PROJECT CONTEXT

LORE

NODE.JSLANCEDBTRANSFORMERS.JSMCP

A cross-project technical knowledge management system for AI agents. It uses semantic search to archive and retrieve architectural lessons learned across different workspaces.

CORE PHILOSOPHY

Designed to provide AI agents with a "global memory," Lore enables the sharing of insights between unrelated projects with full provenance tracking.

SYSTEM ARCHITECTURE

AI Agent (Gemini/Claude)
  ↓ queries
Lore MCP Server (Node.js)
  ↓ indexes/searches
LanceDB (Vector Store)

KEY CAPABILITIES

• Semantic Retrieval: Finds lessons by meaning and intent using local embeddings.
• Automated Harvesting: Scans workspaces to extract lessons from markdown files.
• Proactive Wisdom: Agents check the archive automatically before proposing solutions.

# Feature Tracker

PROJECT TIMELINE

2026-05-24

Automated Knowledge Harvesting: Logic to scan workspaces and extract lessons from LESSONS_LEARNT.md and memory files.

2026-05-22

Semantic Search: Local vector search powered by LanceDB and Transformers.js.

2026-05-22

Proactive Intelligence Protocol: System-wide instructions for agents to search the archive before proposing solutions.

2026-05-22

Hierarchical Knowledge Tiers: Support for both "Raw" (atomic observations) and "Synthesized" (best practices) data.

2026-05-22

Provenance & Ancestry Tracking: Recursive tracing of synthesized rules back to their original source lessons.

2026-05-22

Knowledge Health Diagnostics: Automated fragmentation monitoring with recommendations for synthesis.

2026-05-22

Admin Isolation: Secure separation of destructive maintenance tools (Update/Delete) via Environment Variables.

2026-05-22

Hybrid Storage: Binary vector store for speed with human-readable JSONL backups for portability.

UPCOMING

Visual Knowledge Graph: UI/Web interface to visualize semantic connections between lessons.

UPCOMING

Cloud Embedding Support: Optional integration with OpenAI/Vertex AI for faster/higher-dimension embeddings.

UPCOMING

Advanced De-duplication: Semantic-aware conflict resolution during bulk harvests.

UPCOMING

LORE

CORE PHILOSOPHY

SYSTEM ARCHITECTURE

KEY CAPABILITIES

Automated Knowledge Harvesting: Logic to scan workspaces and extract lessons from LESSONS_LEARNT.md and memory files.

Semantic Search: Local vector search powered by LanceDB and Transformers.js.

Proactive Intelligence Protocol: System-wide instructions for agents to search the archive before proposing solutions.

Hierarchical Knowledge Tiers: Support for both "Raw" (atomic observations) and "Synthesized" (best practices) data.

Provenance & Ancestry Tracking: Recursive tracing of synthesized rules back to their original source lessons.

Knowledge Health Diagnostics: Automated fragmentation monitoring with recommendations for synthesis.

Admin Isolation: Secure separation of destructive maintenance tools (Update/Delete) via Environment Variables.

Hybrid Storage: Binary vector store for speed with human-readable JSONL backups for portability.

Visual Knowledge Graph: UI/Web interface to visualize semantic connections between lessons.

Cloud Embedding Support: Optional integration with OpenAI/Vertex AI for faster/higher-dimension embeddings.

Advanced De-duplication: Semantic-aware conflict resolution during bulk harvests.

Multi-User Permission Layers: Granular access control beyond the simple Admin toggle.