Oracle Cloud Infrastructure Documentation

SQL Tool Issues

Troubleshoot common issues while working with the SQL tool in Generative AI Agents.

Ensure that you review also the Known Issues for Generative AI Agents, the SQL Tool Guidelines for Generative AI Agents, and completed any tasks that are required to use a specific function in an agent with a SQL tool.

SQL Tool Fails to Create

Creating a SQL tool in the Console appears to be successful as Generative AI Agents doesn't display any initial error messages. However, the agent is successfully created but the SQL tool in the agent fails to create. In the CREATE_TOOL work request, an error message that's similar to the following is shown:

Create Tool for ocid1.genaiagenttool.oc1.us-chicago-1.<string> failed due to sql errors: Statement:null[Error:net.sf.jsqlparser.parser.ParseException: Encountered unexpected <something>; at line <number>, column <number>. <more details>

The most common cause is an invalid database schema.

Delete the failed SQL tool and add a new tool.

  1. In the Tools list of the agent, find the failed SQL tool. If you need help finding the list, see Listing the SQL Tools in Generative AI Agents.
  2. From the Actions menu (Actions Menu) of the tool creation that failed, select Delete.
  3. Select Create tool to add a new SQL tool.

    For a valid schema description, see Database Schema.

    You can also copy the DDL of the schema tables, as described in the following optional procedure.

(Optional) Use these steps to get a valid schema from your database.

  1. In the Console, connect to a database and open SQL Worksheet.
  2. In the Navigator, find the schema and tables from which you want to copy the DDL.
  3. Right-click a table and select Quick DDL. Then select Save to File. Repeat for each table.
  4. In a new text file, create the database schema by copying and appending the text from each downloaded file. Modify the DDL in the complete DDL file as needed.
  5. In Generative AI Agents, find your agent. Add a SQL tool by copying and pasting the content of the complete DDL file into the Inline field to import the database schema.

SQL Execution Output File is Missing

A SQL tool is enabled to run the generated SQL queries and store the output result in Object Storage if there are more than 100 rows in the output. Chatting in the agent endpoint appears to be successful, but the output result file isn't saved in Object Storage. An error message that's similar to the following is shown:

Unable to create PreAuthenticated Request url for the result location.

Possible causes are described in the following sections.

Invalid or missing bucket

Select a valid Object Storage bucket in your preferred compartment.

  1. On the agent details page, under Resources, select Endpoints.
  2. Select the endpoint that's enabled for SQL execution.
  3. In the Edit endpoint panel, ensure that Enable storage of large output results is selected.
  4. In the Output details section, select the compartment and bucket in which you want to save output results that are larger than 100 rows.
  5. Select Save changes.

Missing policies

Review the IAM Policies to ensure that you have set up the correct permissions to access Object Storage.

If not already created, add the dynamic group and matching rule. Then add the policies as described for Object Storage, SQL execution, and output storage.