AI Feature Functions

gs_index_advise(text)
Description: Recommends an index for a single query statement.

Parameter: SQL statement string

Return type: record

For details, see section "DBMind: Autonomous Database O&M > AI Sub-functions of DBMind > Index-advisor: Index Recommendation > Single-Query Index Recommendation" in Feature Guide.
hypopg_create_index(text, [text])
Description: Creates a virtual index.

Parameter: character string of the statement for creating an index, level of the created virtual index (optional)

Return type: record

For details, see section "DBMind: Autonomous Database O&M > AI Sub-functions of DBMind > Index-advisor: Index Recommendation > Virtual Index" in Feature Guide.
hypopg_display_index([text])
Description: Displays information about all created virtual indexes.

Parameter: level of the virtual index to be displayed (optional)

Return type: record

For details, see section "DBMind: Autonomous Database O&M > AI Sub-functions of DBMind > Index-advisor: Index Recommendation > Virtual Index" in Feature Guide.
hypopg_drop_index(oid)
Description: Deletes a specified virtual index.

Parameter: OID of the index

Return type: Boolean

For details, see section "DBMind: Autonomous Database O&M > AI Sub-functions of DBMind > Index-advisor: Index Recommendation > Virtual Index" in Feature Guide.
hypopg_reset_index([text])
Description: Clears all virtual indexes.

Parameter: level of the virtual index to be cleared (optional)

Return type: none

For details, see section "DBMind: Autonomous Database O&M > AI Sub-functions of DBMind > Index-advisor: Index Recommendation > Virtual Index" in Feature Guide.
hypopg_estimate_size(oid)
Description: Estimates the space required for creating a specified index.

Parameter: OID of the index

Return type: int8

For details, see section "DBMind: Autonomous Database O&M > AI Sub-functions of DBMind > Index-advisor: Index Recommendation > Virtual Index" in Feature Guide.
check_engine_status(ip text, port text)
Description: Tests whether a predictor engine provides services on a specified IP address and port.

Parameter: IP address and port number of the predictor engine.

Return type: text

This function is unavailable in the current version.
encode_plan_node(optname text, orientation text, strategy text, options text, dop int8, quals text, projection text)
Description: Encodes the plan operator information in the input parameters.

Parameter: plan operator information

Return type: text

This function is an internal function. You are advised not to use it directly.
model_train_opt(template text, model text)
Description: Trains a given query performance prediction model.

Parameters: template name and model name of the performance prediction model

Return type: tartup_time_accuracy FLOAT8, total_time_accuracy FLOAT8, rows_accuracy FLOAT8, peak_memory_accuracy FLOAT8

This function is unavailable in the current version.
track_model_train_opt(ip text, port text)
Description: Returns the training log address of the specified IP address and port predictor engine.

Parameter: IP address and port number of the predictor engine

Return type: text

This function is unavailable in the current version.
encode_feature_perf_hist(datname text)
Description: Encodes historical plan operators collected in the target database.

Parameter: database name

Return type: queryid bigint, plan_node_id int, parent_node_id int, left_child_id int, right_child_id int, encode text, startup_time bigint, total_time bigint, rows bigint, and peak_memory int
gather_encoding_info(datname text)
Description: Calls encode_feature_perf_hist to save the encoded data persistently.

Parameter: database name

Return type: int
db4ai_predict_by_bool (text, VARIADIC "any")
Description: Obtains a model whose return value is of the Boolean type for model inference. This function is an internal function. You are advised to use the PREDICT BY syntax for inference.

Parameter: model name and input column name of the inference task

Return type: Boolean

You are advised not to use it directly.
db4ai_predict_by_float4(text, VARIADIC "any")
Description: Obtains a model whose return value is of the float4 type for model inference. This function is an internal function. You are advised to use the PREDICT BY syntax for inference.

Parameter: model name and input column name of the inference task

Return type: float

You are advised not to use it directly.
db4ai_predict_by_float8(text, VARIADIC "any")
Description: Obtains a model whose return value is of the float8 type for model inference. This function is an internal function. You are advised to use the PREDICT BY syntax for inference.

Parameter: model name and input column name of the inference task

Return type: float

You are advised not to use it directly.
db4ai_predict_by_int32(text, VARIADIC "any")
Description: Obtains a model whose return value is of the int32 type for model inference. This function is an internal function. You are advised to use the PREDICT BY syntax for inference.

Parameter: model name and input column name of the inference task

Return type: int

You are advised not to use it directly.
db4ai_predict_by_int64(text, VARIADIC "any")
Description: Obtains a model whose return value is of the int64 type for model inference. This function is an internal function. You are advised to use the PREDICT BY syntax for inference.

Parameter: model name and input column name of the inference task

Return type: int

You are advised not to use it directly.
db4ai_predict_by_numeric(text, VARIADIC "any")
Description: Obtains a model whose return value is of the numeric type for model inference. This function is an internal function. You are advised to use the PREDICT BY syntax for inference.

Parameter: model name and input column name of the inference task

Return type: numeric

You are advised not to use it directly.
db4ai_predict_by_text(text, VARIADIC "any")
Description: Obtains a model whose return value is of the character type for model inference. This function is an internal function. You are advised to use the PREDICT BY syntax for inference.

Parameter: model name and input column name of the inference task

Return type: text

You are advised not to use it directly.
db4ai_predict_by_float8_array(text, VARIADIC "any")
Description: Obtains a model whose return value is of the character type for model inference. This function is an internal function. You are advised to use the PREDICT BY syntax for inference.

Parameter: model name and input column name of the inference task

Return type: text

You are advised not to use it directly.
gs_explain_model(text)
Description: Obtains the model whose return value is of the character type for text-based model parsing.

Parameter: model name

Return type: text

gs_ai_stats_explain(text, text[])

Description: Prints the intelligent statistics of multiple columns in the corresponding table and columns.

Parameters: table name and column name collection.

Return type: text

Example:

     
          gaussdb=# SET enable_ai_stats=1;
gaussdb=# drop table db4ai_bayesnet;
gaussdb=# create table db4ai_bayesnet(attr1 VARCHAR(256), attr2 VARCHAR(256));
gaussdb=# INSERT INTO db4ai_bayesnet SELECT 'x','x' FROM generate_series(1, 10000) AS s(i);
gaussdb=# ANALYZE db4ai_bayesnet((attr1,attr2));
gaussdb=# select gs_ai_stats_explain('db4ai_bayesnet', ARRAY['attr1', 'attr2']);

ai_watchdog_detection_warnings()
Description: Obtains the risk alarm information of the AI watchdog.

Parameter: none

Return type: record

Example:
1

gaussdb=# select * from ai_watchdog_detection_warnings();
ai_watchdog_monitor_status(int)
Description: Obtains the monitoring information of the AI watchdog.

Parameter: the upper limit of the length of the returned monitoring sequence. The value range is (0,100). The default value is 10.

Return type: record

Example:
1

gaussdb=# select * from ai_watchdog_monitor_status();
ai_watchdog_parameters()
Description: Obtains the internal parameters or status information of the AI watchdog.

Parameter: none

Return type: record

Example:
1

gaussdb=# select * from ai_watchdog_parameters();