Integrating APEX with OCI AI Services

Discover how Oracle APEX leverages OCI AI services to build smarter, more efficient applications. Hosts Lois Houston and Nikita Abraham interview APEX experts Chaitanya Koratamaddi, Apoorva Srinivas, and Toufiq Mohammed about how key services like OCI Vision, Oracle Digital Assistant, and Document Understanding integrate with Oracle APEX.

 

Packed with real-world examples, this episode highlights all the ways you can enhance your APEX apps.

 

Oracle APEX: Empowering Low Code Apps with AI: https://mylearn.oracle.com/ou/course/oracle-apex-empowering-low-code-apps-with-ai/146047/

Oracle University Learning Community: https://education.oracle.com/ou-community

LinkedIn: https://www.linkedin.com/showcase/oracle-university/

X: https://x.com/Oracle_Edu

 

Special thanks to Arijit Ghosh, David Wright, Kris-Ann Nansen, Radhika Banka, and the OU Studio Team for helping us create this episode.

 

---------------------------------------------------------------

 

Episode Transcript:

00:00

Welcome to the Oracle University Podcast, the first stop on your cloud journey. During this series of informative podcasts, we’ll bring you foundational training on the most popular Oracle technologies. Let’s get started!

00:25

Lois: Hello and welcome to the Oracle University Podcast. I’m Lois Houston, Director of Innovation Programs with Oracle University, and with me is Nikita Abraham, Team Lead: Editorial Services.

Nikita: Hi everyone! Last week, we looked at how generative AI powers Oracle APEX and in today’s episode, we’re going to focus on integrating APEX with OCI AI Services.

Lois: That’s right, Niki. We’re going to look at how you can use Oracle AI services like OCI Vision, Oracle Digital Assistant, Document Understanding, OCI Generative AI, and more to enhance your APEX apps.

01:03

Nikita: And to help us with it all, we’ve got three amazing experts with us, Chaitanya Koratamaddi, Director of Product Management at Oracle, and senior product managers, Apoorva Srinivas and Toufiq Mohammed. In today’s episode, we’ll go through each Oracle AI service and look at how it interacts with APEX. Apoorva, let’s start with you. Can you explain what the OCI Vision service is?

Apoorva: Oracle Cloud Infrastructure Vision is a serverless multi-tenant service accessible using the console or REST APIs. You can upload images to detect and classify objects in them. With prebuilt models available, developers can quickly build image recognition into their applications without machine learning expertise.

OCI Vision service provides a fully managed model infrastructure. With complete integration with OCI Data Labeling, you can build custom models easily. OCI Vision service provides pretrained models-- Image Classification, Object Detection, Face Detection, and Text Recognition. You can build custom models for Image Classification and Object Detection.

02:24

Lois: Ok. What about its use cases? How can OCI Vision make APEX apps more powerful?

Apoorva: Using OCI Vision, you can make images and videos discoverable and searchable in your APEX app. 

You can use OCI Vision to detect and classify objects in the images. OCI Vision also highlights the objects using a red rectangular box. This comes in handy in use cases such as detecting vehicles that have violated the rules in traffic images.

You can use OCI Vision to identify visual anomalies in your data. This is a very popular use case where you can detect anomalies in cancer X-ray images to detect cancer. These are some of the most popular use cases of using OCI Vision with your APEX app. But the possibilities are endless and you can use OCI Vision for any of your image analysis.

03:29

Nikita: Let’s shift gears to Oracle Digital Assistant. Chaitanya, can you tell us what it’s all about?

Chaitanya: Oracle Digital Assistant is a low-code conversational AI platform that allows businesses to build and deploy AI assistants. It provides natural language understanding, automatic speech recognition, and text-to-speech capabilities to enable human-like interactions with customers and employees. Oracle Digital Assistant comes with prebuilt templates for you to get started. 

04:00

Lois: What are its key features and benefits, Chaitanya? How does it enhance the user experience?

Chaitanya: Oracle Digital Assistant provides conversational AI capabilities that include generative AI features, natural language understanding and ML, AI-powered voice, and analytics and insights.

Integration with enterprise applications become easier with unified conversational experience, prebuilt chatbots for Oracle Cloud applications, and chatbot architecture frameworks. Oracle Digital Assistant provides advanced conversational design tools, conversational designer, dialogue and domain trainer, and native multilingual support. Oracle Digital Assistant is open, scalable, and secure. It provides multi-channel support, automated bot-to-agent transfer, and integrated authentication profile.

04:56

Nikita: And what about the architecture? What happens at the back end?

Chaitanya: Developers assemble digital assistants from one or more skills. Skills can be based on prebuilt skills provided by Oracle or third parties, custom developed, or based on one of the many skill templates available.

05:16

Lois: Chaitanya, what exactly are “skills” within the Oracle Digital Assistant framework? 

