System

A system comprising:

• a communication interface including a network interface controller coupled to a packet-switched network and configured to transmit and receive data packets conforming to at least one of Ethernet and wireless communication protocols;
• a memory storing instructions, a generative neural network model comprising an embedding layer, a plurality of self-attention layers, a plurality of feed-forward layers, and an output projection layer, and an emotion classification model;
• and
• circuitry comprising one or more processors coupled to the memory and configured to execute the instructions to:
• receive, via the communication interface, a data packet from a terminal device coupled to the packet-switched network;
• perform natural language processing on payload data extracted from the data packet to generate a feature vector representing a semantic interpretation including at least a relationship label, an urgency label, and a classification label;
• construct a prompt data structure encoding the feature vector and contextual parameters extracted from the data packet, and transmit the prompt data structure to the generative neural network model via an inference interface;
• receive, from the generative neural network model, inference output data including an evaluation score and at least one query data structure, and parse the inference output data using a deterministic parser to extract structured fields;
• transmit a query message via the communication interface to at least one external storage device coupled to the packet-switched network storing reference information including at least one of flagged identifiers and anomalous pattern records, and receive a comparison result;
• compute a composite score by applying a weighted combination of the evaluation score, the comparison result, and a deviation metric derived from historical data stored in a database coupled to the packet-switched network;
• when the composite score satisfies a predetermined threshold, transmit, via the communication interface, a control message including a notification payload and the at least one query data structure to the terminal device, and set a state flag that restricts execution of a pending operation at the terminal device;
• receive, via the communication interface, a response message from the terminal device, construct an updated prompt data structure including data from the response message, transmit the updated prompt data structure to the generative neural network model to obtain updated inference output data, and recompute the composite score based on the updated inference output data;
• and
• write, to a log storage coupled to the packet-switched network, a record associating the prompt data structure, the inference output data, the comparison result, the control message, and a final operation outcome with a session identifier.

A method performed by circuitry of a server coupled to a packet-switched network via a communication interface, the method comprising:

• receiving, via the communication interface, a data packet from a terminal device coupled to the packet-switched network;
• performing natural language processing on payload data extracted from the data packet to generate a feature vector representing a semantic interpretation including at least a relationship label, an urgency label, and a classification label;
• constructing a prompt data structure encoding the feature vector and transmitting the prompt data structure to a generative neural network model via an inference interface;
• receiving, from the generative neural network model, inference output data including an evaluation score and at least one query data structure;
• transmitting a query message via the communication interface to at least one external storage device coupled to the packet-switched network, and receiving a comparison result indicating whether the payload data corresponds to a pattern that deviates from a baseline pattern stored in the at least one external storage device;
• computing a composite score based on the evaluation score, the comparison result, and historical data retrieved from a database coupled to the packet-switched network;
• when the composite score satisfies a predetermined threshold, transmitting, via the communication interface, a control message including a notification payload and the at least one query data structure to the terminal device, and setting a state flag that restricts execution of a pending operation at the terminal device;
• receiving, via the communication interface, a response message from the terminal device, constructing an updated prompt data structure including data from the response message, and transmitting the updated prompt data structure to the generative neural network model to obtain updated inference output data;
• and
• writing, to a log storage coupled to the packet-switched network, a record associating the prompt data structure, the inference output data, the comparison result, and the control message with a session identifier.