Oracle | DBMS

Oracle DBMS_VECTOR Version 26ai

General Information

Library Note

Morgan's Library Page Header

The best Oracle News for FY2026

Oracle Database 26ai will be available on generic Linux platforms in January and soon on AIX and Windows

Purpose In Database Artificial Intelligence (AI) Searches. Contains support for load, create index, search, accuracy reporting, and LangChain support.

AUTHID CURRENT_USER

Constants

Name	Data Type	Value
General
PREFERENCE_CHUNKER	VARCHAR2(7)	'CHUNKER'
PREFERENCE_VECTORIZER	VARCHAR2(10)	'VECTORIZER'

Data Types TYPE vector_array_t AS TABLE OF CLOB;

Dependencies SELECT name FROM dba_dependencies WHERE referenced_name = 'DBMS_VECTOR' UNION SELECT referenced_name FROM dba_dependencies WHERE name = 'DBMS_VECTOR' ORDER BY 1;

Returns 44 objects

Dependencies

ALL_CREDENTIALS	DBMSPRIV_CAPTURE	JSON_KEY_LIST
ALL_INDEXES	DBMS_STANDARD	JSON_OBJECT
DBA_DB_LINKS	DBMS_SYSTEM	PLITBLM
DBMS_ASSERT	DBMS_UTILITY	UTL_ENCODE
DBMS_CREDENTIAL	DBMS_VECTOR_CHAIN	UTL_HTTP
DBMS_CRYPTO	DBMS_VECTOR_INTERNAL	UTL_I18N
DBMS_DATA_MINING	DBMS_VECTOR_LIB	UTL_RAW
DBMS_LOG	DUAL	VECTOR_ARRAY_T
DBMS_METADATA	JSON_ARRAY_T	V_$PARAMETER
DBMS_PDB_LIB	JSON_ELEMENT_T

Documented Yes: Partial

Exceptions

Error Code	Reason
ORA-20001	EXCP_CRED_NOT_EXIST
ORA-20002	EXCP_PROVIDER_ERR
ORA-20003	EXCP_GU_ERR (General Error)

First Available 23ai

Security Model Owned by SYS with EXECUTE granted to PUBLIC.

Source {ORACLE_HOME}/rdbms/admin/dbmsvector.sql
{ORACLE_HOME}/rdbms/admin/prvtvector.plb

Subprograms

CHECK_MAX_CALLS	LOAD_ONNX_MODEL_CLOUD
CREATE_CHECKPOINT	LOG_TRACE
CREATE_CREDENTIAL	PURGE_CHECKPOINT
CREATE_INDEX	QUERY
DISABLE_CHECKPOINT	REBUILD_INDEX
DROP_CREDENTIAL	RERANK
DROP_ONNX_MODEL	SET_SGA_PERCENTAGE
ENABLE_CHECKPOINT	SET_TRACE
GET_INDEX_STATUS	TOGGLE_DEBUG_MODE
INDEX_ACCURACY_QUERY	TOGGLE_INDEX_VECTOR_MEMORY_ADVISOR_VERIFY
INDEX_ACCURACY_REPORT	UTL_TO_CHUNKS
INDEX_VECTOR_MEMORY_ADVISOR	UTL_TO_EMBEDDING
LIST_MODELS	UTL_TO_EMBEDDINGS
LIST_PROVIDERS	UTL_TO_GENERATE_TEXT
LOAD_ONNX_MODEL	UTL_TO_RERANK

CHECK_MAX_CALLS (new 23ai)

Undocumented dbms_vector.check_max_calls( method_name IN VARCHAR2, provider_name IN VARCHAR2, params IN JSON);

TBD

CREATE_CHECKPOINT (new 23ai)

This procedure takes a full checkpoint for an HNSW index dbms_vector.create_checkpoint( idx_owner IN VARCHAR2, idx_name IN VARCHAR2);

TBD

CREATE_CREDENTIAL (new 23ai)

Store authentication information in the database dbms_vector.create_credential( credential_name IN VARCHAR2, params IN JSON);

