Skip to content

Releases: qualixar/superlocalmemory

v2.8.0 — Memory Lifecycle, Behavioral Learning, Enterprise Compliance

26 Feb 19:09
@varun369 varun369
Immutable release. Only release title and notes can be modified.
v2.8.0
5ade37c

Choose a tag to compare

View all tags
v2.8.0 — Memory Lifecycle, Behavioral Learning, Enterprise Compliance Latest
Latest

What's New

Memory Lifecycle Management — Memories automatically organize themselves over time based on usage patterns, keeping your system fast and relevant.

Behavioral Learning — The system learns from action outcomes and transfers knowledge across projects. Zero LLM inference — pure local pattern recognition.

Enterprise Compliance — Full access control, immutable audit trails, and retention policy management for GDPR, HIPAA, and EU AI Act.

6 New MCP Tools

Tool Purpose
report_outcome Record action outcomes for behavioral learning
get_lifecycle_status View memory lifecycle state distribution
set_retention_policy Configure retention policies
compact_memories Trigger lifecycle transitions
get_behavioral_patterns View learned behavioral patterns
audit_trail Query the compliance audit trail

Upgrade

npm install -g superlocalmemory@latest

Drop-in upgrade. All existing memories, settings, and configurations preserved. Upgrade Guide →

Full Changelog: CHANGELOG.md
Website: superlocalmemory.com
Research: superlocalmemory.com/research

Assets 3
Loading

SuperLocalMemory v2.7.4

16 Feb 14:01
@varun369 varun369
Immutable release. Only release title and notes can be modified.
v2.7.4
37cdbd2

Choose a tag to compare

View all tags
SuperLocalMemory v2.7.4

What's New

Added

  • Per-profile learning — each profile has its own preferences and feedback
  • Thumbs up/down and pin feedback on memory cards
  • Learning data management in Settings (backup + reset)
  • "What We Learned" summary card in Learning tab

Improved

  • Smarter learning from your natural usage patterns
  • Recall results improve automatically over time
  • Privacy notice for all learning features
  • Learning and backup databases protected together
  • All dashboard tabs refresh on profile switch

Full changelog: https://github.com/varun369/SuperLocalMemoryV2/blob/main/CHANGELOG.md

Loading

v2.7.0 — Your AI Learns You

15 Feb 22:35
@varun369 varun369
Immutable release. Only release title and notes can be modified.
v2.7.0
b400eb4

Choose a tag to compare

View all tags
v2.7.0 — Your AI Learns You

SuperLocalMemory v2.7.0 — "Your AI Learns You"

Your memory system now learns your patterns, adapts to your workflow, and personalizes recall — all 100% locally on your machine.

Highlights

  • Adaptive Learning — Three-layer learning: tech preferences, project context, workflow patterns
  • Personalized Ranking — LightGBM LambdaRank re-ranking with 9 personalized features
  • Privacy by Design — Separate learning.db for behavioral data, GDPR Article 17 compliant, zero telemetry
  • Day-1 Intelligence — Synthetic bootstrap trains ML model from existing memories immediately
  • Multi-Channel Feedback — memory_used MCP tool, slm useful CLI, dashboard clicks
  • 12 MCP Tools — 3 new: memory_used, get_learned_patterns, correct_pattern
  • 319 Tests — Unit, E2E, regression, fresh-install, and edge-case coverage

Install / Upgrade

# npm (recommended)
npm install -g superlocalmemory

# Or clone
git clone https://github.com/varun369/SuperLocalMemoryV2.git
cd SuperLocalMemoryV2 && bash install.sh

Research Foundations

Built on peer-reviewed research: LambdaRank (Burges 2010), TSW-PrefixSpan (IEEE 2020), MACLA (arXiv:2512.18950), cold-start mitigation (LREC 2024), privacy-preserving feedback design.

Documentation

Created by Varun Pratap Bhardwaj — Solution Architect

Loading

v2.6.0 — Security Hardening & Performance

15 Feb 13:06
@varun369 varun369
Immutable release. Only release title and notes can be modified.
v2.6.0
170349f

Choose a tag to compare

View all tags
v2.6.0 — Security Hardening & Performance

What's New

38 improvements across 6 phases, consolidating findings from 7 independent security audits. All changes are backward compatible — existing installations upgrade seamlessly.

