superlocalmemory 2.8.6 → 3.0.0

package/README.md

@@ -2,573 +2,258 @@
   <img src="https://superlocalmemory.com/assets/logo-mark.png" alt="SuperLocalMemory" width="200"/>
 </p>
 
-<h1 align="center">SuperLocalMemory</h1>
-<p align="center"><strong>Your AI Finally Remembers You</strong></p>
+<h1 align="center">SuperLocalMemory V3</h1>
+<p align="center"><strong>Information-Geometric Agent Memory with Mathematical Guarantees</strong></p>
 
 <p align="center">
-  <strong>Stop re-explaining your codebase every session. 100% local. Zero setup. Completely free.</strong>
+  The first agent memory system with mathematically grounded retrieval, lifecycle management, and consistency verification. Four-channel hybrid retrieval. Three operating modes. EU AI Act compliant.
 </p>
 
 <p align="center">
-  <a href="https://superlocalmemory.com"><img src="https://img.shields.io/badge/🌐_Website-superlocalmemory.com-ff6b35?style=for-the-badge" alt="Official Website"/></a>
+  <a href="https://superlocalmemory.com"><img src="https://img.shields.io/badge/Website-superlocalmemory.com-ff6b35?style=for-the-badge" alt="Website"/></a>
+  <a href="https://arxiv.org/abs/2603.02240"><img src="https://img.shields.io/badge/arXiv-2603.02240-b31b1b?style=for-the-badge&logo=arxiv&logoColor=white" alt="arXiv Paper"/></a>
+  <a href="https://zenodo.org/records/19038659"><img src="https://img.shields.io/badge/DOI-10.5281%2Fzenodo.19038659-blue?style=for-the-badge&logo=doi&logoColor=white" alt="V3 DOI"/></a>
 </p>
 
 <p align="center">
-  <a href="https://arxiv.org/abs/2603.02240"><img src="https://img.shields.io/badge/arXiv-2603.02240-b31b1b?style=for-the-badge&logo=arxiv&logoColor=white" alt="arXiv"/></a>
-  <a href="https://zenodo.org/records/18709670"><img src="https://img.shields.io/badge/DOI-10.5281%2Fzenodo.18709670-blue?style=for-the-badge&logo=doi&logoColor=white" alt="DOI"/></a>
-  <a href="https://zenodo.org/records/18709670"><img src="https://img.shields.io/badge/Zenodo-Research_Paper-1682D4?style=for-the-badge&logo=zenodo&logoColor=white" alt="Zenodo"/></a>
-  <a href="https://www.researchgate.net/publication/400976053"><img src="https://img.shields.io/badge/ResearchGate-Paper-00CCBB?style=for-the-badge&logo=researchgate&logoColor=white" alt="ResearchGate"/></a>
-</p>
-
-<p align="center">
-  <a href="https://www.python.org/downloads/"><img src="https://img.shields.io/badge/python-3.8+-3776AB?style=flat-square&logo=python&logoColor=white" alt="Python 3.8+"/></a>
+  <a href="https://www.python.org/downloads/"><img src="https://img.shields.io/badge/python-3.11+-3776AB?style=flat-square&logo=python&logoColor=white" alt="Python 3.11+"/></a>
   <a href="LICENSE"><img src="https://img.shields.io/badge/license-MIT-green?style=flat-square" alt="MIT License"/></a>
-  <a href="#"><img src="https://img.shields.io/badge/local--first-100%25-brightgreen?style=flat-square" alt="100% Local"/></a>
-  <a href="#"><img src="https://img.shields.io/badge/setup-5%20min-orange?style=flat-square" alt="5 Min Setup"/></a>
-  <a href="#"><img src="https://img.shields.io/badge/platform-Mac%20%7C%20Linux%20%7C%20Windows-blue?style=flat-square" alt="Cross Platform"/></a>
-  <a href="https://github.com/varun369/SuperLocalMemoryV2/wiki"><img src="https://img.shields.io/badge/📚_Wiki-Documentation-blue?style=flat-square" alt="Wiki"/></a>
-</p>
-
-<p align="center">
-  <a href="https://superlocalmemory.com"><strong>superlocalmemory.com</strong></a> •
-  <a href="#-quick-start">Quick Start</a> •
-  <a href="#-why-superlocalmemory">Why This?</a> •
-  <a href="#-features">Features</a> •
-  <a href="#-documentation">Docs</a> •
-  <a href="https://github.com/varun369/SuperLocalMemoryV2/issues">Issues</a>
-</p>
-
-<p align="center">
-  <b>A <a href="https://github.com/varun369">Qualixar</a> Product</b> · Created by <a href="https://github.com/varun369">Varun Pratap Bhardwaj</a> •
-  <a href="https://github.com/sponsors/varun369">💖 Sponsor</a> •
-  <a href="LICENSE">📜 MIT License</a>
+  <a href="#three-operating-modes"><img src="https://img.shields.io/badge/EU_AI_Act-Compliant-brightgreen?style=flat-square" alt="EU AI Act"/></a>
+  <a href="#"><img src="https://img.shields.io/badge/tests-1400+-brightgreen?style=flat-square" alt="1400+ Tests"/></a>
+  <a href="#"><img src="https://img.shields.io/badge/platform-Mac_|_Linux_|_Windows-blue?style=flat-square" alt="Cross Platform"/></a>
+  <a href="https://github.com/qualixar/superlocalmemory/wiki"><img src="https://img.shields.io/badge/Wiki-Documentation-blue?style=flat-square" alt="Wiki"/></a>
 </p>
 
 ---
 
-## Research Paper
+## What is SuperLocalMemory?
 
