📘 Complete Code example

Here's the complete example code to implement a hybrid search against the books collection.

db.books.aggregate([
  // Vector search on book synopsis embeddings
  {
    $vectorSearch: {
      index: "books_synopsis_vector",
      path: "embeddings",
      queryVector: [0.023, -0.345, 0.122, ...], // Example query vector
      numCandidates: 100,
      limit: 20
    }
  },
  // Add vector search score
  {
    $set: {
      vector_score: { $meta: "vectorSearchScore" }
    }
  },
  // Combine with text search on title and author
  {
    $search: {
      index: "books_text_index",
      compound: {
        should: [
          {
            text: {
              query: "mystery novel",
              path: "title",
              score: { boost: { value: 3 } }
            }
          },
          {
            text: {
              query: "mystery novel",
              path: "authors.name",
              score: { boost: { value: 2 } }
            }
          }
        ]
      }
    }
  },
  // Add text search score
  {
    $set: {
      text_score: { $meta: "searchScore" }
    }
  },
  // Combine scores
  {
    $addFields: {
      combined_score: {
        $add: [
          { $multiply: ["$vector_score", 0.6] },
          { $multiply: ["$text_score", 0.4] }
        ]
      }
    }
  },
  // Sort and limit results
  { $sort: { combined_score: -1 } },
  { $limit: 5 },
  // Project final results
  {
    $project: {
      title: 1,
      authors: 1,
      synopsis: 1,
      combined_score: 1,
      vector_score: 1,
      text_score: 1
    }
  }
])