Function searchBySimilarity

Search for chunks similar to a query using vector similarity

Uses SurrealDB's built-in vector search with KNN operator for efficiency.

Process:

1. Generate embedding for the query
2. Use SurrealDB's <|limit,COSINE|> operator to find nearest neighbors
3. Join with chunk table to get chunk content and metadata
4. Filter by threshold and optional filters (project, file extension)
5. Enrich with file metadata

Example

const results = await searchBySimilarity(
  "function that validates user input",
  { limit: 5, threshold: 0.3, projectId: "project:abc123" }
);
Copy