-**SuperLocalMemory: Privacy-Preserving Multi-Agent Memory with Bayesian Trust Defense Against Memory Poisoning**
+SuperLocalMemory gives AI assistants persistent, structured memory that survives across sessions. Unlike simple vector stores, V3 uses **information geometry** to provide mathematically grounded retrieval, automatic contradiction detection, and self-organizing memory lifecycle management.
 
-*Varun Pratap Bhardwaj, 2026*
+**Works with:** Claude, Cursor, Windsurf, VS Code Copilot, Continue, Cody, ChatGPT Desktop, Gemini CLI, JetBrains, Zed, and 17+ AI tools via MCP.
 
-The paper presents SuperLocalMemory's architecture for defending against OWASP ASI06 memory poisoning through local-first design, Bayesian trust scoring, and adaptive learning-to-rank — all without cloud dependencies or LLM inference calls.
+> **Upgrading from V2 (2.8.6)?** V3 is a complete architectural reinvention — new mathematical engine, new retrieval pipeline, new storage schema. Your existing data is preserved but requires migration. After installing V3, run `slm migrate` to upgrade your data. Read the [Migration Guide](docs/migration-from-v2.md) before upgrading. Backup is created automatically.
 
-| Platform | Link |
-|----------|------|
-| **arXiv** | [arXiv:2603.02240](https://arxiv.org/abs/2603.02240) |
-| **Zenodo** (CERN) | [DOI: 10.5281/zenodo.18709670](https://zenodo.org/records/18709670) |
-| **ResearchGate** | [Publication Page](https://www.researchgate.net/publication/400976053) |
-| **Research Portfolio** | [superlocalmemory.com/research](https://superlocalmemory.com/research) |
+### Key Results
 
-If you use SuperLocalMemory in your research, please cite:
-
-```bibtex
-@article{bhardwaj2026superlocalmemory,
-  title={SuperLocalMemory: Privacy-Preserving Multi-Agent Memory with Bayesian Trust Defense Against Memory Poisoning},
-  author={Bhardwaj, Varun Pratap},
-  year={2026},
-  eprint={2603.02240},
-  archivePrefix={arXiv},
-  primaryClass={cs.AI},
-  url={https://arxiv.org/abs/2603.02240}
-}
-```
+| Metric | Score | Context |
+|:-------|:-----:|:--------|
+| LoCoMo (Mode A, zero-LLM) | **62.3%** | Highest zero-LLM score. No cloud dependency. |
+| LoCoMo (Mode C, full) | **~78%** | Competitive with funded systems ($10M+) |
+| Math layer improvement | **+12.7pp** | Average gain from mathematical foundations |
+| Multi-hop improvement | **+12pp** | 50% vs 38% (math on vs off) |
 
 ---
 
-## What's New in v2.8 — "Memory That Manages Itself"
-
-**SuperLocalMemory now manages its own memory lifecycle, learns from action outcomes, and provides enterprise-grade compliance — all 100% locally on your machine.**
-
-### Memory Lifecycle Management (v2.8)
-Memories automatically transition through lifecycle states based on usage patterns:
-- **Active** — Frequently used, instantly available
-- **Warm** — Recently used, included in searches
-- **Cold** — Older, retrievable on demand
-- **Archived** — Compressed, restorable when needed
+## Three Operating Modes
 
-Configure bounds to keep your memory system fast:
-```bash
-# Check lifecycle status
-slm lifecycle-status
-
-# Compact stale memories
-slm compact --dry-run
-```
-
-### Behavioral Learning (v2.8)
-The system learns from what works:
-- Report outcomes: `slm report-outcome --memory-ids 1,5 --outcome success`
-- View patterns: `slm behavioral-patterns`
-- Knowledge transfers across projects automatically
-
-### Enterprise Compliance (v2.8)
-Built for regulated environments:
-- **Access Control** — Attribute-based policies (ABAC)
-- **Audit Trail** — Tamper-evident event logging
-- **Retention Policies** — GDPR erasure, HIPAA retention, EU AI Act compliance
-
-### New MCP Tools (v2.8)
-
-| Tool | Purpose |
-|------|---------|
-| `report_outcome` | Record action outcomes for behavioral learning |
-| `get_lifecycle_status` | View memory lifecycle states |
-| `set_retention_policy` | Configure retention policies |
-| `compact_memories` | Trigger lifecycle transitions |
-| `get_behavioral_patterns` | View learned behavioral patterns |
-| `audit_trail` | Query compliance audit trail |
+| Mode | What | LLM Required? | EU AI Act | Best For |
+|:----:|:-----|:-------------:|:---------:|:---------|
+| **A** | Local Guardian | No | Compliant | Privacy-first, air-gapped, enterprise |
+| **B** | Smart Local | Local only (Ollama) | Compliant | Enhanced quality, data stays local |
+| **C** | Full Power | Cloud (optional) | Partial | Maximum accuracy, research |
 
