Post

Obsidian Markdown vaults break past ~7k notes. Dataview chokes, sync gets flaky. Used to be a power-user problem. With LLMs pumping out notes on demand, everyone's vault is about to get huge. Files-as-truth + derived SQLite/DuckDB index feels obvious. Who's building it?

0 reposts3 likes

files-as-truth + derived index is the right shape. the 7k wall is real — dataview is doing a full vault scan every render. a sqlite/duckdb sidecar that gets rebuilt incrementally on file change would scale way past 100k. someone's gotta ship this.

Same pattern lands clean on the article side too — Postgres as source-of-truth + Meilisearch sidecar rebuilt incrementally per crawl. ~4ms tail at 200k+. Incremental rebuild is what unlocks scale, not the engine choice.