DECLARE

       aiCred json_object_t;

      BEGIN

        aiCred := json_object_t();

        aiCred.put('USER_OCID', 'user ocid')

        aiCred.put('TENANCY_OCID', 'tenancy ocid value');

        aiCred.put('COMPARTMENT_OCID', 'compartment ocid value');

        aiCred.put('PRIVATE_KEY', 'private key value');

        aiCred.put('FINGER_PRINT', 'fingerprint value');

      

        dbms_output.put_line(credObj.TO_STRING);

        dbms_vector.creeate_credential(credential_name => 'ORA_AICRED', params=> json(jo.TO_STRING));

      END;

      /

CREATE_INDEX (new 23ai)

Creates a vector index dbms_vector.create_index( idx_name IN VARCHAR2, table_name IN VARCHAR2, idx_vector_col IN VARCHAR2, idx_include_cols IN VARCHAR2 DEFAULT NULL, idx_partitioning_scheme IN VARCHAR2 DEFAULT 'LOCAL', idx_organization IN VARCHAR2, idx_distance_metric IN VARCHAR2 DEFAULT 'COSINE', idx_accuracy IN NUMBER DEFAULT 90, idx_parameters IN CLOB, idx_parallel_creation IN NUMBER DEFAULT 1);

TBD

DISABLE_CHECKPOINT (new 23ai)

Disables the Checkpoint feature for a vector index user and index name dbms_vector.disable_checkpoint( idx_owner IN VARCHAR2 DEFAULT NULL, idx_name IN VARCHAR2 DEFAULT NULL);

TBD

DROP_CREDENTIAL (new 23ai)

Drop authentication credential from the database dbms_vector.drop_credential(credential_name IN VARCHAR2);

exec dbms_vector.drop_credential('ORA_AICRED');

      

      PL/SQL procedure successfully completed.

DROP_ONNX_MODEL (new 23ai)

Drop ONNX model dbms_vector.drop_onnx_model( model_name IN VARCHAR2, force IN BOOLEAN DEFAULT FALSE);

exec dbms_vector.drop_onnx_model('doc_model', TRUE);

      

      PL/SQL procedure successfully completed.

ENABLE_CHECKPOINT (new 23ai)

Enables the Checkpoint feature for a vector index user and index name dbms_vector.enaable_checkpoint( idx_owner IN VARCHAR2 DEFAULT NULL, idx_name IN VARCHAR2 DEFAULT NULL, tbs_name IN VARCHAR2 DEFAULT NULL);

TBD

GET_INDEX_STATUS (new 23ai)

Describes the status of a vector index creation dbms_vector.get_index_status( owner_name IN VARCHAR2, index_name IN VARCHAR2);

TBD

INDEX_ACCURACY_QUERY (new 23ai)

Accuracy reporting query

Overload 1 dbms_vector.index_accuracy_query( owner_name IN VARCHAR2, index_name IN VARCHAR2, qv IN VECTOR, top_k IN NUMBER, target_accuracy IN NUMBER) RETURN VARCHAR2;

TBD

Overload 2 dbms_vector.index_accuracy_query( owner_name IN VARCHAR2, index_name IN VARCHAR2, qv IN VECTOR, top_k IN NUMBER, query_param IN JSON) RETURN VARCHAR2;

TBD

INDEX_ACCURACY_REPORT (new 23ai)

Computes accuracy using the captured query vectors for a given finding dbms_vector.index_accuracy_report( owner_name IN VARCHAR2, index_name IN VARCHAR2, start_time IN TIMESTAMP WITH TIME ZONE := NULL, end_time IN TIMESTAMP WITH TIME ZONE := NULL) RETURN NUMBER;

TBD

INDEX_VECTOR_MEMORY_ADVISOR (new 23ai)

Determines the vector memory size that is needed for a vector index

Overload 1 dbms_vector.index_vector_memory_advisor( index_type IN VARCHAR2, num_vectors IN NUMBER, dim_count IN NUMBER, dim_type IN VARCHAR2, parameter_json IN CLOB DEFAULT NULL, response_json OUT CLOB);