Chaitanya: Skills are individual chatbots that are designed to interact with users and fulfill specific type of tasks. Each skill helps a user complete a task through a combination of text messages and simple UI elements like select list.

When a user request is submitted through a channel, the Digital Assistant routes the user's request to the most appropriate skill to satisfy the user's request.

Skills can combine multilingual NLP deep learning engine, a powerful dialogflow engine, and integration components to connect to back-end systems. 

Skills provide a modular way to build your chatbot functionality. Now users connect with a chatbot through channels such as Facebook, Microsoft Teams, or in our case, Oracle APEX chatbot, which is embedded into an APEX application.

06:21

Nikita: That’s fascinating. So, what are some use cases of Oracle Digital Assistant in APEX apps?

Chaitanya: Digital assistants streamline approval processes by collecting information, routing requests, and providing status updates.

Digital assistants offer instant access to information and documentation, answering common questions and guiding users. Digital assistants assist sales teams by automating tasks, responding to inquiries, and guiding prospects through the sales funnel. Digital assistants facilitate procurement by managing orders, tracking deliveries, and handling supplier communication. Digital assistants simplify expense approvals by collecting reports, validating receipts, and routing them for managerial approval. Digital assistants manage inventory by tracking stock levels, reordering supplies, and providing real-time inventory updates. Digital assistants have become a common UX feature in any enterprise application.

07:28

Want to learn how to design stunning, responsive enterprise applications directly from your browser with minimal coding?

The new Oracle APEX Developer Professional learning path and certification enables you to leverage AI-assisted development, including generative AI and Database 23ai, to build secure, scalable web and mobile applications with advanced AI-powered features.

From now through May 15, 2025, we’re waiving the certification exam fee (valued at $245).

So, what are you waiting for? Visit mylearn.oracle.com to get started today.

08:09

Nikita: Welcome back! Thanks for that, Chaitanya. Toufiq, let’s talk about the OCI Document Understanding service. What is it?

Toufiq: Using this service, you can upload documents to extract text, tables, and other key data. This means the service can automatically identify and extract relevant information from various types of documents, such as invoices, receipts, contracts, etc. The service is serverless and multitenant, which means you don't need to manage any servers or infrastructure. You can access this service using the console, REST APIs, SDK, or CLI, giving you multiple ways to integrate.

08:55

Nikita: What do we use for APEX apps? 

Toufiq: For APEX applications, we will be using REST APIs to integrate the service. Additionally, you can process individual files or batches of documents using the ProcessorJob API endpoint. This flexibility allows you to handle different volumes of documents efficiently, whether you need to process a single document or thousands at once. With these capabilities, the OCI Document Understanding service can significantly streamline your document processing tasks, saving time and reducing the potential for manual errors.

09:36

Lois: Ok.  What are the different types of models available? How do they cater to various business needs?

Toufiq: Let us start with pre-trained models. These are ready-to-use models that come right out of the box, offering a range of functionalities. The available models are Optical Character Recognition (OCR) enables the service to extract text from documents, allowing you to digitize, scan the documents effortlessly.

You can precisely extract text content from documents. Key-value extraction, useful in streamlining tasks like invoice processing. Table extraction can intelligently extract tabular data from documents. Document classification automatically categorizes documents based on their content. OCR PDF enables seamless extraction of text from PDF files.

Now, what if your business needs go beyond these pre-trained models. That's where custom models come into play. You have the flexibility to train and build your own models on top of these foundational pre-trained models. Models available for training are key value extraction and document classification.

10:50

Nikita: What does the architecture look like for OCI Document Understanding?

Toufiq: You can ingest or supply the input file in two different ways. You can upload the file to an OCI Object Storage location. And in your request, you can point the Document Understanding service to pick the file from this Object Storage location. 

Alternatively, you can upload a file directly from your computer. Once the file is uploaded, the Document Understanding service can process the file and extract key information using the pre-trained models.

You can also customize models to tailor the extraction to your data or use case. After processing the file, the Document Understanding service stores the results in JSON format in the Object Storage output bucket. Your Oracle APEX application can then read the JSON file from the Object Storage output location, parse the JSON, and store useful information at local table or display it on the screen to the end user.

11:52

Lois: And what about use cases? How are various industries using this service?

Toufiq: In financial services, you can utilize Document Understanding to extract data from financial statements, classify and categorize transactions, identify and extract payment details, streamline tax document management.

Under manufacturing, you can perform text extraction from shipping labels and bill of lading documents, extract data from production reports, identify and extract vendor details.

In the healthcare industry, you can automatically process medical claims, extract patient information from forms, classify and categorize medical records, identify and extract diagnostic codes. This is not an exhaustive list, but provides insights into some industry-specific use cases for Document Understanding.

