Changelog
AppService status
Back to All

v255 Import/export of NLU models & new GPTs 💾

2 days ago by Michał Brzezicki

“Everything happens for a reason." - Coach AI, Black Mirror, Hand the DJ, S4E4 🤖
SentiOne Automate version 255 was released on 27.09.2025. This version introduces new way to export, backup and upload trained NLU models together with multiple improvements.

Export and import of trained NLU models

Exporting, importing and updating of NLU models was available for longer time time. However, only training sets could be downloaded and uploaded to system. With this release the export file can include trained model as well.

This feature exports binary copy of trained NLU model so that training model is not needed after update. Since it is a binary copy it should work exactly the same as on the environment where model was exported from. With seed function which is known to be unstable depending on CPU architecture, this method of promoting NLU models to higher environments is much more stable and therefore recommended.

Added new LLMs

Following new models are now available in LLM Say and LLM Integration:

OpenAI:

  • GPT-5
  • GPT-5-mini
  • GPT-5-nano
  • o4-mini
  • o3
  • GPT-4.1
  • GPT-4.1-mini
  • GPT-4.1-nano

Google Gemini:

  • gemini-2.5-flash
  • gemini-2.5-flash-lite
  • gemini-2.5-pro

Please note that platform support hundreds of other LLMs though sites like OpenRouter, which include open source models like Bielik, PLLuM, DeepSeek, Qwen, LLama, etc. To enable them use LLM Compatible with OpenAI connector in AI Credentials

💊 IMPROVEMENTS

  • [API] Listing AI Credentials and Knowledge Base Documents in API
  • [Flow] Smart handling on longer processing time in LLM Say, LLM Integration and Integration blocks. If certain timeout is exceeded user will be given message about prolonged processing time.
  • [Security] Keys, passwords and other secret items are now blurred by default in Knowledge Base
  • [Flow] Bot can now set user name through commands SetAuthor and SetVerifiedAuthor
  • [NLU] All NLU modules now use Redis as a distributed storage of the models
  • [Conversations] You can now see the delay of the responses
  • [Genesys] Adding Genesys conversation attributes with SetExternalSessionAttributes command
  • [Genesys] Informing Genesys agent about closed webchat conversation
  • [Knowledge Base] Rendering markdown for Knowledge Base Documents

🐛 FIXES

  • [Analytics] Optimized fetching tags in Analytics view
  • [NLU] Fix for cross validation of big NLUs
  • [Webchat] In webchat agent icon was flashing while replying
  • [Webchat] Fixed restoring webchat websocket connection
  • [Webchat] Fixed issue of carousel view in conversation analytics
  • [Organization] Fixed coping of invitation URL
  • [Flows] Fixed play audio button in chat tester

Follow the changelog for more exciting new features! 👋