logancyang
Index Only Frontmatter/Properties for Vault Q&A
Description Currently, the Vault Q&A feature indexes entire note contents, which can be slow and inefficient when searching for specific files. I propose an option to index only the frontmatter/properties of notes.
Why This Matters
-
Faster Search: Since most of my notes use frontmatter extensively (aliases, descriptions, tags ...), searching only within frontmatter will significantly speed up retrieval.
-
Reduced Token Usage: By limiting the indexed content to frontmatter, queries will require fewer tokens, improving efficiency and reducing API costs for remote models.
-
More Accurate Results: Instead of scanning entire note texts, which may contain irrelevant information, this approach ensures responses focus only on structured metadata.
Proposed Implementation
- Add a setting to enable "Frontmatter-Only Indexing" for Vault Q&A.
Frontmatter / properties are usually small pieces of keywords, dates and tags that are usually aimed for structured query and full-text search. Leveraging semantic search over it will be awkward. Not sure if it works for most people's use cases.
It would work for me, as i add a short description of the notes topic and role into a frontmatter tag specifically with the purpose of being good for semantic search
