Obsidian has been used a lot for agent memory. And I wondered...what if you could query any file or cloud database directly from Obsidian and cache the result? That's what the DuckDB + MotherDuck Obsidian plugin does. Run SQL on local files (Parquet, CSV) or your cloud database, inline result in any markdown note. Results are cached as markdown tables, so your agent reads data instantly, no re-query needed. And thanks to the obsidian CLI, everything can be done through your favorite agent (writing/running/caching the query)