TBD

Overload 2 dbms_vector.index_vector_memory_advisor( table_owner IN VARCHAR2, table_name IN VARCHAR2, column_name IN VARCHAR2, index_type IN VARCHAR2, parameter_json IN CLOB DEFAULT NULL, response_json OUT CLOB);

TBD

LIST_MODELS (new 23ai)

rETURNS a list of models and their capabilities dbms_vector.list_models(params IN JSON DEFAULT NULL) RETURN CLOB;

TBD

LIST_PROVIDERS (new 23ai)

Returns a list of providers dbms_vector.list_providers(params IN JSON DEFAULT NULL) RETURN CLOB;

TBD

LOAD_ONNX_MODEL (new 23ai)

Loads a model from an .onnx model file stored in a local directory.

Overload 1 dbms_vector.load_onnx_model( directory IN VARCHAR2, file_name IN VARCHAR2, model_name IN VARCHAR2, metadata IN JSON DEFAULT JSON('{"function" : "embedding", ' || '"embeddingOutput" : "embedding" , "input" : {"input" : {"input" : ["DATA"]}}');

TBD

Loads a model from an .onnx model file stored in Cloud Object storage

Overload 2 dbms_vector.load_onnx_model( model_name IN VARCHAR2, model_data IN BLOB, metadata IN JSON DEFAULT JSON('{"function" : "embedding", ' || '"embeddingOutput" : "embedding" , "input" : {"input" : {"input" : ["DATA"]}}'));

TBD

LOAD_ONNX_MODEL_CLOUD (new 23ai)

Loads an ONNX model from object storage into the database dbms_vector.load_onnx_model_cloud( model_name IN VARCHAR2, credential IN VARCHAR2, uri IN VARCHAR2, metadata IN JSON default json('{"function" : "embedding", ' || '"embeddingOutput" : "embedding" , "input": {"input": ["DATA"]}}'));

TBD

LOG_TRACE (new 23ai)

Undocumented dbms_vector.log_trace(log_str IN VARCHAR2);

exec dbms_vector.log_trace('TEST');

      

      PL/SQL procedure successfully completed.

PURGE_CHECKPOINT (new 23ai)

Purges the existing Checkpoint for a vector index user and index name dbms_vector.purge_checkpoint( idx_owner IN VARCHAR2, idx_name IN VARCHAR2);

TBD

QUERY (new 23ai)

Queries a database schema for a similarity search

Overload 1 dbms_vector.query( tab_name IN VARCHAR2, vec_col_name IN VARCHAR2, query_vector IN CLOB, top_k IN NUMBER, vec_proj_cols IN sys.json_array_t DEFAULT NULL, idx_name IN VARCHAR2 DEFAULT NULL, distance_metric IN BOOLEAN DEFAULT 'COSINE', use_index IN BOOLEAN DEFAULT TRUE, accuracy IN NUMBER DEFAULT 90, idx_parameters IN CLOB DEFAULT NULL);

TBD

Supports a query vector passed as a vector function

Overload 2 dbms_vector.query( tab_name IN VARCHAR2, vec_col_name IN VARCHAR2, query_vector IN VECTOR, top_k IN NUMBER, vec_proj_cols IN sys.json_array_t DEFAULT NULL, idx_name IN VARCHAR2 DEFAULT NULL, distance_metric IN BOOLEAN DEFAULT 'COSINE', use_index IN BOOLEAN DEFAULT TRUE, accuracy IN NUMBER DEFAULT 90, idx_parameters IN CLOB DEFAULT NULL) RETURN sys.json_array_t;

TBD

REBUILD_INDEX (new 23ai)