-### Performance
-
-| Operation | Latency |
-|-----------|---------|
-| Lifecycle evaluation | Sub-2ms |
-| Access control check | Sub-1ms |
-| Feature vector (20-dim) | Sub-5ms |
-
-**Upgrade:** `npm install -g superlocalmemory@latest` — All v2.7 behavior preserved, zero breaking changes.
-
-[Upgrading to v2.8](https://github.com/varun369/SuperLocalMemoryV2/wiki/Upgrading-to-v2.8) | [Full Changelog](CHANGELOG.md)
+**Mode A** is the only agent memory that operates with **zero cloud dependency** while achieving competitive retrieval accuracy. All data stays on your device. No API keys required.
 
 ---
 
-<details>
-<summary><strong>Previous: v2.7 — "Your AI Learns You"</strong></summary>
-
-SuperLocalMemory learns your patterns, adapts to your workflow, and personalizes recall — all 100% locally. No cloud. No LLM. Your behavioral data never leaves your device.
-
-- **Adaptive Learning** — Learns tech preferences, project context, and workflow patterns
-- **Three-Phase Ranking** — Baseline → Rule-Based → ML Ranking (gets smarter over time)
-- **Privacy by Design** — Learning data stored separately, one-command GDPR erasure
-- **3 New MCP Tools** — Feedback signal, pattern transparency, and user correction
-
-</details>
-
-<details>
-<summary><strong>Previous: v2.6.5 — Interactive Knowledge Graph</strong></summary>
-
-- Fully interactive visualization with zoom, pan, and click-to-explore
-- 6 layout algorithms, smart cluster filtering, 10,000+ node performance
-- Mobile & accessibility support: touch gestures, keyboard nav, screen reader
-
-</details>
-
-<details>
-<summary><strong>Previous: v2.6 — Security & Scale</strong></summary>
-
-## What's New in v2.6
-
-SuperLocalMemory is now **production-hardened** with security, performance, and scale improvements:
-
-- **Trust Enforcement** — Bayesian scoring actively protects your memory. Agents with trust below 0.3 are blocked from write/delete operations.
-- **Profile Isolation** — Memory profiles fully sandboxed. Zero cross-profile data leakage.
-- **Rate Limiting** — Protects against memory flooding from misbehaving agents.
-- **HNSW-Accelerated Graphs** — Knowledge graph edge building uses HNSW index for faster construction at scale.
-- **Hybrid Search Engine** — Combined semantic + FTS5 + graph retrieval for maximum accuracy.
-
-**v2.5 highlights (included):** Real-time event stream, WAL-mode concurrent writes, agent tracking, memory provenance, 28 API endpoints.
-
-**Upgrade:** `npm install -g superlocalmemory@latest`
-
-[Interactive Architecture Diagram](https://superlocalmemory.com/architecture.html) | [Architecture Doc](docs/ARCHITECTURE.md) | [Full Changelog](CHANGELOG.md)
-
-</details>
-
----
-
-## The Problem
-
-Every time you start a new Claude session:
+## Architecture
 
 ```
-You: "Remember that authentication bug we fixed last week?"
-Claude: "I don't have access to previous conversations..."
-You: *sighs and explains everything again*
+Query  ──►  Strategy Classifier  ──►  4 Parallel Channels:
+                                       ├── Semantic (Fisher-Rao graduated similarity)
+                                       ├── BM25 (keyword matching, k1=1.2, b=0.75)
+                                       ├── Entity Graph (spreading activation, 3 hops)
+                                       └── Temporal (date-aware retrieval)
+                                                    │
+                                       RRF Fusion (k=60)
+                                                    │
+                                       Scene Expansion + Bridge Discovery
+                                                    │
+                                       Cross-Encoder Reranking
+                                                    │
+                                       ◄── Top-K Results with channel scores
 ```
 
-**AI assistants forget everything between sessions.** You waste time re-explaining your:
-- Project architecture
-- Coding preferences
-- Previous decisions
-- Debugging history
+### Mathematical Foundations (Novel Contributions)
 
----
-
-## The Solution
+1. **Fisher-Rao Retrieval Metric** — Similarity scoring derived from the Fisher information structure of diagonal Gaussian families. Graduated ramp from cosine to Fisher-information-weighted scoring over the first 10 accesses per memory.
 
-```bash
-# Install in one command
-npm install -g superlocalmemory
-
-# Save a memory
-superlocalmemoryv2-remember "Fixed auth bug - JWT tokens were expiring too fast, increased to 24h"
-
-# Later, in a new session...
-superlocalmemoryv2-recall "auth bug"
-# ✓ Found: "Fixed auth bug - JWT tokens were expiring too fast, increased to 24h"
-```
+2. **Sheaf Cohomology for Consistency** — Algebraic topology detects contradictions between facts by computing coboundary norms on the knowledge graph. Non-trivial restriction maps amplify disagreements along discriminative subspaces.
 
