Our longstanding practice of creating comprehensive documentation for all aspects of our business has been a cornerstone of our success. This practice has proven invaluable in educating and assisting users who seek help through our dedicated Slack channel, specifically designed to support our OAC (Oracle Analytics Cloud) community. However, we have become victims of our own documentation success and I must acknowledge that searching through numerous documents for specific information has become an increasingly challenging task.
To streamline our processes and enhance productivity we partnered with the Oracle Generative AI Agents Product Development Team and the Oracle EMEA Generative AI Agents Specialists Team, and we have recently introduced a cutting-edge Gen AI-powered chat assistant tailored for our extensive OAC user community.
This innovative assistant is designed to revolutionize the user experience by providing prompt and reliable responses to a wide array of internal OAC-related inquiries. As you read this, our chat assistant is actively transforming the way we operate by delivering quick answers to various OAC topics, including subject area specific knowledge, instance specific variables, simple calculations, user privileges, and much more. The assistant has been trained on a vast collection of documents, enabling it to address a diverse range of OAC queries.
So, let’s talk a little bit about the technical design. Our chat assistant is a sophisticated RAG (Retrieval Augmented Generation) application, meticulously crafted using our internal Confluence pages as its knowledge base. This technology leverages the capabilities of an LLM (Large Language Model) by augmenting its responses with relevant information retrieved from this dedicated knowledge base, ensuring accuracy and reliability.
The core components of our chat assistant’s architecture are as follows:
The Powerhouse aka Oracle Gen AI Agents
The engine of this vehicle is the new Oracle Gen AI Agents, where all the fun background action takes place. AI Agents provides RAG capabilities (no coding needed), so this is where we fed all our documentation pages, created a knowledge base, and a subsequent dedicated agent based on this knowledge. We have currently used one agent, although the way we have set up this assistant, we can in future, if need comes, create multiple dedicated agents with specific knowledge bases.
The Curator aka Oracle Digital Assistant
While Oracle Gen AI Agents does all the groundwork, all the communications are overseen by ODA (Oracle Digital Assistant). With the help of ODA, we control the dialog flow of all conversations, so our agent is given a call only when a user has entered a query in chat. From a custom welcome message to a fallback answer to enabling easy integration with apex as user interface, everything is being handled by ODA. Not only that, with many other capabilities like conversation insights, custom metrics, amongst others, ODA is truly a game changer.
The Face aka Oracle APEX
The final piece of the puzzle, as of now, is Oracle APEX where users can interact with the chat assistant. We have setup an Oracle APEX application which is serving as a convenient user interface where they can come in, fire queries, and get quick, reliable, and beautifully summarized answers. The application is easily accessible via mobile as well, quite handy when you are on the go or even in a meeting room without your laptop.
Our Chat Assistant in Action
Never a dull a moment in the realm of AI, right?
- As next steps, we are looking to integrate our chat assistant into our dedicated slack channel.
- Other than this, we would love our chat assistant to also address general queries about OAC, and for this, we plan to bring in the publicly available information into our assistant, of course not literally, but by directly leveraging a Large Language Model (LLM) instead. Of course, we’ll need to tame the beast, so it gives answers only for OAC queries.
- Finally, to enhance the user experience further, we plan to do few other things like showing citations for each response, build explicit user feedback to help improve our assistant’s performance even further, and last but not the least, keep feeding relevant documentation into our agent to make sure it is learning something new (for us) every day.
Call to Action
It’s your turn now, go ahead and start building your own chat assistant.
If you don’t already have an OCI tenancy, sign up for the free tier at oracle.com/cloud/free.
Learn more here : OCI Generative AI Agents, Oracle Digital Assistant and Oracle APEX.