Oracle Digital Assistant is a leading conversational AI platform that has historically dealt with predictive AI. Predictive AI takes in a user’s message and categorizes it based on the data it has been trained with. From this categorization, the conversation can go down pre-defined workflows that give standard responses and connect to different data sources. With the advent of generative AI, new possibilities for robust conversations have opened up. Users can have dynamic conversations with a chatbot that do not feel prescriptive but still connect to the business’s knowledge source. ODA has developed some new features to allow for generative AI to exist alongside its existing predictive AI platform. These new features are Knowledge Dialogs and LLM Blocks.
Knowledge Dialogs
The Knowledge Dialog feature is built directly into ODA using an internal language model (LLM) and a vector database to execute advanced searches on indexed documents. One of its standout advantages is simplicity: there is no need for external integrations since you can feed documents directly into the application. This setup not only simplifies deployment but also enhances security by keeping all data on the same server as the chat application, eliminating the need for separate API keys.
However, as of June 2024, Knowledge Dialog is still in limited release and undergoing development. It currently lacks some advanced features such as user-specific content filtering, the ability to refine prompts based on user inputs, and customization of responses generated by the LLM. These are areas where ongoing updates are expected to enhance functionality.
LLM Blocks
For broader integrations with external generative AI services like Azure OpenAI, Cohere, or Oracle Generative Search, ODA offers LLM Blocks. These blocks act as a versatile connector, facilitating interactions with external LLMs through standard REST calls. LLM Blocks come with features such as message streaming, initial prompting, and output formatting, which make integration seamless and development straightforward. This feature can be called from anywhere in a conversation with any input parameters, which can allow for dynamic and accurate conversations that are relevant to the user’s specific needs and profile.
In essence, ODA’s accommodation for generative AI marks a significant advancement in enhancing user interactions. While Knowledge Dialog showcases ODA’s native capabilities in generative AI, LLM Blocks extend its reach by enabling connections with external AI services. These features demonstrate Oracle’s commitment to providing flexible and cutting-edge AI solutions that enhance digital assistant capabilities across various enterprise applications.
