What’s new in APEX 26.1

Along with the long-awaited APEXlang, APEX 26.1 was launched last week and it is a big update; AI integrations, item and region improvements, a new layout, and a shiny new theme have all been delivered. 

Along with the new features of APEXlang there is a lot to get to grips with inside and outside the application builder. Here is a brief summary of some of the parts we have been most excited about, and what they will mean to our ways of working.

APEXlang

Not to be confused with APEX_LANG, APEXlang is Oracle’s new file based format for APEX Applications. Until now, exporting an app gave you a pile of complicated and hard-to-read files with long strings of static IDs that made editing files locally or using git cumbersome and awkward. 

With APEX 26.1 when you export an application you have a new option to export in APEXlang. This gives you your application in a whole new structure that is human-readable, doesn’t contain all those pesky IDs, and can be edited locally in your IDE of choice. It gives you the application in a package of structured .apx text files that can be exported, imported, stored in source control, diffed, merged, validated, scanned, and reviewed.

The code is somewhere between YAML and Javascript, without quite being either. At a glance it is much easier to read and understand, and you can immediately do some useful stuff like edit your SQL for reports, update page help and items, and finally do a proper diff of two versions of the same page and see what has actually changed.

Of course, they haven’t made such a major change just to help out all us developers. This new format is also much more readable for AI, making all the new integrations in APEX work much better and have higher quality suggestions. You can also leverage your own AI deployments and APIs more effectively; the agents can now browse your local files and make more accurate changes, and even create all new applications from scratch.

All the changes APEXlang brings are too much to go into in this blog post. We have been experimenting with it internally and right off the bat it is already working much more smoothly with AI and version control, which is a huge improvement. To look further for yourself you can read some of the Oracle Blogs and documentation here:

https://www.oracle.com/apex/ai-application-generator/

For upgrading your AI agents to work with APEXlang you can deploy the Oracle Skills for AI Agents that you can find on their github:

https://github.com/oracle/skills/tree/main

AI Agents

Another big change that Oracle is promoting for the new APEX version is the upgrade of the ‘AI Assistant’ from APEX 24 to the ‘AI Agents’ in APEX 26.

As their image shows, previously the Assistant was simply able to make recommendations based on your data, whereas now the Agent can be implemented directly into your application workflows and make informed decisions based on the wider application context, even if the prompts are vague or ambiguous.

Of course, the strength of the suggestions still depends mainly on the quality of the model attached to your Workspace, but the availability of information to the model, along with its power to take action, and suggest changes has had a significant upgrade from the previous version. You can also apply guardrails to these actions to ensure there is adequate human oversight where it is required.

Extended Model Support

Another big win for those of us already leveraging AI within APEX; previously in version 24 the list of providers was pretty limited when connecting a generative AI service via API:

This has now been increased, allowing developers much more scope to use their models of choice, including their own localised models:

This is great, especially if you’ve already got accounts or API subscriptions with one of the newly added services like Anthropic or Gemini, or have taken the leap to deploy and manage your own LLM. It also gives developers more scope to experiment with finding the right models for the right jobs.

AI Interactive Reports

If like us you were watching the APEX Special Update on the release date, you’ll know they spent quite a bit of time on their integration of AI into Interactive Reports:

This is one area that certainly has a lot of potential for end-users to save time and retrieve more accurate information using natural language. However, something to note that we found in our initial testing is that once it is enabled it replaces the default search bar, meaning that potentially every search your end-users makes on the report will be consuming your precious tokens, even if the user is searching for a specific person's name, or a product code, where it could be more efficient to still be using the standard style search. 

Another factor is the time to return a result. In the example above, if I ask the model “Where did Bob go on holiday” it will think for a couple of seconds before giving me an answer. Not a huge delay, but given there are only 3 rows in the result set I was expecting it to be a bit more efficient. To be completely fair the use-case for this is when there is much more data across multiple pages of pagination that users are regularly filtering or pivoting, where AI Interactive Reports will really improve efficiency. It may also be related to the strength of the model I was using for testing the API calls, which was the Google free-tier of Gemini, which is not the fastest.

Data Reporter

The other major feature given a lot of attention in the Special Update demo was the new Data Reporter, and perhaps it is best to let Oracle’s own documentation explain its purpose:

“Data Reporter is a fully self-service module which enables business users to quickly and easily build applications and reports with little to no developer expertise. With Data Reporter, an administrator creates and manages datasets built on tables available in the APEX workspace. Business users can then use these datasets to quickly build reporting applications containing powerful reports without any knowledge of the APEX development environment.”

Introduction to Data Reporter

So at a high level it's a sort of sandboxed tool for allowing certain users a much greater level of customisation and reporting capabilities on restricted data sets. I think the usefulness of this new area will be quite dependant on the type of users and data, but for those that are regularly exporting large amounts of data into Tableau and PowerBI this might provide some functionality to keep them within the APEX environment reducing the need for context switching in the middle of tasks.

APEX Enhancements

Along with these ‘big ticket’ upgrade items, there is a whole host of smaller additions, upgrades and improvements in 26.1. 

Some of the ones we are most excited for are;

• Copy and Paste for Interactive Grid (including from external sources like Excel)

• Dynamic Quick Picks - no more using loads of Javascript to dynamically set dates

• The Document Generator function now supports creating password-protected PDFs

• Natural language AI Generation can be used for creating new pages

• Generative AI Token Usage Limits can be set granularly

• CSS Classes can now be used for Interactive Report Columns

Along with these (and many more) smaller enhancements, Oracle have also implemented a lot of user suggestions from the Ideas App. 

Some of these have already been mentioned above, and there is a lot of great stuff in there. It is always great to see user and developer suggestions being not only acknowledged, but also implemented in new versions. 

You can view the whole list of implemented suggestions here.

So there you have it, that’s just a quick overview of some of the key features we are excited about, and have had the most noise around them. For the full list of changes, deprecated items and documentation, you can view the full 26.1 release notes here.