Minutes Ingest

Works with PARA, Obsidian, and wiki formats. The default mode only pulls from parsed frontmatter so there's no hallucination risk, though you can opt into LLM extraction if you want deeper analysis of transcript bodies. The safety model is solid: confidence thresholds, deduplication, provenance tracking, and dry run mode for first timers. Most useful when you've accumulated dozens of meeting files and want to backfill your knowledge base without manually copying facts into profile pages. If your meetings don't have structured frontmatter yet, this won't extract anything, which is correct behavior but worth knowing upfront.