# Release Notes: 2026 Sprint 7 **Period: April 16, 2026 – April 28, 2026** This sprint, the MyAlice engineering team shipped 18 features and improvements alongside 10 bug fixes — including a long-term chat memory MVP for Raven AI agents, a full-featured rich text email editor in Inbox, automated credit top-up for Billing, and a dedicated AI Agents queue, alongside a wide range of stability and usability improvements across the platform. *** ### 🚀 Features and Innovations **Inbox** * **Dedicated AI Agents Queue** — A new AI Agents queue now appears in the Inbox sidebar alongside the existing Bot queue, displaying all conversations handled by AI Agents. The queue supports tags, priority, and pin options, and auto-hides when no AI Agent tickets exist. * **Email Thread Included in Replies** — When replying to emails in Inbox, the full email thread (previous messages) is now automatically included, ensuring agents always have full conversation context visible to the recipient. * **Rich Text Editor for Email** — Email composition in Inbox now supports a rich text editor with formatting options including bold, italic, and more — giving agents more expressive control over their email communication. * **Unassigned Queue View Restriction** — The Unassigned queue now respects group access permissions. Agents without a supervisor role only see unassigned tickets and tickets assigned to their own groups, while supervisors retain full visibility. * **SLA Pause for On-Hold Tickets** — SLA calculations are now automatically paused when a ticket is placed on hold and resumed when the ticket is taken off hold. A yellow "Paused" label is displayed on the ticket for clear status visibility. **Raven** * **Long-Term Chat Memory** — Raven AI agents can now retain conversation context across sessions using a vector-based long-term memory system (powered by Qdrant). This enables personalized, context-aware responses without overloading the token window. * **Team Access Controls for Long-Term Memory** — Selective team permissions are now available for the long-term chat memory feature, allowing administrators to control which teams have access to AI agent memory capabilities. * **Model and Provider Switching** — AI agents can now be configured to use different AI models and providers — including OpenAI, Anthropic, Google, VertexAI, xAI, Alibaba, OpenRouter, and Ollama — with runtime switching, validation, and fallback handling. * **Reasoning Token Optimization** — Irrelevant reasoning tokens are now trimmed from runtime history for reasoning models, reducing hallucinations and improving AI agent response accuracy. * **Knowledge Base Tool Output Control** — An optional agent flag now allows KB tool outputs to be excluded from conversation history, keeping context clean and reducing noise while preserving other tool outputs. * **Expanded Conversation History Window** — The conversation history limit for AI agents has been increased to 12 message pairs, enabling better context retention and more accurate responses in longer conversations. * **Thinking Visibility in Response Trace** — The model's reasoning process is now visible in the response trace, giving teams greater transparency into how AI agents arrive at their decisions — useful for configuration, debugging, and quality review. **Broadcast** * **Refined CSV Audience Journey** — The broadcast audience selection flow for CSV uploads now includes a dedicated "Map Audiences" step. Users can toggle whether to retain mapping for previously used files and have the flexibility to re-select or upload new CSVs without being locked into the mapping screen. **Billing** * **Automatic Credit Top-Up** — Teams can now configure automatic top-up for both WhatsApp and Raven credits separately. Settings include a configurable threshold, top-up amount, payment method validation, failure warnings, and a disable confirmation prompt. Billing history is recorded automatically, and email notifications are sent for both success and failure events. * **Billing Email Standardization** — Notification emails for WhatsApp and Raven credit top-up events are now standardized across all partners, with consistent left-aligned text formatting for both success and failure notifications. **Integrations** * **Store ID Added for Single eCommerce Integration** — When only one eCommerce store is connected, the store ID is now automatically applied to all channels, and the Channel Preferences option is hidden to simplify the setup experience. * **Template Category and Status Filters** — The Manage Templates screen now includes dedicated filters for WhatsApp template category and approval status, making it faster to find and manage specific templates. **Settings** * **Updated Form Entry Created Webhook Payload** — The Form Entry Created webhook payload has been updated to include richer, more complete data fields — giving developers more information to work with on each form submission event. *** ### 🐛 Bug Fixes **Inbox** * Fixed multiple issues in Inbox Email: the "Assign to me" button was missing when a ticket was already assigned to someone else; messages from the previous queue were carrying over when switching between queues or between Email and Feed tickets; and note height was incorrectly expanding when it should remain fixed. * Fixed an issue where the email compose area appeared shrunk due to the rich text editor panel, making it difficult to view the full email while composing. The compose area now provides adequate space for both viewing and writing. * Fixed an issue where AI agents were appearing in the Agent filter dropdown under the All Assigned queue. AI agents are now correctly excluded from this filter. * Fixed AI Agent queue behavior: AI-handled tickets were incorrectly appearing in the All Assigned queue and Home section, the Hold action was showing in the AI Agent queue topbar, and the AA avatar was not displaying in the left bar channel name for AI-assigned tickets. All three issues have been resolved. * Fixed issues with TikTok DM's 48-hour messaging window: the expiry timer was being calculated from the wrong reference point, and the text input remained active with no restriction message after the window closed. The timer now correctly counts down from the customer's last inbound message, and the text box is disabled with a restriction notice once the 48-hour window has passed. **Raven** * Fixed an issue where the Save button remained disabled after pasting content directly into the Bot Instructions field. The Save button now activates immediately when content is present, regardless of whether it was typed or pasted. * Fixed an issue where cloning a Comm Agent resulted in it being created as a DM Agent, and associated tools (Hide, Delete, Send Private Reply) were not being copied over. Comm Agents now clone correctly with all tools preserved. * Fixed an issue preventing proper response generation and delivery for email channel messages handled by the Email AI Agent. **Reports** * Fixed an issue where column order in Agent Performance Metrics exports was inconsistent, causing data columns to appear in different sequences with each download. Column order is now fixed and consistent across all exports. **Broadcast** * Fixed an issue where the Broadcast Audience report displayed client names as "anonymous" even when names were already recorded in the broadcast sheet. Client names now appear correctly in the report. *** *These updates are live for all MyAlice users. As always, if you have any questions or run into anything unexpected, reach out to our support team — we're here to help.* --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://docs.myalice.ai/release-log/release-notes-2026-sprint-7.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.