The Vocode hosted API provides dials to tune the responsiveness and conversational mechanics of your agent. These parameters are as follows:

  • interrupt_sensitivity: How responsive the bot is to speech from the human
  • endpointing_sensitivity: Controls the sensitivity of detecting the end of the human’s speech
  • conversation_speed: Determines how quickly the bot responds for its turns in the conversation.

These parameters can be customized via the create and update agent APIs.

Interrupt Sensitivity

Vocode phone agents are designed to handle interruptions from the humans, and the interrupt_sensitivity parameter controls how sensitive the bot is to these interruptions.

Interrupt sensitivity has two options: low (default) and high. Low sensitivity makes the bot ignore backchannels (e.g. “sure”, “uh-huh”) while the bot is speaking. High sensitivity makes the agent treat any word from the human as an interruption.

In summary:

  • Possible values: low, high
  • Default value: low
  • low = ignore non-interrupting words
  • high = stop at any word

Endpointing Sensitivity

endpointing_sensitivity controls how the bot determines when the human is done speaking so it can kick off its response. It can be one of sensitive, relaxed, or auto (default and currently equivalent to relaxed).

A bot with endpointing_sensitivity == 'sensitive' treats the end of a sentence or a short pause as the end of the human’s turn. This works well for quick conversations in which you expect short, single-sentence responses from the human. A bot with this sensitivity runs the risk of interrupting the human if they continue speaking after a single sentence.

A bot with endpointing_sensitivity == 'relaxed' waits a short time after the end of a human’s sentence before considering their turn over. Alternatively, when a sentence is not finished, relaxed sensitivity waits for a longer pause before marking the end of the human’s speech. This sensitivity has less risk of interrupting the human but can potentially have a slower response time. Note that auto sensitivity is equivalent to relaxed.

Note: the amount of time the agent waits during a pause adapts to the human’s utterance speed. If the human speaks fast, the bot will wait less, and if the human speaks slow the bot will wait longer.

In summary:

  • Possible values: sensitive, relaxed, auto
  • Default value: auto
  • sensitive = end of first sentence or shorter pause
  • relaxed/auto = pause after end of first sentence or longer pause

Conversation Speed

conversation_speed controls the length of endpointing pauses, i.e. how long the bot will wait before responding to the human. This includes normal utterances from the human as well as interruptions.

The amount of time the bot waits inversely scales with the conversation_speed value. So a bot with conversation_speed of 2 responds in half the time compared to a conversation_speed of 1. Likewise a conversation_speed of 0.5 means the bot takes twice as long to respond.

In summary:

  • Possible values: between 0.5 and 2
  • Default value: 1
  • Response time inversely scales with this value