Security

  • Profile isolation on all query endpoints — memories from one profile can never leak to another
  • Trust enforcement gate — agents with trust score below 0.3 are blocked from write operations
  • Rate limiting on API and dashboard (100 writes/min, 300 reads/min per IP)
  • SSRF protection for webhook URLs (private/internal IP blocking)
  • SQL injection defense — table name whitelist validation in reset operations
  • Error message sanitization — internal paths and table names stripped from all error responses
  • Optional API key auth — opt-in authentication for write endpoints via ~/.claude-memory/api_key
  • Event content redaction — event payloads no longer leak memory content by default

Performance & Scalability

  • Graph cap raised from 5,000 → 10,000 memories with intelligent sampling beyond the limit
  • HNSW-accelerated graph building — O(n log n) edge construction instead of O(n²)
  • HNSW-accelerated search — O(log n) memory lookups wired into the search pipeline
  • Hybrid search engine — BM25 + Graph + TF-IDF fusion available via SLM_HYBRID_SEARCH=true
  • TF-IDF incremental optimization — skip rebuilds when memory count changes less than 5%
  • Read connection pool limit (50 connections) with graceful reuse

Reliability

  • Database integrity check (PRAGMA quick_check) on every startup
  • Incremental auto-vacuum for automatic space reclamation
  • CORS restricted to known service ports (8765, 8417, 8766)
  • Bounded write queue (1,000 entries) for backpressure protection
  • Auto-prune for event table (every 100 events or 24 hours)

Infrastructure

  • GitHub Actions CI — automated testing on Python 3.8, 3.10, 3.12
  • Ghost agent cleanup — dashboard filters out disconnected agents
  • SECURITY.md updated with encryption posture and defense-in-depth documentation

Documentation

  • Graph scaling limits documented in README, wiki, and website
  • All public docs harmonized for v2.6 features
  • Wiki synced with 9 updated pages

Upgrade

cd ~/path/to/SuperLocalMemoryV2
git pull origin main
./install.sh

Your existing ~/.claude-memory/memory.db is fully compatible — no migration needed.

Full Changelog

v2.5.0...v2.6.0

Loading

v2.5.0 - Your AI Memory Has a Heartbeat

12 Feb 06:07
@varun369 varun369
Immutable release. Only release title and notes can be modified.
v2.5.0
2005934

Choose a tag to compare

View all tags
v2.5.0 - Your AI Memory Has a Heartbeat

🎉 SuperLocalMemory v2.5.0 - "Your AI Memory Has a Heartbeat"

Major Feature Release | February 12, 2026 | Backward Compatible

SuperLocalMemory transforms from passive storage to active coordination layer. Every memory operation now triggers real-time events. Agents are tracked. Trust is scored. The dashboard is live.

🆕 What's New

Real-Time Event System

  • Event Bus: Every memory write/update/delete fires real-time events via SSE, WebSocket, and Webhook
  • Live Events Tab: Watch memory operations happen in real-time across all AI tools
  • Event Subscriptions: Durable + ephemeral subscriptions with filters

Multi-Agent Coordination

  • Agent Registry: Track which AI agents connect - protocol, trust score, write/recall counters
  • Agents Tab: Dashboard view showing all connected agents with live statistics
  • Provenance Tracking: Know who created each memory and via which protocol

Concurrency Fixed

  • DbConnectionManager: SQLite WAL mode + write queue eliminates "database is locked" errors
  • Thread-Safe: Multiple agents can write simultaneously without conflicts
  • 5-second busy timeout: Patient retry instead of immediate failure

Trust Scoring (Silent Collection)

  • Bayesian Trust Signals: Track agent behavior (cross-agent recalls, burst writes, quick deletes)
  • Research-Grounded: Beta-Binomial posterior calculation
  • Silent in v2.5: Scores collected but not enforced (enforcement in v2.6)

Complete Visual Documentation (85 Assets)