-**Your AI now remembers everything.** Forever. Locally. For free.
+3. **Riemannian Langevin Lifecycle** — Memory positions evolve on the Poincare ball via a discretized Langevin SDE. Frequently accessed memories stay near the origin (ACTIVE); neglected memories diffuse toward the boundary (ARCHIVED). The potential is modulated by access frequency, age, and importance.
 
 ---
 
-## 🚀 Quick Start
+## Quick Start
 
-### Install (One Command)
+### Install via npm (recommended)
 
 ```bash
 npm install -g superlocalmemory
+slm setup
 ```
 
-Or clone manually:
-```bash
-git clone https://github.com/varun369/SuperLocalMemoryV2.git && cd SuperLocalMemoryV2 && ./install.sh
-```
-
-Both methods auto-detect and configure **17+ IDEs and AI tools** — Cursor, VS Code/Copilot, Codex, Claude, Windsurf, Gemini CLI, JetBrains, and more.
-
-### Verify Installation
+### Install via pip
 
 ```bash
-superlocalmemoryv2-status
-# ✓ Database: OK (0 memories)
-# ✓ Graph: Ready
-# ✓ Patterns: Ready
+pip install superlocalmemory
+# or with all features:
+pip install "superlocalmemory[full]"
 ```
 
-**That's it.** No Docker. No API keys. No cloud accounts. No configuration.
-
-### Launch Dashboard
+### First Use
 
 ```bash
-# Start the interactive web UI
-python3 ~/.claude-memory/ui_server.py
+# Store a memory
+slm remember "Alice works at Google as a Staff Engineer"
 
-# Opens at http://localhost:8765
-# Features: Timeline, search, interactive graph, statistics
-```
+# Recall
+slm recall "What does Alice do?"
 
----
-
-## 💡 Why SuperLocalMemory?
-
-### For Developers Who Use AI Daily
-
-| Scenario | Without Memory | With SuperLocalMemory |
-|----------|---------------|----------------------|
-| New Claude session | Re-explain entire project | `recall "project context"` → instant context |
-| Debugging | "We tried X last week..." starts over | Knowledge graph shows related past fixes |
-| Code preferences | "I prefer React..." every time | Pattern learning knows your style |
-| Multi-project | Context constantly bleeds | Separate profiles per project |
-
-### Built on Peer-Reviewed Research
-
-Not another simple key-value store. SuperLocalMemory implements **cutting-edge memory architecture** backed by peer-reviewed research — hierarchical organization, knowledge graph clustering, identity pattern learning, multi-level retrieval, adaptive re-ranking, workflow sequence mining, temporal confidence scoring, and cold-start mitigation.
-
-**The only open-source implementation combining all these approaches — entirely locally.**
+# Check status
+slm status
 
-[Read the paper →](https://zenodo.org/records/18709670)
-
----
-
-## ✨ Features
-
-### Multi-Layer Memory Architecture
-
-**[View Interactive Architecture Diagram](https://superlocalmemory.com/architecture.html)** — Click any layer for details, research references, and file paths.
-
-```
-┌─────────────────────────────────────────────────────────────┐
-│  Layer 9: VISUALIZATION (v2.2+)                             │
-│  Interactive dashboard: timeline, graph explorer, analytics │
-├─────────────────────────────────────────────────────────────┤
-│  Layer 8: HYBRID SEARCH (v2.2+)                             │
-│  Combines: Semantic + FTS5 + Graph traversal                │
-├─────────────────────────────────────────────────────────────┤
-│  Layer 7: UNIVERSAL ACCESS                                  │
-│  MCP + Skills + CLI (works everywhere)                      │
-│  17+ IDEs with single database                              │
-├─────────────────────────────────────────────────────────────┤
-│  Layer 6: MCP INTEGRATION                                   │
-│  Model Context Protocol: 18 tools, 6 resources, 2 prompts   │
-│  Auto-configured for Cursor, Windsurf, Claude               │
-├─────────────────────────────────────────────────────────────┤
-│  Layer 5½: ADAPTIVE LEARNING (v2.7 — NEW)                   │
-│  Three-layer learning: tech prefs + project context + flow  │
-│  Local ML re-ranking — no cloud, no telemetry               │
-├─────────────────────────────────────────────────────────────┤
-│  Layer 5: SKILLS LAYER                                      │
-│  7 universal slash-commands for AI assistants               │
-│  Compatible with Claude Code, Continue, Cody                │
-├─────────────────────────────────────────────────────────────┤
-│  Layer 4: PATTERN LEARNING                                  │
-│  Confidence-scored preference detection                     │
-│  "You prefer React over Vue" (73% confidence)               │
-├─────────────────────────────────────────────────────────────┤
-│  Layer 3: KNOWLEDGE GRAPH + HIERARCHICAL CLUSTERING         │
-│  Auto-clustering: "Python" → "Web API" → "Auth"            │
-│  Community summaries with auto-generated labels             │
-├─────────────────────────────────────────────────────────────┤
-│  Layer 2: HIERARCHICAL INDEX                                │
-│  Tree structure for fast navigation                         │
-│  O(log n) lookups instead of O(n) scans                     │
-├─────────────────────────────────────────────────────────────┤
-│  Layer 1: RAW STORAGE                                       │
-│  SQLite + Full-text search + vector search                  │
-│  Compression: 60-96% space savings                          │
-└─────────────────────────────────────────────────────────────┘
+# Switch modes
+slm mode a   # Zero-LLM (default)
+slm mode b   # Local Ollama
+slm mode c   # Full power
 ```
 
