Vector search

In the left-side secondary navigation pane, click Similar Vector Search. Fill in the required information and click Search. The search returns the results for similar vectors.

1. Vector search in a single-vector collectionThe parameters for the vector search are described below.

   Parameter	Corresponding API parameter name	Description	Required
   Query Vector	vector	The vector data. For example, `[1.0,2.0,3.0,4.0]`. Note The vector dimensions and data type must be the same as those of the collection.	Yes
   Filter Condition	filter	The filter condition. It must follow the SQL `WHERE` clause specification. For more information, see Filtered search.	No
   Partition	partition	The partition name. The default value is `default`. Select a partition as needed.	Yes
   TopK	topk	The maximum number of vectors to return. The default value is 10. The maximum value is 1024.	Yes

   The parameters in the returned results are described below.

   Parameter	Corresponding API parameter name	Description
   Distance	score	Vector similarity. The numeric representation of the distance between vectors varies with the distance metric. For more information, see What is a vector?. The results are sorted by vector similarity in descending order.
   Primary Key ID	id	The primary key ID of the similar vector.
   Vector	vector	The vector data. For example, [1.0,2.0,3.0,4.0].
   Fields	fields	The JSON field parameters. For example, {"price":100,"type":"dress"}.
   Partition	partition	The partition where the similar vector is located.

2. Vector search in a multi-vector collection

   Note

   • Click Add Query Vector to add a record. The number of query vectors cannot exceed the number of vector fields defined in the collection.

   • Click the delete button on the right to delete the record.

   The parameters for the vector search are described below.

   Parameter	Corresponding API parameter name	Description	Required
   Vector Name	{VectorName}	Custom: The following requirements must be met: The name must be 3 to 32 characters in length. The name must consist of uppercase letters, lowercase letters, digits, underscores (_), and hyphens (-). For example, `vector1`, `vector_1`, or `vector_a_name`. The vector name must be unique within the collection. Two identical vector names cannot exist at the same time. Note You can only select a vector name that was specified when the collection was created.	Yes
   Query Vector	vector	The vector data. For example, `[1.0,2.0,3.0,4.0]`. Note The vector dimensions and data type must be the same as those defined when the collection was created.	Yes
   Number of candidate vectors	num_candidates	The number of results to recall for a single vector. The default value is the same as `topk` (10).	No
   Weight	Optional[Dict[str, float]	This parameter is required only when **Sorting Method** is set to **WeightRank**. By default, the weights are equal (1.0:1.0:1.0...). For more information, see WeightedRanker.	Yes
   Sorting Method	RrfRanker/ WeightedRanker	Supports RRFRank and WeightRank. For more information, see RrfRanker and WeightedRanker.	Yes
   Constant	rank_constant	This parameter is valid only when **Sorting Method** is set to **RRFRank**. For example, if `rank_constant` is 10, the 10 most similar results are returned for each vector. The default value is 60. For more information, see RrfRanker.	Yes
   Filter Condition	filter	The filter condition. It must follow the SQL `WHERE` clause specification. For more information, see Filtered search.	No
   Partition	partition	The partition name. The default value is `default`. Select a partition as needed.	Yes
   TopK	topk	The maximum number of vectors to return. The default value is 10. The maximum value is 1024.	Yes

   The parameters in the returned results are described below.

   Parameter	Corresponding API parameter name	Description
   Distance	score	Vector similarity. The numeric representation of the distance between vectors varies with the distance metric. For more information, see What is a vector?. The results are sorted by vector similarity in descending order.
   Primary Key ID	id	The primary key ID of the similar vector.
   Vector	vector	The vector data. For example, [1.0,2.0,3.0,4.0].
   Properties	fields	The JSON field parameters. For example, {"price":100,"type":"dress"}.
   Partition	partition	The partition where the similar vector is located.