How does interrupt work in NYRA-Agent

Overview

The interrupt mechanism in the TEN-Agent framework is composed of two core components: Interrupt Detection and Interrupt Response. This document provides an in-depth explanation of both components and outlines how the interrupt command propagates through the AI agent graph.

Part 1: Interrupt Detection

1. Current Interrupt Detection Implementation

The current (interrupt_detector_python) extension implements a text-based interrupt detection mechanism:

def on_data(self, nyra: NyraEnv, data: Data) -> None:
    text = data.get_property_string(TEXT_DATA_TEXT_FIELD)
    final = data.get_property_bool(TEXT_DATA_FINAL_FIELD)

    # Trigger interrupt when text is final or reaches threshold length
    if final or len(text) >= 2:
        self.send_flush_cmd(nyra)

The interrupt detector is activated under the following conditions:

1. Upon receiving final text (is_final = true).

2. When the text length meets or exceeds the defined threshold (≥ 2 characters).

2. Customize Interrupt Detection

To create a custom interrupt detection logic, you can use the implementation of interrupt_detector_python as a reference, tailoring the interrupt conditions to align with your specific requirements.

Part 2: Interrupt Response

Chain Processing in AI Agent Graph

Within a conventional AI agent graph, the interrupt command (flush) is propagated through a structured sequential processing chain.

Interrupt Detector
       ↓
    LLM/ChatGPT
       ↓
      TTS
       ↓
   agora_rtc

Each extension within the processing chain executes two critical actions upon receiving a flush command:

1. Resource Clean up: Clears its own resources and resets its internal state.

2. Command Propagation: Forwards the flush command to downstream extensions.

This structured approach ensures:

• Extensions are cleaned and reset in the proper sequence.

• Residual data is effectively eliminated from the system.

• Each extension is restored to a pristine state, ready for subsequent operations.

Conclusion

Nyra's interrupt mechanism employs a structured chain-processing paradigm to enforce the systematic cleanup of all extensions within the AI agent graph. Upon an interrupt event, each extension sequentially performs two critical tasks: first, it meticulously cleans its internal state and resources; second, it propagates the flush command to the subsequent extensions in the chain. This rigorous approach guarantees a coherent and pristine system state, enabling seamless execution of subsequent operations.