The AlloyDB for PostgreSQL hybrid_search
function accepts several parameters to control the search
and fusion process. Understanding these parameters helps you refine search
results to meet your specific needs. This document provides an overview of the
available parameters. To learn how to perform hybrid search using the vector
extension, see Run a hybrid vector similarity search
.
Main parameters
| Parameter | Type | Description |
|---|---|---|
search_inputs
|
JSONB[]
|
Required. An array of JSONB objects, where each object defines a search component. |
id_type
|
anyelement
|
The data type of the output ID column. Defaults to NULL::TEXT
. |
include_json_output
|
BOOLEAN
|
Includes detailed JSON output in the results. Defaults to FALSE
. Setting this to TRUE
can impact query performance. |
fusion_algorithm
|
TEXT
|
The fusion algorithm to use. The only supported value is 'RRF'
. Defaults to 'RRF'
. |
Vector search parameters
Set the data_type
field to 'vector'
in the search_inputs
JSONB object.
| Parameter | Type | Description |
|---|---|---|
data_type
|
TEXT
|
Required. Must be 'vector'
. |
table_name
|
TEXT
|
Required. The name of the table to search. |
key_column
|
TEXT
|
Required. The name of the document ID column. |
vec_column
|
TEXT
|
Required. The name of the vector column. |
distance_operator
|
TEXT
|
Required. The distance operator to use, for example, '<=>'
for cosine distance. |
query_vector
|
TEXT
|
Required. The string representation of the vector to search for. |
limit
|
INTEGER
|
Required. The maximum number of results to return from this search component. |
weight
|
FLOAT
|
The contribution of this component to the final RRF score, from 0.0 to 1.0. If you don't provide weights, they are distributed evenly. |
filter_condition
|
TEXT
|
A SQL WHERE
clause to apply to this component's subquery. |
Text search parameters
Set the data_type
field to 'text'
in the search_inputs
JSONB object.
| Parameter | Type | Description |
|---|---|---|
data_type
|
TEXT
|
Required. Must be 'text'
. |
table_name
|
TEXT
|
Required. The name of the table to search. |
key_column
|
TEXT
|
Required. The name of the document ID column. |
text_column
|
TEXT
|
Required. The name of the tsvector
or text column. |
query_text_input
|
TEXT
|
Required. The text string for keyword matching. |
limit
|
INTEGER
|
Required. The maximum number of results to return from this search component. |
weight
|
FLOAT
|
The contribution of this component to the final RRF score, from 0.0 to 1.0. If you don't provide weights, they are distributed evenly. |
ranking_function
|
TEXT
|
The full-text search ranking function. For GIN indexes, set to 'ts_rank'
.
For RUM indexes, set to '<=>'
. Defaults to 'ts_rank'
. |
query_parser
|
TEXT
|
The full-text search query parser function. Defaults to 'ai.g_to_tsquery'
. |
filter_condition
|
TEXT
|
A SQL WHERE
clause to apply to this component's subquery. |
What's next
- Learn how to run a hybrid vector similarity search .
