Skip to main content

JAICP updates in May 2025

New

  • AI agents in J‑Graph.
  • Knowledge base in AI agents.
  • Configure barge-in for LLM in telephony.
  • Support for the raw reply type in edna.chatCenter.
  • Roles for Yandex v3 voices.

Improved

  • LLM temperature in phone channel.
  • Logging update in ECMAScript 6.

Docs

  • Patterns in NLU test widget.
  • Aimylogic bot with Jay Knowledge Hub.

AI agents in J‑Graph​

We got a new block in J‑Graph: AI agent. This block lets you create and configure AI agents in your scripts. Agents use LLMs to engage in dialogs and handle any user requests.

AI agent block

You can specify general settings, pass dialog history to the agent, and configure logic of transitions to other states.

Knowledge base in AI agents​

Now you can use a RAG knowledge base in AI agents. An agent gets small fragments of text (chunks) from the base for each user request and takes them into account when generating a response.

info

To access a RAG knowledge base, contact your account manager or email us at support@just-ai.com.

To get started:

  1. Prepare a knowledge base in Jay Knowledge Hub.
  2. Connect the knowledge base in JAICP.
  3. Connect the knowledge base to an AI agent using one of the options below:

Configure barge-in for LLM in telephony​

A user can now interrupt the bot if you use llmRequest.

The LLM in telephony article demonstrates how to configure simple barge-in and conditional barge-in.

caution

Contextual barge-in is not supported for llmRequest.

raw reply type in edna.chatCenter​

The raw reply type is now available in the edna.chatCenter channel. You can use it to send parameters supported by edna.

Roles for Yandex v3 voices​

A role is a characteristic of the voice. For example, the speaker can sound friendlier or whisper.

If you use Yandex TTS v3, you can now specify roles for voices:

tip

See the available roles for different voices in the Yandex documentation.

LLM temperature in phone channel​

You can now specify the temperature property in the llmRequest response. It lets you adjust the creativity of LLM responses.

Logging update in ECMAScript 6​

We added new features for logging in the ECMAScript 6 runtime:

  • You can now call log inside asynchronous functions, messages are displayed in server log.
  • You can also specify different logging levels for messages.

Patterns in NLU test widget​

We updated the article about the NLU test widget and added information about working with patterns. If a pattern was triggered:

  • is displayed next to the intent.
  • Click to view the triggered pattern.

Aimylogic bot with Jay Knowledge Hub​

info

Jay Knowledge Hub is an intelligent question-answering system by Just AI. It uses large language models (LLMs) and the RAG technology. This system finds the required information in a set of documents and generates answers for users. For more information about Jay Knowledge Hub, see the website and documentation.

We added a tutorial on how to access Jay Knowledge Hub in Aimylogic scripts. The HTTP request block is used to send requests to the knowledge base.