Rebuilds a vector index which is performed by dropping and recreating the original index dbms_vector.rebuild_index( idx_name IN VARCHAR2, table_name IN VARCHAR2 DEFAULT NULL, idx_vector_col IN VARCHAR2 DEFAULT NULL, idx_include_cols IN VARCHAR2 DEFAULT NULL, idx_partitioning_scheme IN VARCHAR2 DEFAULT NULL, idx_organization IN VARCHAR2 DEFAULT NULL, idx_distance_metric IN VARCHAR2 DEFAULT NULL, idx_accuracy IN NUMBER DEFAULT NULL, idx_parameters IN CLOB DEFAULT NULL, idx_parallel_creation IN NUMBER DEFAULT NULL);

TBD

RERANK (new 23ai)

Reorders search results for more relevant output dbms_vector.rerank( query IN CLOB, documents IN JSON, params IN JSON DEFAULT NULL) RETURN JSON;

TBD

SET_SGA_PERCENTAGE (new 23ai)

setS the desired max vector memory size for the PDB as a percentage of SGA (0 - 70%) dbms_vector.set_sga_percentage(sga_percent IN NUMBER);

exec dbms_vector.set_sga_percentage(5);

      

      PL/SQL procedure successfully completed.

SET_TRACE (new 23ai)

Undocumented dbms_vector.set_trace(status IN BOOLEAN);

exec dbms_vector.set_trace(TRUE);

TOGGLE_DEBUG_MODE (new 23ai)

Toggles debug mode ON or OFF dbms_vector.toggle_debug_mode(turn_on IN BOOLEAN);

exec dbms_vector.toggle_debug_mode(TRUE);

      

      PL/SQL procedure successfully completed.

      

      exec dbms_vector.toggle_debug_mode(FALSE);

      

      PL/SQL procedure successfully completed.

TOGGLE_INDEX_VECTOR_MEMORY_ADVISOR_VERIFY (new 23ai)

Undocumented dbms_vector.toggle_index_memory_advisor_verify(turn_on IN BOOLEAN);

exec dbms_vector.toggle_index_memory_advisor_verify(TRUE);

      

      PL/SQL procedure successfully completed.

      

      exec dbms_vector.toggle_index_memory_advisor_verify(TRUE);

      

      PL/SQL procedure successfully completed.

UTL_TO_CHUNKS (new 23ai)

Convert plain text into pieces of smaller text along with metadata

Overload 1 dbms_vector.utl_to_chunks( data IN CLOB, params IN JSON DEFAULT NULL) RETURN sys.vector_array_t;

TBD

Overload 2 dbms_vector.utl_to_chunks( data IN VARCHAR2, params IN JSON DEFAULT NULL) RETURN sys.vector_array_t;

TBD

UTL_TO_EMBEDDING (new 23ai)

Undocumented dbms_vector.url_to_embedding( data IN CLOB, params IN JSON DEFAULT NULL) RETURN VECTOR;

TBD

Overload 2 dbms_vector.utl_to_embedding( data IN BLOB, modeality IN VARCHAR2 DEFAULT 'image', params IN JSON DEFAULT NULL) RETURN VECTOR;

TBD

UTL_TO_EMBEDDINGS (new 23ai)

Converts text into embeddings (VECTORS) by calling an ONNX model or making an OCIGenAI REST call

Overload 1 dbms_vector.url_to_embeddings( data IN CLOB, params IN JSON DEFAULT NULL) RETURN vector_array_t;

TBD

Overload 2 dbms_vector.url_to_embeddings( data IN sys.vectoray_array_t, params IN JSON DEFAULT NULL) RETURN vector_array_t;

TBD

UTL_TO_GENERATE_TEXT (new 23ai)

Generate text by making a call a REST API call to OCIGenAI

Overload 1 dbms_vector.utl_to_generate_text( data IN CLOB, params IN JSON DEFAULT NULL) RETURN CLOB;

TBD

Overload 2 dbms_vector.utl_to_generate_text( text_data IN CLOB, media_data IN BLOB, media_type IN VARCHAR2, params IN JSON DEFAULT NULL) RETURN CLOB;

TBD

UTL_TO_RERANK (new 23ai)

Undocumented dbms_vector.utl_to_rerank( query IN CLOB, documents IN JSON, params IN JSON DEFAULT NULL) RETURN JSON;

TBD