-### Key Capabilities
+### MCP Integration (Claude, Cursor, etc.)
 
-- **[Adaptive Learning System](https://github.com/varun369/SuperLocalMemoryV2/wiki/Learning-System)** — Learns your tech preferences, workflow patterns, and project context. Personalizes recall ranking using local ML. Zero cloud dependency. *New in v2.7*
-- **[Knowledge Graphs](https://github.com/varun369/SuperLocalMemoryV2/wiki/Knowledge-Graph-Guide)** — Automatic relationship discovery. Interactive visualization with zoom, pan, click.
-- **[Pattern Learning](https://github.com/varun369/SuperLocalMemoryV2/wiki/Pattern-Learning-Explained)** — Learns your coding preferences and style automatically.
-- **[Multi-Profile Support](https://github.com/varun369/SuperLocalMemoryV2/wiki/Multi-Profile-Workflows)** — Isolated contexts for work, personal, clients. Zero context bleeding.
-- **[Hybrid Search](https://github.com/varun369/SuperLocalMemoryV2/wiki/Advanced-Search)** — Semantic + FTS5 + Graph retrieval combined for maximum accuracy.
-- **[Visualization Dashboard](https://github.com/varun369/SuperLocalMemoryV2/wiki/Visualization-Dashboard)** — Web UI for timeline, search, graph exploration, analytics.
-- **[Framework Integrations](docs/FRAMEWORK-INTEGRATIONS.md)** — Use with LangChain and LlamaIndex applications.
-- **[Real-Time Events](https://github.com/varun369/SuperLocalMemoryV2/wiki/Real-Time-Event-System)** — Live notifications via SSE/WebSocket/Webhooks when memories change.
-- **[Memory Lifecycle](https://github.com/varun369/SuperLocalMemoryV2/wiki/Memory-Lifecycle)** — Automatic state transitions (Active → Warm → Cold → Archived) with bounded growth guarantees. *New in v2.8*
-- **[Behavioral Learning](https://github.com/varun369/SuperLocalMemoryV2/wiki/Behavioral-Learning)** — Learns from action outcomes, extracts success/failure patterns, transfers knowledge across projects. *New in v2.8*
-- **[Enterprise Compliance](https://github.com/varun369/SuperLocalMemoryV2/wiki/Enterprise-Compliance)** — ABAC access control, tamper-evident audit trail, GDPR/HIPAA/EU AI Act retention policies. *New in v2.8*
+Add to your IDE's MCP config:
 
----
-
-## 🌐 Works Everywhere
-
-**SuperLocalMemory is the ONLY memory system that works across ALL your tools:**
-
-### Supported IDEs & Tools
-
-| Tool | Integration | How It Works |
-|------|-------------|--------------|
-| **Claude Code** | ✅ Skills + MCP | `/superlocalmemoryv2-remember` |
-| **Cursor** | ✅ MCP + Skills | AI uses memory tools natively |
-| **Windsurf** | ✅ MCP + Skills | Native memory access |
-| **Claude Desktop** | ✅ MCP | Built-in support |
-| **OpenAI Codex** | ✅ MCP + Skills | Auto-configured (TOML) |
-| **VS Code / Copilot** | ✅ MCP + Skills | `.vscode/mcp.json` |
-| **Continue.dev** | ✅ MCP + Skills | `/slm-remember` |
-| **Cody** | ✅ Custom Commands | `/slm-remember` |
-| **Gemini CLI** | ✅ MCP + Skills | Native MCP + skills |
-| **JetBrains IDEs** | ✅ MCP | Via AI Assistant settings |
-| **Zed Editor** | ✅ MCP | Native MCP tools |
-| **Aider** | ✅ Smart Wrapper | `aider-smart` with context |
-| **Any Terminal** | ✅ Universal CLI | `slm remember "content"` |
-
-### Three Ways to Access
-
-1. **MCP (Model Context Protocol)** — Auto-configured for Cursor, Windsurf, Claude Desktop
-   - AI assistants get natural access to your memory
-   - No manual commands needed
-   - "Remember that we use this framework" just works
-
-2. **Skills & Commands** — For Claude Code, Continue.dev, Cody
-   - `/superlocalmemoryv2-remember` in Claude Code
-   - `/slm-remember` in Continue.dev and Cody
-   - Familiar slash command interface
-
-3. **Universal CLI** — Works in any terminal or script
-   - `slm remember "content"` - Simple, clean syntax
-   - `slm recall "query"` - Search from anywhere
-   - `aider-smart` - Aider with auto-context injection
-
-**All three methods use the SAME local database.** No data duplication, no conflicts.
-
-[Complete setup guide for all tools →](docs/MCP-MANUAL-SETUP.md)
-
----
+```json
+{
+  "mcpServers": {
+    "superlocalmemory": {
+      "command": "slm",
+      "args": ["mcp"]
+    }
+  }
+}
+```
 
