Skip to main content

JAICP updates in October 2025

Improved

  • List of pause fillers for LLM for the phone channel.
  • LLM parameters for the phone channel.

List of pause fillers for LLM for the phone channel​

Previously, you could specify only one pause filler for the llmRequest response.

To make interactions with the bot livelier and more diverse, you can now specify a list of phrases using fillerPhrasesList. Each time, the bot will choose a random phrase from this list.

LLM parameters for the phone channel​

For the llmRequest response type, you can pass an object with LLM parameters. The set of parameters, their allowed values, and how they affect the response depend on the specified model and version.

In particular, the max_completion_tokens parameter can be used to produce shorter responses from the LLM, which are preferred over the phone channel. Older GPT models may use max_tokens.