What problem does it solve? Searching a large personal knowledge base by hand is slow and error-prone. This Skill answers natural-language questions against a compiled Obsidian wiki using a tiered retrieval strategy, returning synthesized answers with citations instead of raw file dumps. ## Core Features & Use Cases - Tiered Retrieval: Escalates from cheap index lookups to section greps to full page reads, minimizing context usage while maximizing answer quality. - Multi-hop Graph Traversal: Traces typed relationship edges (e.g., "contradicts", "derived_from") across pages to answer path queries like "how is X connected to Y". - Semantic and Hybrid Search: Supports embedding-based cosine similarity and BM25 hybrid scoring via a SQLite index, plus optional QMD integration for lexical and vector search. - Use Case: Ask "what do I know about transformer attention mechanisms" and receive a synthesized answer with [[wikilink]] citations, staleness annotations, and identified knowledge gaps. ## Quick Start Ask the assistant to search your Obsidian wiki for everything related to a topic and synthesize an answer with page citations.
