22 November, 2026

SELECT AI in the Oracle Autonomous Database

 How to configure and run SELECT AI in the Oracle Autonomous Database.


UPDATE : Video Demo of the usage of SELECT AI is at 

https://youtu.be/qhnBciuFC78


--- first setup Network ACL for the Database to connect to the API
--- in this case, I have configured an API at anthropic.com
--- you could be using an OCI AI Service  OR   OpenAI or other provider
BEGIN
    DBMS_NETWORK_ACL_ADMIN.APPEND_HOST_ACE(
         host => 'api.anthropic.com',
         ace  => xs$ace_type(privilege_list => xs$name_list('http'),
                             principal_name => 'HEMANT',
                             principal_type => xs_acl.ptype_db)
   );
END;
/

--- grants to the account that will be configuring the credential
GRANT EXECUTE ON UTL_HTTP TO HEMANT;

GRANT EXECUTE ON DBMS_CLOUD_AI TO HEMANT;

--- login as HEMANT

BEGIN
  DBMS_CLOUD.DROP_CREDENTIAL(
    credential_name => 'ANTHROPIC_CRED'
  );
END;
/

--- create the Credential to connect to the API
--- give it a name and label
--- specify the API Key as a password
--- verify with that the API Key is valid and has not expired
BEGIN
  DBMS_CLOUD.CREATE_CREDENTIAL(
    credential_name => 'ANTHROPIC_CRED',
    username        => 'ANTHROPIC_USER', -- User-defined label
    password        => 'MYAPIKEY'  -- example value "sk-ant-usr-15-ABcdefghijklmnopqrstuvwxYZ"
   );
END;
/

--- create a profile to map to the API Credential
--- if you use different LLMs and different API Keys, specify a different profile for each !!
BEGIN
  DBMS_CLOUD_AI.DROP_PROFILE(
    profile_name => 'MY_ANTHROPIC_PROFILE'
  );
END;
/

--- in the profile speciy which database schema and tables would be available to SELECT AI
BEGIN
  DBMS_CLOUD_AI.CREATE_PROFILE(
    profile_name => 'MY_ANTHROPIC_PROFILE',
    attributes   => '{
      "provider": "anthropic", 
      "credential_name": "ANTHROPIC_CRED",
      "model": "claude-sonnet-4-6",
      "object_list": [
         {"owner": "HEMANT", "name": "DEPARTMENTS"},
         {"owner": "HEMANT", "name": "EMPLOYEES"},
         {"owner": "HEMANT", "name": "LOCATIONS"}
      ]
    }',
    description  => 'AI profile for translating natural language to SQL'
  );
END;
/


------------------------------------------------------------------------------------------------
--- Here we start the user session by specifying the profile to be used
EXEC DBMS_CLOUD_AI.SET_PROFILE('MY_ANTHROPIC_PROFILE');

---  this test returns output in JSON format
SELECT DBMS_CLOUD_AI.GENERATE(
  prompt       => 'Name all the Deparments and the number of Employees in each Department where a department has more than 5 employees', 
  profile_name => 'MY_ANTHROPIC_PROFILE',
  action       => 'runsql' 
) FROM DUAL;


SELECT AI chat 'Name all the Deparments and the number of Employees in each Department where a department has more than 5 employees';

SELECT AI narrate 'Name all the Deparments and the number of Employees in each Department where a department has more than 5 employees';


SELECT AI chat 'Which location has the largest number of employees ?';

SELECT AI narrate 'Which location has the largest number of employees ?';


SELECT AI chat 'Which are the Top 3 Departments by Salary, list the Department names and total Salary in each';

SELECT AI narrate 'Which are the Top 3 Departments by Salary, list the Department names and total Salary in each';


SELECT AI chat  'how many employees are paid a commision ?' ;

SELECT AI narrate  'how many employees are paid a commision ?' ;

SELECT AI narrate  'what is the highest commission rate being paid ?' ;


The above shows how I configured my Autonomous Database to connect to an LLM and configured a profile for three specific tables to be used by SELECT AI

20 November, 2026

Using an LLM to generate Annotations and run Queries

 A video demonstration of using an LLM to

1. create "description" annotations on an Oracle Database Schema
2. query the database schema using Natural Language ("English")

The combination of tools used :
Oracle Database 26c (26.3)
Oracle SQL Developer 26.3 Extension in VSCode
Oracle sqlcl (installed with SQL Developer)
Oracle sqlcl MCP Server
GitHub Copilot
Claude Sonnet model
(Other than the Database all the tools running from within VSCode)