  • 20 dashboard screenshots (real UI, light/dark modes)
  • 7 CLI terminal mockups (professional styling)
  • 5 animated GIFs (demos of key features)
  • 3 video walkthroughs (installation, quick start, dashboard tour)
  • 24 thumbnails (web-optimized previews)
  • Updated wiki (74 image references, all placeholders removed)
  • Enhanced website (carousel, feature galleries, visual guides)

📦 Installation

Quick Install (All Platforms)

```bash
npm install -g superlocalmemory
```

From Source

```bash
git clone https://github.com/varun369/SuperLocalMemoryV2.git
cd SuperLocalMemoryV2
./install.sh # Mac/Linux

or: .\install.ps1 # Windows

```

🔧 What Changed

Core Modules

  • `memory_store_v2.py`: DbConnectionManager integration, event bus hooks, provenance tracking
  • `mcp_server.py`: Singleton instance, agent registration, no more per-call overhead
  • `ui_server.py`: Refactored to 9 modular route files (from 2008-line monolith)
  • `ui/app.js`: Split into 13 modular files (from 1588-line monolith)

New Components

  • `src/db_connection_manager.py` - Concurrency fix (WAL mode + write queue)
  • `src/event_bus.py` - Real-time event broadcasting
  • `src/subscription_manager.py` - Event subscriptions
  • `src/webhook_dispatcher.py` - HTTP webhook delivery
  • `src/agent_registry.py` - Agent tracking
  • `src/provenance_tracker.py` - Memory origin tracking
  • `src/trust_scorer.py` - Bayesian trust scoring

Database Schema

  • 4 new tables: `memory_events`, `subscriptions`, `agent_registry`, `trust_signals`
  • 4 new columns on memories: `created_by`, `source_protocol`, `trust_score`, `provenance_chain`

🐛 Fixes

  • "Database is locked" errors: Eliminated via WAL mode + write serialization
  • Fresh database crashes: Graceful empty responses when tables don't exist
  • Per-call overhead: MCP handlers no longer rebuild vectorizer on every call

🧪 Testing

  • 63 pytest tests: Unit, concurrency, regression, security, edge cases
  • 27 end-to-end tests: All v2.5 components + 50 concurrent writes
  • 15 fresh-database tests: New user scenarios
  • 17 backward compatibility tests: Against live v2.4.2 database
  • Playwright dashboard testing: All 8 tabs, SSE stream, profile switching

🌐 Website & Documentation

🔄 Upgrade Instructions

Automatic (Recommended)

```bash
npm update -g superlocalmemory
```

Manual

```bash
cd SuperLocalMemoryV2
git pull origin main
./install.sh # Mac/Linux

or: .\install.ps1 # Windows

```

Database Migration: Automatic. New tables and columns are created on first run.

🚀 What's Next

  • v2.5.1 (April 2026): BMW Dashboard - Identity home, pattern viz, NL search
  • v2.6 (May 2026): A2A Protocol - Agent-to-agent communication
  • v2.7 (Jul-Aug 2026): Identity export/import + EU AI Act compliance
  • v3.0 (Oct 2026): Enterprise multi-tenant platform

🙏 Credits

Created by: Varun Pratap Bhardwaj (Solution Architect)
License: MIT
Repository: https://github.com/varun369/SuperLocalMemoryV2

Research Foundations:

  • Event Bus: Real-time memory systems
  • Multi-Agent: A2A Protocol (Google/Linux Foundation)
  • Trust Scoring: MACLA (arXiv:2512.18950)
  • Knowledge Graph: GraphRAG (Microsoft, arXiv:2404.16130)
  • Pattern Learning: MemoryBank (AAAI 2024)

Full Changelog: https://github.com/varun369/SuperLocalMemoryV2/blob/main/CHANGELOG.md

Loading

v2.3.7 - Critical Fix: Smart Content Truncation

09 Feb 17:38
@varun369 varun369
Immutable release. Only release title and notes can be modified.
v2.3.7
da6972b

Choose a tag to compare

View all tags
v2.3.7 - Critical Fix: Smart Content Truncation

v2.3.7 - Critical Fix: Smart Content Truncation

🐛 Critical Bug Fix

Problem: Recall was showing truncated content (200 chars max) even for short memories. Affected ALL IDEs using SuperLocalMemory.

