One of the “marquee” features of the new Oracle APEX 26.1 release is the ability to add natural language queries to Interactive Reports (IR). So, as well as having full IR ability, the user can ask a question like “show me all employees in Sales earning more than 1500” as an alternative to manually setting filters and it’s APEX which then applies report settings such as filters, sorts, highlights, groups, control breaks, charts and pivots. The user can see the changes and can tweak those settings in the usual way.
In order to use the facility, there is some AI setup required as well as context setting at report, column and also reference data level. This blog explores this, shows the setup and some examples of the technology in use.
Let’s create an example to show this in action. Firstly we need to have “AI” setup
Turn On Generative AI Within APEX
Navigate to Workplace Utilities –> Generative AI
I setup a link to use an Open AI gpt-5.4-nano model (“Our cheapest GPT-5.4-class model for simple high-volume tasks“).
Generate a new Application
Now let’s create an application that has an IR in it that we can use as an example. As we’re going to be using a new AI feature, let’s go AI all the way and use AI to generate the initial app for us.
In our database we have installed the EMP and DEPT tables so we will ask the application AI assistant to create us a simple application to manage these tables just by providing the follow prompt.
We will accept the generated content as the first page that is created is the interactive report that we need. Note though that we could continue the dialog and add/change content via a dialog.
Once created, we can see all the artefacts.
In the generated application, go into the shared components and select the AI attributes from the Generative AI.
Select the AI we created.
Using the new AI Narrative Facility
Let’s run the application and see the IR as it is created by default.
The AI IR is not enabled by default, but in order to activate it we just need to toggle the Natural Language Support switch to enable the facility.
When we toggle on the Natural Language Support, we can fill in the Report Context to give the AI service enough business context to more reliably map user’s narrative to report settings ( filters, etc).
The IR now has a new AI bar added where we can make narrative suggestions to filter the report rather than having to manually make those changes.
We can enter the following narrative suggestion to highlight the employees with a high commission.
The AI interprets the narrative request and performs the appropriate highlighting.
As well as simply filtering and highlighting, we can ask some more complex question, such as grouping.
As well as the narrative bar, we can also invoke the AI assistant to have a conversation. Here we will start by asking to restrict the employees to those where KING is their manager and then also order the results by a specific metric too.
At column level we can add Column Context which gives the AI a better understanding of what each column means. By default there’s no values and so we will need to set it to an appropriate value.
For example, let’s set some context for the Salary.
Let’s also give the commission some context too.
Also set the JOB column context
Also add some context to the MGR column to improve the context for this too.
As additional context we can also add Reference Data. This is the other section within the Generative AI block, so let’s add a LOV for JOBS so the AI can reference all the values for jobs without us actually listing them in the comments ( as this list can change over time ).
Take a look at the help for the field to see more information on how to do that for SQL.
So for JOBS we can just write this SQL and validate it.
Now we can ask for all employees who have the role of Manager to be highlighted in a specific colour.
As we desired, this filter is applied immediately.
What about AI Annotations in the 26ai database?
In my database currently, I have NOT performed any AI annotations on these tables as we can see here using the SQL Developer plug-in for VS Code
These are useful for AI features such as Select AI, where database metadata can help the model understand table and column meaning.
For APEX 26.1 AI Interactive Reports, I would still configure the APEX built in Report Context, Column Context and Reference Data as a database column can be used in multiple reports, and each report may have a different user purpose. The metadata at the database should provide a description of the what the data means and the IR should describe how the report want to see/use it.
Summary
AI Interactive Reports are an augmentation of the existing IR, and provide a natural language interface for users to configure reports they already have, or to create/tweak new ones.
After some experimentation this works well when the report is well designed, the purpose of the columns are clear, and the AI context is specific enough to reduce ambiguity. In the example I used above I did not write a lot of context nor did I add that every column or add in LOVs for metadata lookup, but for the ones that I did then this is already giving the results I was expecting.