To build a prompt generator from K-graph, add the graphs in the promptset builder and select the pattern from Actor, Asset, Action & Process. You can select multiple patterns from the graphs. You can select the elements from universal graphs, or graph groups or individual graphs.

When working with an AI model, one tends to start with a simple question. And gradually they add more conditions and context to the topic. To improve on the number of iterations required to achieve the desired results, you can build a contextful prompt with all the caveats of the topic derived from knowledge graph. This type of prompt has shown better results and achieves the desired goal in single attempt.

The assembly of graph elements to generate multiple prompts around a topic is called as promptset, in FastBuilder. The promptsets can be iterated over the graph to generate multiple sentences which hold true as ground truth. These sentences can be used as control or approach for querying an AI.

For large and complex topics, such as supply chain bottlenecks, one maynot know all that is to discover in the beginning. For such scenarios, one may go for exploratory prompts. It is simular to the famous "Five Why" problem solving technique. After each question, the graph traversal can generate the best next path to explore through meaningful questions.

A graph depicting the actor/action/asset triplets of a human worker like doctor, lawyer, accountant, etc is called a doer/role graph. You can upload a user manual or user stories from their usage applications etc to build such graphs. A series of prompts or promptsets, can be generated for scenarios to explain a software bot about how to perform a human task. Such techniques can be used for various task of task automation or task assistance. In such cases, a bot would know about the next logical step. Such bots can guide or war the human doer for deviations or questions, respectively