Options for the find method.

Field

sort - The sort order to pick which document to return if the filter selects multiple documents.

Field

vector - An optional vector to use for the appropriate dimensionality to perform an ANN vector search on the collection.

Field

projection - Specifies which fields should be included/excluded in the returned documents.

Field

limit - Max number of documents to return in the lifetime of the cursor.

Field

skip - Number of documents to skip if using a sort.

Field

includeSimilarity - If true, include the similarity score in the result via the $similarity field.

See

Collection.find

interface FindOptions {
    includeSimilarity?: boolean;
    limit?: number;
    projection?: Projection;
    skip?: number;
    sort?: Sort;
    vector?: number[];
    vectorize?: string;
}

Properties

includeSimilarity?: boolean

If true, include the similarity score in the result via the $similarity field.

If false, do not include the similarity score in the result.

Defaults to false.

Default Value

false

Example

const doc = await collection.findOne({}, {
  sort: {
  $vector: [.12, .52, .32],
  },
  includeSimilarity: true,
});

console.log(doc?.$similarity);
limit?: number

Max number of documents to return. Applies over the whole result set, not per page. I.e. if the result set has 1000 documents and limit is 100, only the first 100 documents will be returned, but it'll still be fetched in pages of some N documents, regardless of if N < or > 100.

projection?: Projection

Specifies which fields should be included/excluded in the returned documents.

If not specified, all fields are included.

When specifying a projection, it's the user's responsibility to handle the return type carefully, as the projection will, of course, affect the shape of the returned documents. It may be a good idea to cast the returned documents into a type that reflects the projection to avoid runtime errors.

Example

interface User {
  name: string;
  age: number;
}

const collection = db.collection<User>('users');

const doc = await collection.findOne({}, {
  projection: {
  _id: 0,
  name: 1,
  },
  vector: [.12, .52, .32],
  includeSimilarity: true,
}) as { name: string, $similarity: number };

// Ok
console.log(doc.name);
console.log(doc.$similarity);

// Causes type error
console.log(doc._id);
console.log(doc.age);
skip?: number

Number of documents to skip. Only works if a sort is provided.

sort?: Sort

The order in which to apply the update if the filter selects multiple documents.

If multiple documents match the filter, only one will be updated.

Defaults to null, where the order is not guaranteed.

Default Value

null
vector?: number[]

An optional vector to use of the appropriate dimensionality to perform an ANN vector search on the collection to find the closest matching document.

This is purely for the user's convenience and intuitiveness—it is equivalent to setting the $vector field in the sort field itself. The two are interchangeable, but mutually exclusive.

If the sort field is already set, an error will be thrown. If you really need to use both, you can set the $vector field in the sort object directly.

Deprecated

  • Prefer to use sort: { $vector: [...] } instead
vectorize?: string

NOTE: This feature is under current development.

Deprecated

  • Prefer to use sort: { $vectorize: '...' } instead