h2oGPTe v1.6.54​

New features​

User interface and experience​

• Inline reference pills: Chat responses now include hoverable citation pills that reveal document details on hover and open the source on click, enabling users to easily validate information and navigate supporting documents. Click any reference pill to view the relevant passage.
• PDF highlighting: When you click an inline reference, the PDF viewer opens and automatically highlights the exact passage, making it easier to verify sources.
• PDF original view mode: View PDF documents in their original format with a new viewing option.
• Side panel file viewer: Open inline images and file links directly in the side panel without leaving your current view.

Search and filtering​

• RAG query filters: Include or exclude specific content using filters on your chat queries for more precise results.

Connectors and integrations​

• Confluence attachments: The Confluence connector now supports document attachments, allowing you to import attached files along with page content, with improved reliability and expanded content import capabilities.

Models​

• GPT-5.2: Added support for OpenAI's GPT-5.2 model.
• Gemini-3-pro-preview: Added Google's Gemini-3-pro-preview model option.
• Embedded models: Added support for new embedded models including embeddinggemma-300m-qat-q8_0-unquantized and mxbai-embed-large-v1.

Improvements​

User interface and experience​

• Markdown document preview: Markdown files now correctly display as text previews instead of raw content.
• Auto-generated turn titles: Agent turn message panels now automatically generate titles when none is provided.
• Simplified prompt template view: The prompt template view is now streamlined when using agentic RAG.

Bug fixes​

User interface and experience​

• Page previews: Fixed page previews incorrectly displaying the full PDF document instead of the intended view section.
• View selector dropdown: Fixed view selector dropdown appearing empty and preventing proper view selection.
• Vision UI toggle: Fixed regression where the enable vision UI toggle no longer displayed the Off/Auto/On options, restoring full vision control functionality.

Chat and sessions​

• Chat history stability: Fixed issues with chat history management and message alignment on large screens.

Security​

• Security updates: Enhanced system security with base image updates that address multiple reported CVEs and other security vulnerabilities.

h2oGPTe v1.6.46​

New features​

User interface​

• Math formula toggle: Control how mathematical formulas display in chat responses. Enable or disable formatted equation rendering in your preferences to customize your viewing experience.

Improvements​

User interface and experience​

• Chat history display: Chat history button now appears consistently across all conversations, making it easier to review and reference past interactions.
• Agent response filtering: Improved filtering of thinking sections in agent responses for cleaner, more focused content that's easier to read.

Security​

• Security updates: Enhanced system security with updates that address multiple security vulnerabilities.

Bug fixes​

Content rendering​

• Text-to-speech: Fixed an issue where the text-to-speech feature wasn't working. You can now use the speech button to hear responses read aloud.

h2oGPTe v1.6.45​

Improvements​

Content formatting and display​

• Markdown math rendering: Mathematical equations written in markdown now display as properly formatted equations, making technical documentation and formulas easier to read and understand.

User interface and navigation​

• Settings button visibility: The + New Settings button on the System Settings page is now disabled when no configuration options are available, providing immediate visual feedback about which areas can be customized.

Bug fixes​

User interface and messaging​

• Chat history clarity: Internal system messages are now hidden from chat history, keeping conversations focused on your questions and the agent's responses.
• Error message accuracy: Error messages throughout the application have been updated to provide clearer descriptions of issues and actionable guidance for resolution.

System reliability​

• Workspace validation: Added validation checks for workspaces and system configuration to help prevent errors when required setup steps are incomplete, enhancing overall system stability.

h2oGPTe v1.6.44​

Improvements​

User interface and experience​

• Chat toolbar location: Moved the toolbar into the chat input area so actions like starting the agent or attaching files sit next to the message box.
• Collection name in chats: Show the collection name in collection-based chats so you can see which content the agent uses.
• References slider: Added a slider to focus on the highest‑scoring referenced passages instead of showing every match.
• Topic model management: Added a button to delete Topic Model directly from the UI to simplify analytics cleanup.
• Cloud ingestion jobs: Renamed cloud ingestion jobs with clearer, more descriptive titles so you can identify each job quickly.
• Licensing information: Standardized licensing information display so deployment details are easier to review.