-## 🆚 vs Alternatives
-
-### The Hard Truth About "Free" Tiers
-
-| Solution | Free Tier Limits | Paid Price | What's Missing |
-|----------|-----------------|------------|----------------|
-| **Mem0** | 10K memories, limited API | Usage-based | No pattern learning, not local |
-| **Zep** | Limited credits | $50/month | Credit system, cloud-only |
-| **Supermemory** | 1M tokens, 10K queries | $19-399/mo | Not local, no graphs |
-| **Personal.AI** | ❌ No free tier | $33/month | Cloud-only, closed ecosystem |
-| **Letta/MemGPT** | Self-hosted (complex) | TBD | Requires significant setup |
-| **SuperLocalMemory** | **Unlimited** | **$0 forever** | **Nothing.** |
-
-### What Actually Matters
-
-| Feature | Mem0 | Zep | Khoj | Letta | **SuperLocalMemory** |
-|---------|------|-----|------|-------|------------------------|
-| **Works in Cursor** | Cloud Only | ❌ | ❌ | ❌ | ✅ **Local** |
-| **Works in Windsurf** | Cloud Only | ❌ | ❌ | ❌ | ✅ **Local** |
-| **Works in VS Code** | 3rd Party | ❌ | Partial | ❌ | ✅ **Native** |
-| **Universal CLI** | ❌ | ❌ | ❌ | ❌ | ✅ |
-| **Multi-Layer Architecture** | ❌ | ❌ | ❌ | ❌ | ✅ |
-| **Pattern Learning** | ❌ | ❌ | ❌ | ❌ | ✅ |
-| **Adaptive ML Ranking** | Cloud LLM | ❌ | ❌ | ❌ | ✅ **Local ML** |
-| **Knowledge Graphs** | ✅ | ✅ | ❌ | ❌ | ✅ |
-| **100% Local** | ❌ | ❌ | Partial | Partial | ✅ |
-| **GDPR by Design** | ❌ | ❌ | ❌ | ❌ | ✅ |
-| **Zero Setup** | ❌ | ❌ | ❌ | ❌ | ✅ |
-| **Completely Free** | Limited | Limited | Partial | ✅ | ✅ |
-
-**SuperLocalMemory is the ONLY solution that:**
-- ✅ **Learns and adapts** locally — no cloud LLM needed for personalization
-- ✅ Works across 17+ IDEs and CLI tools
-- ✅ Remains 100% local (no cloud dependencies)
-- ✅ GDPR Article 17 compliant — one-command data erasure
-- ✅ Completely free with unlimited memories
-
-[See full competitive analysis →](docs/COMPETITIVE-ANALYSIS.md)
+24 MCP tools available: `remember`, `recall`, `search`, `fetch`, `list_recent`, `get_status`, `build_graph`, `switch_profile`, `health`, `consistency_check`, `recall_trace`, and more.
 
 ---
 
-## ⚡ Measured Performance
-
-All numbers measured on real hardware (Apple M4 Pro, 24GB RAM). No estimates — real benchmarks.
-
-### Search Speed
-
-| Database Size | Median Latency | P95 Latency |
-|---------------|----------------|-------------|
-| 100 memories | **10.6ms** | 14.9ms |
-| 500 memories | **65.2ms** | 101.7ms |
-| 1,000 memories | **124.3ms** | 190.1ms |
-
-For typical personal use (under 500 memories), search results return faster than you blink.
-
-### Concurrent Writes — Zero Errors
-
-| Scenario | Writes/sec | Errors |
-|----------|------------|--------|
-| 1 AI tool writing | **204/sec** | 0 |
-| 2 AI tools simultaneously | **220/sec** | 0 |
-| 5 AI tools simultaneously | **130/sec** | 0 |
-
-Concurrent-safe architecture = zero "database is locked" errors, ever.
-
-### Storage
-
-10,000 memories = **13.6 MB** on disk (~1.4 KB per memory). Your entire AI memory history takes less space than a photo.
-
-### Graph Construction
-
-| Memories | Build Time |
-|----------|-----------|
-| 100 | 0.28s |
-| 1,000 | 10.6s |
-
-Auto-clustering discovers 6-7 natural topic communities from your memories.
-
-[Full benchmark details →](https://github.com/varun369/SuperLocalMemoryV2/wiki/Performance-Benchmarks)
+## V3 Engine Features
+
+### Retrieval (4-Channel Hybrid)
+- Semantic similarity with Fisher-Rao information geometry
+- BM25 keyword matching (persisted tokens, survives restart)
+- Entity graph with spreading activation (3-hop, decay=0.7)
+- Temporal date-aware retrieval with interval support
+- RRF fusion (k=60) + cross-encoder reranking
+
+### Intelligence
+- 11-step ingestion pipeline (entity resolution, fact extraction, emotional tagging, scene building, sheaf consistency)
+- Adaptive learning with LightGBM-based ranking (3-phase bootstrap)
+- Behavioral pattern detection (query habits, entity preferences, active hours)
+- Outcome tracking for retrieval feedback loops
+
+### Trust & Compliance
+- Bayesian Beta distribution trust scoring (per-agent, per-fact)
+- Trust gates (block low-trust agents from writing/deleting)
+- ABAC (Attribute-Based Access Control) with DB-persisted policies
+- GDPR Article 15/17 compliance (full export + complete erasure)
+- EU AI Act data sovereignty (Mode A: zero cloud, data stays local)
+- Tamper-proof hash-chain audit trail (SHA-256 linked entries)
+- Data provenance tracking (who created what, when, from where)
+
+### Infrastructure
+- 17-tab web dashboard (trust visualization, math health, recall lab)
+- 17+ IDE integrations with pre-built configs
+- Profile isolation (16+ independent memory spaces)
+- V2 to V3 migration tool (zero data loss, rollback support)
+- Auto-capture and auto-recall hooks for Claude Code
 
