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

March 2026

24 days ago by Ibtihal

Version 3.31.6

Released on March 26th, 2026

✨ New Feature & Enhancements

1. New: Real-Time API Mode Support for LLM Providers

Integration of the Real-Time (Chat Completion) mode within the AI Deployment interface. This functionality is now available for LLM Proxy, OpenAI, and Microsoft Azure OpenAI providers, allowing for streaming responses and lower latency interactions.

2. Improved: AWS Bedrock Visual Branding

Addition of the AWS Bedrock icon in the AI Deployment selector. This update ensures consistent UI representation and easier identification when selecting AWS as an infrastructure provider.

3. New: Model Context Protocol (MCP) Connector Management

4. New: MCP Server Integration for LLM Proxy & Function Calling

Full support for Model Context Protocol (MCP) server connections within the LLM Proxy layer. This integration enables Smartly Agents to perform standardized function calls and query custom external tools directly through the proxy. By developing and connecting an MCP server, agents can now interact with private data sources and execute complex tool-based workflows with unified management.

🐞 Bug Fixes

Fixed: Dashboard Activity History Filtering

Resolution of an issue where the "Recent Activity" history failed to display when a date filter was applied. Default date ranges have been implemented to ensure the dashboard remains populated with data upon loading.

Fixed: UI Overflow in AI Deployment Selector

Resolution of a layout issue where the dropdown options in the AI Deployment selector were partially hidden or cut off at the bottom of the modal. The interface now ensures that the selection list is fully visible and that the scroll position is correctly managed when the menu is opened.

Fixed: Multi-Language Save Issue in Voice Welcome Manager

Resolution of a bug that prevented saving configurations within the Voice Mode's Welcome Manager when a second language was added. This fix ensures that greeting messages and locale settings are correctly persisted for all configured languages.


Version 3.31.5

Released on March 19th, 2026

✨ New Feature & Enhancements

1. New: Voice Mode for Webchat Integration

Introduction of a dedicated Voice Mode within the Webchat configuration. This feature allows assistants to engage in interactive voice conversations.

  • Select Voice Assistant: Choose a specific neural voice model for the integration.
  • Welcome Manager: Configure a custom spoken greeting message that triggers when the voice session starts.

  • 2. Improved: Team & Permissions Modal for Bot Sharing Visual update of the bot sharing and permissions interface. User access for specific modules (Analytics, AI Assistants, Conversations, etc.) is now managed through intuitive, toggleable Read/Write icons, simplifying team role assignments.

  • 3. New: MAXit and In-App Channel Support for API Dialog Expansion of channel filtering capabilities within Conversations and Analytics. Two new identifiers have been added to the API Dialog channel:

    • MAXit: Dedicated filter for in-app integrations within the MAXit ecosystem (keyword: maxit).
    • Application: Filter for generic in-app integrations outside of MAXit (keyword: application). This improvement allows for precise traffic segmentation, similar to existing Messenger and WhatsApp filters. The API Dialog documentation has also been updated to include these new technical specifications.

  • 4. New: Multi-Masterbot Filtering in Analytics Expansion of filter capabilities within the dashboard and analytics sections. Users can now select one or multiple Masterbots simultaneously to refine their view. This allows for consolidated data analysis across specific projects, providing more flexibility beyond the existing date-based filtering.

  • 5. New LLM Provider: AWS Bedrock Support Integration of AWS Bedrock as a selectable provider for AI Deployments. This allows Smartly Agents to leverage foundation models hosted on AWS (such as the Anthropic Claude 3 series) with full support for specific configuration parameters including region selection, access keys, and auto-continue attempts.

🔐 Security & Governance

  • New: Dedicated SSPO Support Role for MEA Monitoring

    Implementation of a specialized permission profile for regional support teams (SSPO), enabling comprehensive bot monitoring while ensuring platform integrity.

    • Enforced Read-Only Access: The Builder and Dataset modules (including Intents, Slots, and Knowledge Base components) are restricted to "Observer" status for this role.
    • UI Hardening: Critical action buttons, such as "Train" and "Add Language," are automatically hidden within the interface to prevent unauthorized modifications.
    • Fixed: Permission Synchronization: Resolution of a bug where the previous bot's access rights remained in memory when switching bots via the selector in the Datasets module.
    • Improved Inbox UX: Enhancement of the Inbox interface allowing for direct actions upon clicking a message row, streamlining support workflows.

🐞 Bug Fixes

  • Fixed: Duplicate Date Markers in Conversations Correction of a UI rendering bug where date separators (e.g., "27 Feb 2026") were appearing twice within the conversation history view.


Version 3.31.4

Released on March 2nd, 2026

✨ New Feature & Enhancements

  • Improved: Rate Limit Management for Licensed Users

    Increase of the default rate limit to 150 req/min for users with an active license. The configuration for rate limits (per minute or per month) is now optional, as long as at least one of the two parameters is defined.

  • Improved: Vectorizer Model Display in Embeddings The Document Search (Embeddings) section now explicitly displays the specific vectorizer model name instead of a generic base model label. This provides better technical clarity on the specific model used for the vectorization and indexing process within the Knowledge Base.

🐞 Bug Fixes

  • Fixed: AI Assistant Replacement in Integration Flow Resolution of a bug preventing the addition of a new AI Assistant after stopping and deleting an existing one within the bot integration section.

  • Fixed: NLU Model Configuration & Hybrid Mode Stability Correction of an issue where changing the NLU model incorrectly disabled the Hybrid NLU mode and caused inconsistencies in the "AI Deployment" selection after saving.

  • Fixed: PPTX to ZIP File Downloads in Webchat Resolution of a technical issue preventing the correct downloading of .pptx files when delivered as part of a .zip archive through the Webchat interface.

  • Fixed: Redundant Authentication on Knowledge Base Inactivity Elimination of a double-authentication prompt that appeared following account inactivity within the Knowledge Base (KB) module.

  • Fixed: Persistent Modal Overlay Blocking UI Resolution of a bug where the modal-backdrop component would occasionally remain active after closing a modal via the "ESC" key or the close button. This fix prevents the interface from becoming non-responsive due to a persistent background overlay when no modal is visible.

  • Improved: Website Indexing Reliability with Puppeteer & Cheerio Resolution of ingestion failures when indexing external websites for Knowledge Bases. Connectivity has been optimized to prevent server IP blocking by third-party security rules, ensuring consistent data extraction when using Puppeteer and Cheerio libraries.