GuidesChangelogsAPI ReferencePublic RoadmapService Status
Changelogs
Start typing to search…
Back to All

February 2026

about 1 month ago by Ibtihal

Version 3.31.3

Released on February 18th, 2026

✨ New Feature & Enhancements

  • New Google Gemini Voice Model

Support has been added for the gemini-2.5-flash-native-audio-preview-9-2025 model. This new option is available when configuring AI Deployments for voice assistants with the Google Vertex AI provider.

  • Improved: AI Deployment Creation Flow

The alert message displayed when creating an AI Assistant without any available AI Deployments has been improved. It now includes a direct "+ Create AI model" button, allowing users to immediately open the creation modal and streamline the setup process.

  • Improved: Fully Customizable Main Dashboard

The main dashboard is now fully customizable. A new 'Personalize' button on the Home page opens a configuration modal where users can toggle the visibility of all available widgets.

Widgets are organized into two categories: KPIs & Metrics and Graphs. This allows for the creation of a personalized overview, enabling users to focus on the data that matters most to them.

  • Dynamic chunking management with custom instructions

Bug Fixes

  • Fixed: Duplicate CO2 Footprint Graph in Analytics

Corrected a display bug in the "AI Tokens" section of Analytics where the CO2 footprint graph was being rendered twice. The duplicate graph has been removed.

  • Fixed: Special Character Handling in "My Bots" Search

Resolved an issue on the "My Bots" page where the search filter would ignore special characters, such as underscores (_). The search now correctly handles these characters, ensuring that all bots are accurately found when searching by names that include them.

  • Fixed: Pagination in Core Modules

Corrected a significant bug where the pagination was incorrectly capped at 4 pages in several key modules. This fix applies to the Assistants, Knowledge Bases, Tools, and AI Deployments pages. Pagination now correctly displays all available pages, allowing full access to all items.

  • Fixed: Translation of Botman Tags

Addressed an issue where internal bot commands containing spaces (e.g., [GOTO:Mon solde]) were being incorrectly processed by the input translation service. This could cause the command to be partially translated and fail. The engine's regex has been updated to correctly identify and preserve these tags, ensuring reliable navigation and command execution in multilingual bots.

  • Fixed: NLU API Gateway "Socket Hang Up" Errors

Implementation of a Keep-Alive agent within the NLU API Gateway to prevent unexpected connection closures. This fix resolves "socket hang up" errors and ensures stable service communication during high-concurrency periods.

  • Fixed: User Entry Regex in License Management

Correction of the regular expression used for validating user input within the License Management tool.

  • Fixed: Crisp Integration Compatibility with Redis v4

Restoration of Crisp message delivery within the Webchat. The integration layer has been updated to ensure full compatibility with Redis v4, resolving a regression that prevented users from receiving messages following the database upgrade.

🛡️ Security & Infrastructure

  • Improved: Rate Limiting & Localization for MEA Licenses

Implementation of new request limits for MEA region licenses. The associated error message for rate-limit exhaustion (HTTP 429) is now fully localized, ensuring the response matches the bot's configured language instead of defaulting to English.

Version 3.31.2

Released on February 2nd, 2026

Translation Engine Enhancements

A series of improvements have been made to the translation logic to provide more control, reduce costs, and increase reliability.

  • RAG Responses Now Support Manual Translation: RAG-generated answers will now be translated according to the user_data.preferred_translation_language setting, ensuring that both NLU and RAG responses are consistently translated into the desired language.
  • Optimized Translation Logic: The system now intelligently skips the translation step if the user_data.preferred_translation_language is the same as the bot's native language. This prevents unnecessary API calls, saving tokens and reducing latency.
  • Improved Handling of Botman Tags: The translation engine now correctly identifies and preserves internal bot commands (e.g., [GOTO:My Balance]), preventing them from being partially translated and broken.

🛡️ Security Enhancement

Improved: User-Based Rate Limiting on Webchat : To maintain platform stability and prevent abuse, user-based rate limiting has been extended to the Webchat channel. This brings parity with the recent security enhancements implemented for Messenger and WhatsApp.

Bug Fixes & UI Improvements

  • Fixed: Date Filter Persistence in Analytics : Resolved an issue where active date filters were reset when navigating from the main Analytics dashboard to the Surveys (sondage) section. Selected timeframes are now correctly preserved during navigation.
  • Fixed: Modal Displacement in Webchat Integration : Corrected a UI bug where the configuration modal would become misaligned or offset after a specific navigation sequence (Integrations ➔ Builder ➔ Integrations).
  • Fixed: Evaluation Logic for Child Runs : Improved the evaluation workflow to ensure that existing evaluation processes are properly terminated when a new child evaluation (eval fils) is relaunched, preventing redundant runs and ensuring data accuracy.