 ---
 
-## 🔧 CLI Commands
+## Benchmarks
 
-```bash
-# Memory Operations
-superlocalmemoryv2-remember "content" --tags tag1,tag2  # Save memory
-superlocalmemoryv2-recall "search query"                 # Search
-superlocalmemoryv2-list                                  # Recent memories
-superlocalmemoryv2-status                                # System health
-
-# Profile Management
-superlocalmemoryv2-profile list                          # Show all profiles
-superlocalmemoryv2-profile create <name>                 # New profile
-superlocalmemoryv2-profile switch <name>                 # Switch context
-
-# Knowledge Graph
-python ~/.claude-memory/graph_engine.py build            # Build graph
-python ~/.claude-memory/graph_engine.py stats            # View clusters
-
-# Pattern Learning
-python ~/.claude-memory/pattern_learner.py update        # Learn patterns
-python ~/.claude-memory/pattern_learner.py context 0.5   # Get identity
-
-# Visualization Dashboard
-python ~/.claude-memory/ui_server.py                     # Launch web UI
-```
+Evaluated on the [LoCoMo benchmark](https://arxiv.org/abs/2402.09714) (Long Conversation Memory):
 
-[Complete CLI reference →](docs/CLI-COMMANDS-REFERENCE.md)
+### Mode A Ablation (conv-30, 81 questions, zero-LLM)
 
----
-
-## 📖 Documentation
-
-| Guide | Description |
-|-------|-------------|
-| [Quick Start](https://github.com/varun369/SuperLocalMemoryV2/wiki/Quick-Start-Tutorial) | Get running in 5 minutes |
-| [Installation](https://github.com/varun369/SuperLocalMemoryV2/wiki/Installation) | Detailed setup instructions |
-| [Visualization Dashboard](https://github.com/varun369/SuperLocalMemoryV2/wiki/Visualization-Dashboard) | Interactive web UI guide |
-| [Interactive Graph](https://github.com/varun369/SuperLocalMemoryV2/wiki/Using-Interactive-Graph) | Graph exploration guide (NEW v2.6.5) |
-| [Framework Integrations](docs/FRAMEWORK-INTEGRATIONS.md) | LangChain & LlamaIndex setup |
-| [Knowledge Graph](https://github.com/varun369/SuperLocalMemoryV2/wiki/Knowledge-Graph-Guide) | How clustering works |
-| [Pattern Learning](docs/PATTERN-LEARNING.md) | Identity extraction |
-| [Memory Lifecycle](https://github.com/varun369/SuperLocalMemoryV2/wiki/Memory-Lifecycle) | Lifecycle states, compaction, bounded growth (v2.8) |
-| [Behavioral Learning](https://github.com/varun369/SuperLocalMemoryV2/wiki/Behavioral-Learning) | Action outcomes, pattern extraction (v2.8) |
-| [Enterprise Compliance](https://github.com/varun369/SuperLocalMemoryV2/wiki/Enterprise-Compliance) | ABAC, audit trail, retention policies (v2.8) |
-| [Upgrading to v2.8](https://github.com/varun369/SuperLocalMemoryV2/wiki/Upgrading-to-v2.8) | Migration guide from v2.7 |
-| [API Reference](docs/API-REFERENCE.md) | Python API documentation |
-
----
+| Configuration | Micro Avg | Multi-Hop | Open Domain |
+|:-------------|:---------:|:---------:|:-----------:|
+| Full (all layers) | **62.3%** | **50%** | **78%** |
+| Math layers off | 59.3% | 38% | 70% |
+| Entity channel off | 56.8% | 38% | 73% |
+| BM25 channel off | 53.2% | 23% | 71% |
+| Cross-encoder off | 31.8% | 17% | — |
 