Root Cause: Hardcoded [:200] truncation in memory_store_v2.py line 1014.

✨ What's Fixed

Smart Truncation

  • Memories < 5000 chars: Show FULL content (no truncation) ✅
  • Memories ≥ 5000 chars: Smart 2000 char preview with "..."
  • 95%+ of memories now display in full by default

New Feature: --full Flag

Override truncation to always show complete content:

slm recall "query" --full
slm list --full
python3 ~/.claude-memory/memory_store_v2.py search "query" --full

Bonus Fix

  • Fixed get command error (get_memory() → get_by_id())

🚀 How to Update

npm Users (Recommended)

npm update -g superlocalmemory
# or
npm install -g superlocalmemory@latest

Direct Install Users

cd ~/SuperLocalMemoryV2
git pull origin main
bash install.sh

Verify Update

slm status
slm recall "test"  # Should now show full content

📦 Package Stats

  • Version: 2.3.7
  • Size: 403.1 KB compressed
  • Files: 105
  • Commit: da6972b

🔄 What Gets Updated

  • ✅ Python backend (memory_store_v2.py)
  • ✅ Node.js hooks (recall, list)
  • ✅ All CLI commands
  • ✅ MCP server compatibility

Your data is safe: Database and memories preserved.

🧪 Tested With

  • ✅ 74+ existing memories
  • ✅ Backward compatible (same output format)
  • ✅ All IDEs: Claude, Cursor, Windsurf, VS Code, etc.
  • ✅ All access methods: MCP, Skills, CLI

📚 Documentation

🙏 Credits

Issue reported by users across multiple IDEs. Fixed and tested with 74+ real-world memories.

npm: https://www.npmjs.com/package/superlocalmemory
GitHub: https://github.com/varun369/SuperLocalMemoryV2

Loading

v2.3.0 — Universal Integration

08 Feb 18:09
@varun369 varun369
Immutable release. Only release title and notes can be modified.
v2.3.0
4ba217c

Choose a tag to compare

View all tags
v2.3.0 — Universal Integration

SuperLocalMemory V2.3.0 — Universal Integration Release

The architecture was always universal. Now the distribution is too.

What's New

  • 4 new MCP configs: Codex CLI (TOML), VS Code/Copilot, Gemini CLI, JetBrains
  • slm serve command: HTTP transport for ChatGPT and remote MCP clients
  • MCP tool annotations: readOnlyHint, destructiveHint, openWorldHint on all 6 tools
  • Auto-detect 16+ tools: Codex, VS Code/Copilot, Gemini CLI, JetBrains added to install.sh
  • Universal symlink: ~/.superlocalmemory → ~/.claude-memory for non-Claude branding
  • Skills expansion: Cursor + VS Code/Copilot added to install-skills.sh with --auto flag
  • npm is primary install: npm install -g superlocalmemory

Supported Tools (16+)

Tool Integration
Claude Code Skills + MCP
Cursor MCP + Skills
Windsurf MCP + Skills
Claude Desktop MCP
OpenAI Codex MCP + Skills
VS Code / Copilot MCP + Skills
Continue.dev MCP + Skills
Cody Custom Commands
Gemini CLI MCP + Skills
JetBrains IDEs MCP
Zed Editor MCP
OpenCode MCP
Perplexity MCP
Antigravity MCP + Skills
ChatGPT HTTP Transport
Aider Smart Wrapper
Any Terminal CLI

Install

npm install -g superlocalmemory

Or manual:

git clone https://github.com/varun369/SuperLocalMemoryV2.git
cd SuperLocalMemoryV2 && ./install.sh

Backward Compatibility

100% backward compatible. Zero breaking changes. Existing data untouched.

Testing

  • 68/68 regression tests passed
  • Docker fresh install: passed (Python 3.11-slim)
  • Docker backward compatibility: passed (existing data preserved)

100% local. 100% private. 100% universal.

