What is the source of of the Editor AI Assistant's knowledge?
An LLM.
Is the AI Assistant's knowledge tied to a specific LLM?
Yes, AWS Nova is used.
How do we know the AI Assistant's knowledge is up-to-date and relevant?
Because the LLM is one of AWS's foundation models. This does not use RAG, so responses do not leverage Kernels as source material.
What is the target use case?
This functionality is ideal for content creators (think Author or Editor roles) who want to increase their efficiency and improve content quality, whether on an internal-facing or public site.
Are we repurposing another application, or is this self-built?
The final solution is a combination of:
- The user-interface, designed by the editor infrastructure service.
- The LLM, a licensed AWS service.
- Predefined prompts: a combination of proprietary prompts that we defined and added, and commonly used prompts provided by the licensed editor service.
Could formatting content for GenSearch be a part of prompt engineering?
Yes. Using proprietary predefined prompts (shortcuts) within the Editor, authors can optimize content for GenSearch responses. These enhancements are included in the 25.2 NiCE release.
Content contributors can also engineer and define their own prompts to generate custom-tailored responses, using the ExpertAI within the Editor. These responses will be subject to token limitations on the LLM model.
