Technical Awareness: Hybrid GraphRAG (Native Edge Graph & App Mapping)

Problem

While our current semantic vector database (ChromaDB) successfully handles unstructured deep-dive semantic searches across memories and the knowledge base, it lacks relational topology. The AI lacks structural intuition (e.g., knowing the exact inheritance chain from Neo.grid.View → Neo.component.Base → vnode hooks).

Proposed Solution

Pivot to a true Hybrid GraphRAG Model.

• Retain ChromaDB for its high-performance semantic vector search.
• Build the Native Neo Graph DB specifically to store relational topologies (edges) that sit alongside Chroma.

Requirements

1. Native Graph Database Engine:
   • Build a zero-dependency JS Graph Database that sits alongside ChromaDB.
   • Design lightweight node/edge relationship schemas in a purely JS environment to explicitly connect vector chunks (Graph Layer interacting with Semantic Layer).
2. Application Engine Knowledge Graph Mapping:
   • (Deprecation Update): We will not build a standalone static AST class hierarchy ingestor. Instead, the AI will rely on the neo-mjs-neural-link MCP server for powerful runtime scene inspections and the neo-mjs-knowledge-base MCP for static relationship querying.
   • The Native Edge graph should instead focus exclusively on topological relationships mapping external concepts (File System structure, Agent Memories, GitHub Epics, Boardroom KPIs).
   • Expose MCP server endpoints so the AI can traverse this graph dynamically to gain architectural awareness.

Definition of Done

• A zero-dependency JS Graph DB module is functional and integrated into the overarching server architecture.
• A functional FileSystemIngestor natively injects code architecture anchors into the graph.
• Agents can query the graph and receive deterministic answers.