JAICP updates in March 2025
New
- LLM in telephony.
- Get history in the LLM format.
- Deploy bots to Mattermost.
Improved
$gpt.createChatCompletion
method update.
Docs
- Fields in
$nlp.match
response.
LLM in telephony
Beta
The llmRequest
reply type has become available in JAICP.
It lets you use LLMs in the phone channel with minimal pauses:
- The bot receives text from the LLM by sentences and synthesizes speech by sentences as well. The two processes occur in parallel.
- You can specify phrases that the bot will say to fill the pause at the beginning of text generation.
Get history in the LLM format
The new $jsapi.chatHistoryInLlmFormat
lets you get the history of the dialog with the bot in the LLM format.
For example, you can use this history in $gpt.createChatCompletion
or in the llmRequest
reply type.
Deploy bots to Mattermost
Mattermost is a messenger that is used as an internal chat in organizations and companies. It lets you create groups and channels, exchange files, and configure various integrations.
Now, you can deploy bots to this messenger via JAICP or Aimylogic. The Mattermost channel has the following features:
- Send and receive audio, video, images, files.
- Display buttons.
- Launch text campaigns.
- Transfer to agent.
$gpt.createChatCompletion method update
The $gpt.createChatCompletion
method lets you access ChatGPT models from your bot script.
Now you can:
- Select the model that will generate the response. GPT-4o mini is used by default.
- Pass an object with any parameters accepted by OpenAI API. This provides flexible control over the model behavior.
Fields in $nlp.match response
We updated the article about $nlp.match
and added a description of the new fields in the method response:
nBest
nluResults
intent
entities
With these fields, you can get more information about the classification results.