Skip to content

Neural Memory v0.2.0 — Autonomous Desktop Knowledge Graph

Latest

Choose a tag to compare

@Hexecu Hexecu released this 03 Sep 14:17
· 1 commit to main since this release
724f814

Neural Memory v0.2.0 — Autonomous Desktop Knowledge Graph & Second Brain

We are thrilled to announce the release of Neural Memory v0.2.0! This milestone marks the transformation of Neural Memory from a developer context tracker into an autonomous personal cognitive memory assistant and second brain, featuring a Dual-Track Architecture and a zero-configuration standalone DMG installer for macOS.


🚀 Key Highlights & What's New

📦 Standalone Drag-and-Drop macOS Installer (Zero Docker Required)

  • Download NeuralMemoryAgent-0.2.0-Installer.dmg below, mount the image, and drag NeuralMemoryAgent.app to your Applications folder.
  • Zero prerequisites: runs out-of-the-box on macOS 13+ with no Docker, no Python, no Homebrew, and no terminal commands.
  • Features an embedded SQLite Property Graph Engine running in WAL mode at ~/Library/Application Support/NeuralMemory/memory.db.
  • Managed by a native Swift child supervisor (EmbeddedDaemonManager) that starts and stops the embedded Mach-O daemon automatically.

🌌 Native SwiftUI Temporal Knowledge Graph Navigator

  • Topic Clusters Mode: Orbital regional gravity organizes thoughts, meetings, and commitments into thematic galaxies without central collapse.
  • Timeline Stream Mode: Horizontal chronological axis with 5 vertical semantic altitude lanes (Reflections, Decisions, Meetings, Commitments, Topics).
  • Time-Travel Scrubber: Interactive slider to replay memory history with presets (All Time, 30 Days, 7 Days, 3 Days, Today).
  • Ebbinghaus Memory Decay: Mathematical retention curve rendering active thoughts with vivid neon glow while dormant memories gently fade to background awareness.

🧠 Cognitive Life & Work Second Brain Ontology

  • Multimodal Interaction Bundles (POST /api/ingest/bundle): Captures screen context, visual frames, and keystrokes on natural human action boundaries (Enter key, app switch, debounce).
  • Contextual Micro-Feedback Anchoring: Eliminates arbitrary length cutoffs. Assents like "Ok, procedi con l'offerta da 5000€" are anchored to the sender, document, and decision terms.
  • Ontology Nodes: Traces Decision (verdicts & rationale), Commitment (deadlines & debtor/creditor), Meeting (attendees & topics), and Reflection (consolidated insights).

🔒 PrivacyShield Local Redaction Pipeline

  • In-memory regex and Luhn checksum filtering automatically masks credit cards, IBANs, API keys (sk-..., AIza..., GitHub tokens), and passwords before storage or optional LLM enrichment.
  • Strict deny-list ignoring password vaults (1Password, Bitwarden, Keychain) and banking sessions.
  • Ephemeral raw event pruning (48h retention) via Dream Mode (POST /api/memory/consolidate).

🤖 Model Context Protocol (MCP) for AI Coding Agents

  • Native stdio server providing tools for Claude Desktop, Cursor, Antigravity, and OpenCode: recall_decisions, recall_commitments, recall_meetings, get_daily_briefing, and search.

💻 Downloads & Assets

  • macOS Drag-and-Drop Installer: NeuralMemoryAgent-0.2.0-Installer.dmg (112 MB, Apple Silicon / arm64, macOS 13+)

🛠️ Developer Stack (Track A: Open Local)

For developers contributing or using the full Neo4j 5.26 stack:

git clone https://github.com/Hexecu/mcp-neuralmemory.git
cd mcp-neuralmemory
./scripts/bootstrap.sh

Full documentation and guides are available in the docs/ folder of the repository.