12:50

Nikita: Toufiq, let’s switch to the big topic everyone’s excited about—the OCI Generative AI Service. What exactly is it?

Toufiq: OCI Generative AI is a fully managed service that provides a set of state of the art, customizable large language models that cover a wide range of use cases. It provides enterprise grade generative AI with data governance and security, which means only you have access to your data and custom-trained models.

OCI Generative AI provides pre-trained out-of-the-box LLMs for text generation, summarization, and text embedding. OCI Generative AI also provides necessary tools and infrastructure to define models with your own business knowledge.

13:37

Lois: Generally speaking, how is OCI Generative AI useful? 

Toufiq: It supports various large language models. New models available from Meta and Cohere include Llama2 developed by Meta, and Cohere's Command model, their flagship text generation model. Additionally, Cohere offers the Summarize model, which provides high-quality summaries, accurately capturing essential information from documents, and the Embed model, converting text to vector embeddings representation.

OCI Generative AI also offers dedicated AI clusters, enabling you to host foundational models on private GPUs. It integrates LangChain and open-source framework for developing new interfaces for generative AI applications powered by language models.

Moreover, OCI Generative AI facilitates generative AI operations, providing content moderation controls, zero downtime endpoint model swaps, and endpoint deactivation and activation capabilities. For each model endpoint, OCI Generative AI captures a series of analytics, including call statistics, tokens processed, and error counts.

14:58

Nikita: What about the architecture? How does it handle user input?

Toufiq: Users can input natural language, input/output examples, and instructions. The LLM analyzes the text and can generate, summarize, transform, extract information, or classify text according to the user's request. The response is sent back to the user in the specified format, which can include raw text or formatting like bullets and numbering, etc.

15:30

Lois: Can you share some practical use cases for generative AI in APEX apps? 

Toufiq: Some of the OCI generative AI use cases for your Oracle APEX apps include text summarization. Generative AI can quickly summarize lengthy documents such as articles, transcripts, doctor's notes, and internal documents. Businesses can utilize generative AI to draft marketing copy, emails, blog posts, and product descriptions efficiently. Generative AI-powered chatbots are capable of brainstorming, problem solving, and answering questions.

With generative AI, content can be rewritten in different styles or languages. This is particularly useful for localization efforts and catering to diverse audience.

Generative AI can classify intent in customer chat logs, support tickets, and more. This helps businesses understand customer needs better and provide tailored responses and solutions. By searching call transcripts, internal knowledge sources, Generative AI enables businesses to efficiently answer user queries. This enhances information retrieval and decision-making processes.

16:47

Lois: Before we let you go, can you explain what Select AI is? How is it different from the other AI services?

Toufiq: Select AI is a feature of Autonomous Database. This is where Select AI differs from the other AI services. Be it OCI Vision, Document Understanding, or OCI Generative AI, these are all freely managed standalone services on Oracle Cloud, accessible via REST APIs. Whereas Select AI is a feature available in Autonomous Database. That means to use Select AI, you need Autonomous Database. 

17:26

Nikita: And what can developers do with Select AI?

Toufiq: Traditionally, SQL is the language used to query the data in the database. With Select AI, you can talk to the database and get insights from the data in the database using human language.

At the very basic, what Select AI does is it generates SQL queries using natural language, like an NL2SQL capability. 

17:52

Nikita: How does it actually do that?

Toufiq: When a user asks a question, the first step Select AI does is look into the AI profile, which you, as a developer, define.

The AI profile holds crucial information, such as table names, the LLM provider, and the credentials needed to authenticate with the LLM service. Next, Select AI constructs a prompt. This prompt includes information from the AI profile and the user's question. 
Essentially, it's a packet of information containing everything the LLM service needs to generate SQL.

The next step is generating SQL using LLM. The prompt prepared by Select AI is sent to the available LLM services via REST. Which LLM to use is configured in the AI profile. The supported providers are OpenAI, Cohere, Azure OpenAI, and OCI Generative AI.

Once the SQL is generated by the LLM service, it is returned to the application. The app can then handle the SQL query in various ways, such as displaying the SQL results in a report format or as charts, etc. 

19:05

Lois: This has been an incredible discussion! Thank you, Chaitanya, Apoorva, and Toufiq, for walking us through all of these amazing AI tools. If you’re ready to dive deeper, visit mylearn.oracle.com and search for the Oracle APEX: Empowering Low Code Apps with AI course. You’ll find step-by-step guides and demos for everything we covered today. 

Nikita: Until next week, this is Nikita Abraham…

Lois: And Lois Houston signing off!

19:31

That’s all for this episode of the Oracle University Podcast. If you enjoyed listening, please click Subscribe to get all the latest episodes. We’d also love it if you would take a moment to rate and review us on your podcast app. See you again on the next episode of the Oracle University Podcast.