Skip to main content

April 2026 Update

· 3 min read
H&Z digital
H&Z GenAI Platform Team

The April update introduces GPT-5.4, reasoning effort selection, background chats, a unified settings page, and a refreshed landing page.

Highlights​

  • GPT-5.4 support now available as the default model
  • Reasoning effort selection to fine-tune AI response depth
  • Background chats with persistent task management, and background deep research with email on completion
  • Unified settings page with user, agent, prompt, and system settings in one place
  • Improved chat experience with unread indicators, first-unread auto-scroll, and better navigation
  • Enhanced agent management with improved search, refreshed layouts, and support for large agent lists
  • Updated UI including a redesigned landing page, documentation link, and dismissible banners
  • Markdown export for individual messages and full conversations
  • Admin enhancements including usage statistics, user deletion, and credential management
  • Stability & performance improvements across the platform

Unified Settings Page​

  • Consolidated all settings into a single, unified settings page
  • Moved user settings, agent configuration, template prompts, and system settings under one roof
  • Added API keys management UI to system settings
  • Added model settings with max tokens and reasoning effort configuration

Chat & Conversations​

  • Added unread message indicators for conversations
  • Auto-scroll to the first unread message when opening a chat
  • Added floating scroll buttons for easier chat navigation
  • Conversations now remember your last used model, agent, and reasoning effort
  • Improved background task management for chat persistence and recovery
  • Improved Title generation
  • Added reasoning effort selector to fine-tune AI response depth
  • Hide prompt edit button while a response is streaming

Agent Management​

  • Revamped Agent Card layout with refreshed details view
  • Redesigned agent selector with improved icon and search UI
  • Added infinite scrolling and pagination for large agent lists
  • Added reasoning effort dropdown for fine-tuning AI responses
  • Added loading indicator when uploading files to agent file explorer
  • Added ingestion processing toggle for agents
  • Fixed stale event handling in training progress tracking
  • Improved agent loading logic and robustness

Landing Page & UI Overhaul​

  • Complete landing page redesign with updated layout and visuals
  • Added configurable documentation link to the header
  • Dismissible warning banners with persisted state

Export & Markdown​

  • Added export options for individual messages and full conversations in Markdown format
  • Sanitized and truncated Markdown export filenames for safety
  • Consolidated response-cleaning logic for consistent rendering and exports

MCP Server Improvements​

  • Added MCP server connection testing via backend proxy to avoid CORS issues
  • Added email notification option when MCP server processing completes
  • Improved connection test reliability for existing MCP servers

Administration​

  • Added Usage Statistics dashboard for admins
  • Added user deletion functionality in user management
  • Added ability to share API tokens and manage user credentials
  • Added LLM Management UI
  • Personal system prompt option now shown conditionally based on user role

Bug Fixes & Reliability​

  • Fixed authentication response handling across API clients
  • Fixed category count display in template views
  • Improved template processing completion detection
  • Fixed keyboard event propagation in agent selector search