sspaeti/obsidian-note-taking-assistant

# Technical Summary Combines semantic search (BGE-M3 embeddings) with explicit wikilink graph traversal to surface both directly-linked notes and semantically similar unconnected content. The ingestion pipeline parses Obsidian markdown into a DuckDB schema with vector embeddings, chunks, and hyperedge tables, enabling specialized queries like "hidden connections" (semantically similar but unlinked notes) and graph-boosted search that ranks results by both embedding similarity and graph proximity. Includes a Python CLI for querying and a web app for browsing the knowledge graph.