Reference implementation of the semantic layer from "Knowledge Engineering: The Future of AI-Assisted Software Engineering" (https://kn owledgeengineering.substack.com/p/knowledge-engineering-the-future). Turns a directory of Markdown files into a queryable knowledge graph with four primitives — get, search, traverse, and list. Domain experts encode business knowledge as interconnected artifacts; AI agents access it through structured retrieval instead of unstructured context dumps. Includes three skills: /pramana:setup for daemon lifecycle, /pramana:query for token-efficient semantic retrieval, and /pramana:author for creating artifacts with auto-elicited author profiles. Supports multi-tenant serving of multiple knowledge bases from a single daemon.