Function semanticSearch

semanticSearch(
    vault: Vault,
    args: {
        folder?: string;
        limit?: number;
        min_score?: number;
        query: string;
    },
): Promise<
    {
        matches: SemanticHit[];
        method: "tfidf-cosine";
        query: string;
        total_docs: number;
    },
>
Pure-JS lexical-semantic search via TF-IDF cosine similarity.

Builds (or reuses cached) per-vault TF-IDF index, then ranks notes by cosine similarity of the query vector against each body vector. Catches "related-term" recall that the substring path of searchText misses (e.g. searching "retrieval" will surface notes about "recall" if the vocabulary co-occurs). Zero native deps — works on every platform with no model download. For full ML retrieval use embeddingsSearch; for graceful-degradation fusion use searchHybrid.
Parameters
- vault: Vault
  The vault to search.
- args: { folder?: string; limit?: number; min_score?: number; query: string }
  query is required. limit defaults to 10. min_score defaults to 0.05 — anything below is pruned. folder restricts to a subdirectory.
Returns Promise<
    {
        matches: SemanticHit[];
        method: "tfidf-cosine";
        query: string;
        total_docs: number;
    },
>
An envelope with query, total_docs (corpus size), method (always "tfidf-cosine"), and matches sorted by score desc.
Throws
If query is empty / whitespace-only.
Example
```
const result = await semanticSearch(vault, {
  query: "vector retrieval cosine",
  limit: 5
});
for (const hit of result.matches) {
  console.log(hit.path, hit.score, hit.matched_terms);
}
```
- Defined in tools/search.ts:592

Function semanticSearch

Parameters

Returns Promise< { matches: SemanticHit[]; method: "tfidf-cosine"; query: string; total_docs: number; },>

Throws

Example

Settings

Returns Promise<
{
matches: SemanticHit[];
method: "tfidf-cosine";
query: string;
total_docs: number;
},
>