roasbeef/obsidian-claude-code

Overview

This skill provides semantic search and Dataview-style SQL queries across an Obsidian vault. It surfaces related notes by meaning, queries frontmatter metadata, and returns programmatic JSON or human-readable results. Use it to find notes by topic, list tasks, or answer questions about vault contents.

How this skill works

The skill builds a vector index of note chunks and stores frontmatter metadata in a sqlite database. Semantic queries use embeddings and sqlite-vec vector search; metadata queries run SQL over the notes table. Results include snippets, metadata columns, and optional JSON output for automation.

When to use it

• Find notes by meaning or related concepts (semantic search).
• Locate or filter notes by frontmatter metadata (Dataview-style SQL).
• List tasks, open items, or items due within a range.
• Build dashboards or aggregate counts across tags, folders, or projects.
• Combine semantic relevance with metadata filters (e.g., related notes that are open tasks).

Best practices

• Rebuild the index after many file changes or bulk imports to keep results fresh.
• Use --folder or SQL WHERE filters to narrow results and reduce noise.
• Limit semantic results (n-results) when exploring; increase only for broad discovery.
• Use the JSON output for downstream scripts or integrations.
• Exclude system folders and large binary assets from indexing to save space and speed up queries.

Example use cases

• Find notes related to a research topic: semantic search returns the most relevant chunks and source paths.
• Show open high-priority tasks: run a dataview SQL query filtering status and priority.
• List investing positions by folder and metadata like ticker and expiry via SQL.
• Combine: find notes about “portfolio risk” that are in the investing folder and marked open.
• Rebuild the index after a vault import to ensure new notes appear in searches.