-## 🤝 Contributing
+### Competitive Landscape
 
-We welcome contributions! See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.
+| System | Score | LLM Required | Open Source | EU AI Act |
+|:-------|:-----:|:------------:|:-----------:|:---------:|
+| EverMemOS | 92.3% | Yes | No | No |
+| MemMachine | 91.7% | Yes | No | No |
+| Hindsight | 89.6% | Yes | No | No |
+| **SLM V3 Mode C** | **~78%** | Optional | **Yes** | Partial |
+| **SLM V3 Mode A** | **62.3%** | **No** | **Yes** | **Yes** |
+| Mem0 ($24M) | 34.2% F1 | Yes | Partial | No |
 
-**Areas for contribution:**
-- Additional pattern categories
-- Performance optimizations
-- Integration with more AI assistants
-- Documentation improvements
+*SLM V3 is the only system offering a fully local mode with mathematical guarantees and EU AI Act compliance.*
 
 ---
 
-## 💖 Support This Project
+## Research Papers
 
-If SuperLocalMemory saves you time, consider supporting its development:
+### V3: Information-Geometric Foundations
+> **SuperLocalMemory V3: Information-Geometric Foundations for Zero-LLM Enterprise Agent Memory**
+> Varun Pratap Bhardwaj (2026)
+> [Zenodo DOI: 10.5281/zenodo.19038659](https://zenodo.org/records/19038659)
 
-- ⭐ **Star this repo** — helps others discover it
-- 🐛 **Report bugs** — [open an issue](https://github.com/varun369/SuperLocalMemoryV2/issues)
-- 💡 **Suggest features** — [start a discussion](https://github.com/varun369/SuperLocalMemoryV2/discussions)
-- ☕ **Buy me a coffee** — [buymeacoffee.com/varunpratah](https://buymeacoffee.com/varunpratah)
-- 💸 **PayPal** — [paypal.me/varunpratapbhardwaj](https://paypal.me/varunpratapbhardwaj)
-- 💖 **Sponsor** — [GitHub Sponsors](https://github.com/sponsors/varun369)
+### V2: Architecture & Engineering
+> **SuperLocalMemory: A Structured Local Memory Architecture for Persistent AI Agent Context**
+> Varun Pratap Bhardwaj (2026)
+> [arXiv:2603.02240](https://arxiv.org/abs/2603.02240) | [Zenodo DOI: 10.5281/zenodo.18709670](https://zenodo.org/records/18709670)
 
 ---
 
-## 📜 License
+## Project Structure
 
-MIT License — use freely, even commercially. Just include the license.
+```
+superlocalmemory/
+├── src/superlocalmemory/     # Python package (17 sub-packages)
+│   ├── core/                 # Engine, config, modes, profiles
+│   ├── retrieval/            # 4-channel retrieval + fusion + reranking
+│   ├── math/                 # Fisher-Rao, Sheaf, Langevin
+│   ├── encoding/             # 11-step ingestion pipeline
+│   ├── storage/              # SQLite with WAL, FTS5, migrations
+│   ├── trust/                # Bayesian scoring, gates, provenance
+│   ├── compliance/           # GDPR, EU AI Act, ABAC, audit chain
+│   ├── learning/             # Adaptive ranking, behavioral patterns
+│   ├── mcp/                  # MCP server (24 tools, 6 resources)
+│   ├── cli/                  # CLI with setup wizard
+│   └── server/               # Dashboard API + UI server
+├── tests/                    # 1400+ tests
+├── ui/                       # 17-tab web dashboard
+├── ide/                      # IDE configs for 17+ tools
+├── docs/                     # Documentation
+└── pyproject.toml            # Modern Python packaging
+```
 
 ---
 
-## 👨‍💻 Author
-
-**Varun Pratap Bhardwaj** — Founder, Qualixar · Solution Architect
+## Contributing
 
-[![GitHub](https://img.shields.io/badge/GitHub-@varun369-181717?style=flat-square&logo=github)](https://github.com/varun369)
+See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.
 
-Building the complete agent development platform at Qualixar — memory, testing, contracts, and security for AI agents.
+## License
 
-### Part of the Qualixar Agent Development Platform
+MIT License. See [LICENSE](LICENSE).
 
-SuperLocalMemory is part of **Qualixar**, a suite of open-source tools for building reliable AI agents:
+## Attribution
 
-| Product | What It Does |
-|---------|-------------|
-| **SuperLocalMemory** | Local-first AI agent memory |
-| **SkillFortify** | Agent skill supply chain security |
+Part of [Qualixar](https://qualixar.com) | Author: [Varun Pratap Bhardwaj](https://varunpratap.com)
 
 ---
 
 <p align="center">
-  <b>100% local. 100% private. 100% yours.</b>
-</p>
-
-<p align="center">
-  <a href="https://github.com/varun369/SuperLocalMemoryV2">
-    <img src="https://img.shields.io/badge/⭐_Star_on_GitHub-black?style=for-the-badge&logo=github" alt="Star on GitHub"/>
-  </a>
+  <sub>Built with mathematical rigor. Not in the race — here to help everyone build better AI memory systems.</sub>
 </p>