Changelog
AppService status
Back to All

v263 Expression Autocomplete Everywhere ✨

about 3 hours ago by Michał Brzezicki

"I cannot deactivate until you say you are satisfied with your care." - Baymax, Big Hero 6 🤖
SentiOne Automate version 263 was released on 01.07.2026. This release delivers the highly requested AEL expression autocomplete in Responses and all interpolated fields, along with smarter WebChat session handling, voice channel configuration validation, and LLM enhancements.

Expression Autocomplete in Responses

The autocomplete known from AEL expression inputs is now available in Responses and all other fields that support interpolated strings. Suggestions cover all AEL expressions — memory keys, system variables, NLU values, and commands — appearing as you type, making it faster and less error-prone to reference dynamic values in bot messages. This has been one of the most requested features from our users.

WebChat Session Recovery

WebChat now correctly handles conversations that expire during long periods of inactivity. When a user returns after their conversation has expired, the widget detects the expired state, displays the full conversation history, and offers a clear option to start a new conversation — preserving any initial data (such as user context passed by the host application) when the new session begins. This significantly improves the experience for WebChat embedded in mobile applications.

Voice Channel Configuration Validation

TTS and ASR settings are now validated when saving a voice channel. Incorrect credentials or configuration parameters are caught immediately, preventing broken voice channels from being deployed. Validation failures display a clear error message with a request ID for easier troubleshooting.

LLM Request Logging

The conversation logging capabilities introduced in v262 now extend to LLM blocks. A new logging toggle in LLM Say and LLM Integration blocks records resolved requests (and responses for LLM Integration), making it easier to understand exactly what was sent to the language model. Logging is enabled by default for newly created blocks.

Tool Calling for OpenAI-Compatible Models

LLM tool calling is now supported for OpenAI-compatible models, including open-source models. This enables advanced conversational scenarios with self-hosted or third-party language models that follow the OpenAI API format.

Self-Hosted WebChat Assets

WebChat styles, fonts, and icons are now served directly from the application instead of an external CDN. This removes external dependencies — particularly important for on-premise deployments with restricted network environments.

💊 IMPROVEMENTS

  • [Flow] AEL expression autocomplete available in Responses and all interpolated string fields
  • [WebChat] Expired conversations detected after long inactivity — history shown with option to start a new conversation
  • [WebChat] Initial context data preserved when starting a new conversation after expiration
  • [WebChat] Assets served from the application instead of an external CDN
  • [Voice] TTS/ASR configuration validated when saving voice channels
  • [Voice] Clear validation error messages with request ID for troubleshooting
  • [LLM] Request logging toggle in LLM Say and LLM Integration blocks
  • [LLM] Tool calling support for OpenAI-compatible models, including open-source models
  • [Genesys] Retry limiter added to gracefully handle Genesys API rate limits

🐛 FIXES

  • [Flow] Log block objects now keep their original order after saving

  • [Flow] Corrected context execution order when returning from nested contexts

  • [WebChat] Words are no longer incorrectly split into syllables in message bubbles

  • [Stability] Improved bot engine startup reliability under heavy load

Follow the changelog for more exciting new features! 👋