory V2
The foundational architecture. Published on arXiv. 12,000+ users. The product shell that V3's mathematical engine runs inside.
The Foundation V3 Builds Upon
V2 established the product architecture: CLI, MCP server, 17+ IDE integrations, web dashboard, trust scoring, compliance framework, and adaptive learning. V3 added the mathematical retrieval engine — Fisher-Rao similarity, sheaf cohomology, and Langevin dynamics. Together, they form a complete memory system.
If you are using V2, your data is safe. Run slm migrate to upgrade to V3. A backup is created automatically and rollback is available for 30 days.
V2 Feature Highlights
These capabilities are fully preserved in V3. The V2 product shell runs the V3 mathematical engine.
Adaptive Learning
Three-phase LightGBM learning system that adapts retrieval ranking based on user feedback. Bootstrap → Active → Mature phases with automatic progression.
Bayesian Trust Scoring
Beta distribution trust per agent with evidence accumulation. Trust gates block low-trust agents from writing or deleting memories. Burst detection for anomaly alerting.
Multi-Profile Isolation
16+ independent memory contexts. Work, personal, client projects — each with isolated storage, graph, and learning state. Switch instantly.
GDPR & Compliance
Attribute-Based Access Control (ABAC), immutable SHA-256 audit chain, retention policies, full data export (Art. 15), and complete erasure (Art. 17).
Knowledge Graph
Recursive Leiden clustering for community detection. Entity extraction with alias resolution. Graph traversal for relational queries.
Web Dashboard
12-tab interactive dashboard with graph visualization, trust monitoring, learning progress, compliance audit, and profile management.
V2 Research Paper
"SuperLocalMemory: Privacy-Preserving Multi-Agent Memory with Bayesian Trust Defense Against Memory Poisoning"
Varun Pratap Bhardwaj, 2026 — Published on arXiv and Zenodo.
Ready to Upgrade?
V3 adds mathematical foundations on top of everything V2 built. Your data migrates seamlessly.