Skip to content

v1.5.0 — Turso Cloud Sync, Git Project Identity, Prompt Manager & 15 MCP Tools

Choose a tag to compare

View all tags
@Lotargo Lotargo released this 07 Aug 01:54
· 212 commits to main since this release
v1.5.0
d883180

@lotargo/memory_plugin v1.5.0

Major stable release introducing Turso Cloud Database Sync, Git-first Project Identity, System Prompt Management, Agent-Driven Knowledge Graph linking, and a expanded suite of 15 MCP tools.

🚀 Key Features & Architectural Enhancements

  • Turso Cloud Sync & 3 Storage Modes:
    • only-local: 100% local SQLite database.
    • only-cloud: Direct remote LibSQL database connection.
    • hybrid-sync: Local-first performance with automatic background cloud synchronization, conflict resolution, and reverse sync.
  • Cloud Failover & Circuit Breaker: Automatic fallback to local SQLite cache if cloud endpoints encounter network degradation or downtime.
  • Headless Authentication: Token & env-based login (memory_plugin login --api-key <TOKEN> / memory_plugin setup --api-key <TOKEN> --mode hybrid-sync / --from-env) for Docker, Google Jules, and VPS environments.
  • Git-Based Project Identity: Project memories are now keyed to normalized Git repository identity (git:github.com/owner/repo or git:local:<repo>), ensuring memories persist across subdirectories, OSes, and machines.
  • Global Prompt Synchronization: Automatic memory agent instruction injection via memory_plugin enable-prompt / disable-prompt across client configurations (~/.gemini/config/AGENTS.md, ~/.codex/AGENTS.md, ~/.claude/CLAUDE.md).
  • Agent-Driven Knowledge Graph: Associate durable Notebook facts directly with RAG documents, sections, and exact line ranges (:L20-35) using link_knowledge or remember(docId, startLine, endLine).

🛠️ Complete MCP Tool Suite (15 Tools)

  • Notebook Memory (Layer 1): remember, recall, get_fact, update_fact, forget, memory_info.
  • Project Identity Management: link_project_memory, unlink_project_memory, relink_project_memory.
  • RAG Knowledge Base & Graph (Layers 2 & 3): ingest_document, query_knowledge_base, manage_knowledge_base, link_knowledge.
  • MCP & Agent Helpers: list-mcp-tools, mcp-reminder.

📈 Empirical Search Quality Results

Evaluated over a 32-document technical corpus (21 queries):

Search Strategy MRR@5 Recall@5 NDCG@5
BM25 Lexical Search Only 0.6706 76.19% 0.6934
Dense ONNX Vector Only 0.8135 100.00% 0.8612
Hybrid RRF ($k=10$) 0.8810 95.24% 0.8997
Hybrid RSF ($\alpha=0.5$) 0.9286 100.00% 0.9473

🧹 Package & CLI Optimizations

  • Reduced package tarball footprint (115 kB archive / 431 kB unpacked).
  • Production CLI (memory_plugin cli) isolated from local benchmark scripts.
  • Added npm run benchmark and npm test scripts for local repository development.
Assets 2
Loading