Abstract
Conversational assistant interfaces are reactive. No processing is performed until a session has been opened and a well-formed request has been typed into an open input field. Operational problems must therefore be found first on separate monitoring surfaces. Described here is an asynchronous technique by which unstructured backend state events are converted into pre-parameterized, executable natural language prompts. A remote procedure call is issued against a decoupled state database, and returned anomalies are mapped to category and priority constants. A fully formed prompt string is then compiled for each anomaly, with live record variables injected during compilation. Each compiled string is transmitted as a hidden payload bound to a graphical element. Interaction with that element submits the payload along a state-isolated execution path. An anomaly is thereby resolved without manual prompt composition, and without alteration of the memory held by an ongoing conversation.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Sharma, Jayant, "Proactive Generation and Interface-Binding of Parameterized Executable Prompts from Backend State Events", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/12064