Connectors and integrations​

• Connector dialog redesign: Redesigned the connector configuration dialog to clarify fields and simplify setup.
• Connector navigation: Moved connector access from the sidebar to the profile area to group advanced settings.
• SharePoint OAuth settings: Added centralized SharePoint OAuth configuration to the Admin Dashboard (under System Settings) with admin-only access and built-in encryption of sensitive information to enhance security in enterprise environments.
• Confluence sync status: Added clearer progress reporting while downloading Confluence pages.
• Confluence page names: Allowed Confluence pages with slashes in their names.
• Cloud storage integrations: Extended the S3 connector to support AWS IAM Roles for Service Accounts (IRSA) and role chaining.
• Connector ingest hints: Added clearer hints for configuring ingestion through connectors by hovering over the method name in the drop-down menu.

Agents and models​

• Shared agent model option: Added an option for administrators to configure a shared default language model across agents.
• Default guardrails flag: Added a configuration flag H2OGPTE_MUX_DEFAULT_GUARDRAILS_ENABLED to enable or disable default agent guardrails globally.

Administration and identity​

• User identity display: Aligned username display from single sign-on (SSO) providers across the UI and audit logs for consistency.

Performance improvements​

• PDF highlighting: Switched PDF highlighting to use a global threshold for more consistent visual citations.
• Timeout handling: Reduced timeout failures for long‑running operations.
• Storage reliability: Improved storage connection handling for more stable document storage under load.
• Filtered document lists: Fixed document listing with filters so filtered views are more accurate and responsive.

Bug fixes​

User interface and documents​

• File handling: Fixed previews for files with trailing dots or dot‑prefixed names, and prevented empty files from causing indexing errors.
• Drag-and-drop uploads: Fixed a form control error that occurred when drag‑dropping files.
• Extractor messages: Fixed a misleading error message when saving extractors.
• Collections and dialogs: Fixed the add‑documents dialog height, edit‑mode checkmarks, and dark mode display for connector screens.
• Topic model labels: Corrected Topic Model labels so they render as expected.

Authentication and authorization​

• Single sign-on usernames: Fixed inconsistencies in how usernames from SSO providers are displayed.
• API key expiry toggle: Fixed API key expiry toggle behavior on the API page so it aligns with actual key expiry.

Workspaces and configuration​

• Personal workspace assignment: Added stricter checks when setting personal workspaces to prevent incorrect assignments.
• Expiry defaults and limits: Corrected expiry default behavior and global expiration limits to match configured settings.

Connectors​

• Confluence reliability: Improved error handling in the Confluence connector.
• SharePoint reactivity: Ensured SharePoint connector settings update correctly when the main server configuration changes.

Agents and models​

• Agent chat history: Fixed issues with old agent chat history when agents are enabled but chat history is turned off.
• Private mode LLM filtering: Fixed broken private mode LLM filtering.
• Agent isolation: Fixed agent isolation behavior and prevented built‑in code execution tools from sharing keys unexpectedly.

h2oGPTe v1.6.43​

Improvements​

Workspaces and organization​

• Workspace functionality: Added workspace feature to organize and manage your content with tagging and tracking capabilities.
  • Existing personal content automatically migrates to the new workspace system

Agent capabilities and tools​

• Tool Builder Agent: New specialized agent type designed for building and customizing tools
• Automatic model routing: Agents now automatically select the best model based on accuracy requirements for optimal results
• Agent monitoring: Added tracking for agent conversation turns to help monitor agent usage
• Final Answer Mode: New dropdown in agent configuration to control how agents deliver final responses
• Agent streaming and files UI: Improved real-time streaming interface and file management for agent responses
• Human-in-the-loop planning: Added control to review and approve agent planning before execution
• Document auto-tagging: Agents can now automatically tag documents using specialized tools
• Tool defaults: Option for admins to enable tools by default when creating new agents for faster setup
• Authentication improvements: Improved authentication tab interface for configuring agent credentials

New model support​

• Google Vertex AI Gemini 2.5 Pro: Added support for Google's latest Gemini model via Vertex AI
• GPT-5 Codex: Added support for OpenAI's GPT-5 Codex model

Document processing and connectors​

• Video scene descriptions: h2oGPTe now generates automatic scene descriptions for better understanding of video content
• Auto-Sync Collection Connector: New connector that automatically synchronizes collections with external sources
• SharePoint OAuth2 authentication: Enhanced SharePoint connector with OAuth2 user authentication for improved security

