AnalyticDB:QueryCollectionData

Instance ID.

Collection name.

Namespace.

Password for the namespace.

Content for full-text search. When this value is empty, only vector search is used; when it is not empty, both vector and full-text search are used.

Filter conditions for the data to be queried, in SQL WHERE format. It is an expression that returns a boolean value (true or false). Conditions can be simple comparison operators such as equal (=), not equal (<> or !=), greater than (>), less than (<), greater than or equal to (>=), less than or equal to (<=), or more complex expressions combined with logical operators (AND, OR, NOT), as well as conditions using keywords like IN, BETWEEN, and LIKE.

Set the number of top results to return.

Vector data, with the same dimension as specified in the CreateCollection API.

Region ID where the instance is located.

Similarity algorithm used during retrieval. Value descriptions:

• l2: Euclidean distance.
• ip: Inner product (dot product) distance.
• cosine: Cosine similarity.

Whether to return vector data. Value descriptions:

• true: Return vector data.
• false: Do not return vector data, used for full-text search scenarios.

Dual-path recall algorithm, default is empty (i.e., directly compare and sort the scores of vectors and full-text).

Available values:

• RRF: Reciprocal rank fusion, with a parameter k controlling the fusion effect. See HybridSearchArgs configuration for details;
• Weight: Weighted sorting, using a parameter alpha to control the score ratio of vectors and full-text, then sorting. See HybridSearchArgs configuration for details;
• Cascaded: Perform full-text search first, then vector search based on the full-text results;

The parameters of the two-way retrieval algorithm. The following parameters are supported:

• When HybridSearch is set to RRF, the scores are calculated by using the 1/(k+rank_i) formula. The constant k is a positive integer that is greater than 1.

{ 
   "RRF": {
    "k": 60
   }
}

• When HybridSearch is set to Weight, the scores are calculated by using the alpha * vector_score + (1-alpha) * text_score formula. The alpha parameter specifies the proportion of the vector search score and the full-text search score and ranges from 0 to 1. A value of 0 specifies full-text search and a value of 1 specifies vector search.

{ 
   "Weight": {
    "alpha": 0.5
   }
}

Defaults to empty, indicating the field for sorting. Does not support hybrid search scenarios.

The field must belong to metadata or be a default field in the table, such as id. The supported formats are:

• A single field, e.g., chunk_id;
• Multiple fields, separated by commas, e.g., block_id, chunk_id;
• Supports reverse order, e.g., block_id DESC, chunk_id DESC;

Defaults to empty, indicating the starting point for pagination queries. Does not support hybrid search scenarios.

The value must be >= 0. When this value is not empty, it will return Total, which indicates the total number of hits. This parameter works with TopK. For example, to paginate 20 and retrieve chunks with chunk_id from 0 to 44, you need to make three requests:

• Offset=0, TopK=20 returns chunk_id 0~19
• Offset=20, TopK=20 returns chunk_id 20~39
• Offset=30, TopK=20 returns chunk_id 40~44

Defaults to empty, indicating the metadata fields to return. Multiple fields should be separated by commas.

The ID of the Workspace composed of multiple database instances. This parameter and DBInstanceId cannot both be empty. If both are specified, this parameter takes precedence.

Uses another relational table to filter vector data (similar to a Join function).