Created by Varun Pratap Bhardwaj

Loading

SuperLocalMemory V2.1.0 - Universal Integration + Critical Fixes

07 Feb 18:16
@varun369 varun369
Immutable release. Only release title and notes can be modified.
v2.1.0
7a9cf21

Choose a tag to compare

View all tags
SuperLocalMemory V2.1.0 - Universal Integration + Critical Fixes

SuperLocalMemory V2.1.0 - Universal Integration

Release Date: February 8, 2026
Status: Production Ready ✅
Testing: 100% pass rate on fresh Docker python:3.12 container

🚀 What's New

Universal Integration

  • MCP Protocol: Auto-configures 11+ IDEs (Claude Desktop, Cursor, Windsurf, Continue.dev, Zed, OpenCode, Antigravity, Perplexity, ChatGPT Desktop, Cody, Aider)
  • Universal Skills: 6 skills work across 5 AI tools (Claude Code, Continue.dev, Cody, Cursor, Windsurf)
  • Single Database: All tools share one memory database (~/.claude-memory/memory.db)
  • CLI: slm remember, slm recall, slm status available globally

Critical Bug Fixes (Feb 8, 2026)

Found via Docker testing on fresh systems (no pre-installed packages):

  1. scikit-learn missing (06236c1)

    • Added to requirements-core.txt
    • Graph engine requires sklearn for TF-IDF vectorization
    • Fresh installs failed without this dependency

  2. graph_engine.py column name mismatch (82a1ca4)

    • Fixed: name → cluster_name
    • Fixed: member_count → memory_count
    • get_stats() was querying non-existent columns

  3. pattern_learner.py column name mismatch (7a9cf21)

    • Fixed: key → pattern_key
    • Fixed: value → pattern_value
    • Fixed: evidence_count → frequency
    • get_patterns() was querying non-existent columns

📦 Installation

Quick Install (Mac/Linux)

# Download installer
curl -L -o SuperLocalMemory-v2.1.0-macos.dmg https://github.com/varun369/SuperLocalMemoryV2/releases/download/v2.1.0/SuperLocalMemory-v2.1.0-macos.dmg

# Mount and run installer
open SuperLocalMemory-v2.1.0-macos.dmg

From Source

git clone https://github.com/varun369/SuperLocalMemoryV2.git
cd SuperLocalMemoryV2
./install.sh

✅ Testing

Tested on:

  • ✅ macOS (M1/M2)
  • ✅ Docker python:3.12 (fresh container, no pre-installed packages)
  • ✅ Ubuntu 22.04 LTS (simulated)
  • ⏳ Windows 11 (manual testing recommended)

📖 Documentation

🙏 Attribution

Created by Varun Pratap Bhardwaj
Solution Architect & Original Creator

Licensed under MIT License (attribution required)

Co-Authored-By: Claude Sonnet 4.5 noreply@anthropic.com

Website: https://varun369.github.io/SuperLocalMemoryV2
Free Domain: https://superlocalmemory.is-a.dev (pending approval)

Loading

SuperLocalMemory V2 - Initial Release

05 Feb 12:11
@varun369 varun369
v1.0.0
f6aee8d

Choose a tag to compare

View all tags
SuperLocalMemory V2 - Initial Release

SuperLocalMemory V2 - Initial Release

A standalone intelligent memory system with knowledge graphs, pattern learning, and 4-layer architecture.

Features

  • 4-Layer Architecture: Raw Storage → Hierarchical Index → Knowledge Graph → Pattern Learning
  • 6 CLI Commands: remember, recall, list, status, reset, profile
  • Knowledge Graph: Auto-discovers relationships using TF-IDF + Leiden clustering
  • Pattern Learning: Learns your coding preferences with confidence scoring
  • Progressive Compression: 60-96% space savings
  • Multi-Profile Support: Isolated contexts for different projects
  • Claude CLI Integration: Optional skills for /superlocalmemoryv2:* commands
  • FastAPI UI Server: Visual exploration of your memory graph

Installation

git clone https://github.com/varun369/SuperLocalMemoryV2.git
cd SuperLocalMemoryV2
./install.sh

Requirements

  • Python 3.8+
  • SQLite3 (usually pre-installed)

Key Highlights

  • 100% local - zero external APIs
  • Complete privacy - your data stays on your machine
  • Works standalone or with Claude CLI
  • Built on research: GraphRAG, PageIndex, xMemory, A-RAG

Full documentation: README.md

Loading