User interface and experience​

• Clickable table links: Tables now support clickable links for better navigation
• Microphone animation: Added visual animation when microphone is active during voice input
• Chat session creator: Users can now see who created each chat session on chats and collection pages
• Shareable extractors: Extractors page can now be shared with other users
• Collection application: Improved interface for applying collection settings in chat sidebar
• Chat with specific collection: Added ability to start a chat directly with a specific collection
• Document deletion: Users can now delete documents from collections when they have proper permissions

Privacy and security​

• Custom PII detection: Added AI-powered detection for user-defined personal information labels

User experience refinements​

• Global settings discoverability: The global settings are now under System Settings tab for easier access
• Action menus on inactive chats: Added action menu buttons for archived or inactive recent chats
• Searchable user dropdowns: User selection dropdowns now support search for easier user finding
• Searchable model options: Model selection dropdowns now support search functionality
• Dark mode improvements: Fixed font color issues in dark mode for better readability
• Android voice improvements: Enhanced speech synthesis for Android webview users
• Role selector: Added role selector to the side menu in the user interface.

Interface improvements​

• Disabled input when recording: Text input automatically disabled when microphone is active to prevent conflicts
• Title truncation: Long titles are now automatically truncated with ellipsis for cleaner display

Performance improvements​

• Faster collection loading: Reduced API calls for collections and documents, resulting in faster page loads
• Better document retrieval: Markdown and HTML content now processed more effectively for improved search and retrieval

Bug fixes​

User interface​

• Layout fixes: Fixed padding and spacing issues in primary and secondary panels
• Code viewer: Removed width restrictions on code viewer for better code display
• File status badges: Fixed positioning of file status badges
• Height issues: Resolved height calculation problems in various interface components
• Markdown tables: Improved handling of edge cases in markdown table rendering
• UI crash prevention: Fixed crash that occurred when no language models were configured

h2oGPTe v1.6.41​

Improvements​

User interface and experience​

• Spanish language support: Added full Spanish language localization throughout the user interface
• Language-specific voice: Text-to-speech now automatically speaks in the language you've selected or the language of your prompt template
• Collection sharing: Added share button on Collections page to easily share collections with team members
• Prompt sharing: Added share button on Prompts page to share your custom prompt templates with team members
• Share dialogs redesign: Improved sharing dialog box for collections and prompts with clearer options and better layout
• Chat session expiration: Added ability to set expiration dates for shared chat sessions for better security control
• Document filtering: Added toggle to switch between viewing all documents or only your own documents
• Document references: Document links now display their source location for easier identification and navigation
• WebP image support: You can now upload and view WebP format images in your collections
• Mobile menu: Fixed navigation menu glitch on mobile devices for smoother browsing
• Table display: Fixed table width issues to ensure consistent display when resizing your browser window
• Toggle controls: Resolved toggle button width issues for more consistent appearance
• Status updates: Fixed real-time status updates on cards and tables to display changes immediately
• File information: Fixed display of file names and sizes to show complete information
• Collection previews: Resolved issues with document preview displays in collections
• Markdown links: Fixed issue where clickable links were being removed from formatted text responses
• Chat formatting: Chat responses now maintain proper formatting after sensitive information removal

Agent capabilities and personalization​

• Data Science agent: Added a new agent type designed for data analysis, statistics, and analytics workflows
• User Persona selector: Added dropdown in Configuration panel to select your role or persona for more tailored responses
• Custom Tools framework: Administrators can now create, build, and deploy custom tools for specialized tasks
• Tool filtering: Added filter dropdown in agent tools table to find tools by type
• Browser navigation: Added new tool that allows agents to browse and interact with web pages

Administrator features​

• Secret Manager panel: Administrators can now manage all API keys and credentials from a centralized, secure admin panel
• Default prompts: Administrators can customize default prompt templates for your organization
• Output token limits: Added global setting to control maximum response lengths for cost management
• API key expiration: Administrators can set automatic expiration dates for API keys
• Role customization: Administrators can configure custom role settings visible via System Dashboard

Performance improvements​

• Faster document processing: Documents are now indexed faster through concurrent processing
• Improved page rendering: Pages with complex formatted text now load faster
• Smoother email handling: Email documents now display more quickly

Bug fixes​

Security​

• Session security: Improved protection against unauthorized session access
• Security vulnerabilities: Addressed security vulnerabilities CVE-2025-22874 and CVE-2025-4565

Stability​

• Document storage: Resolved timing issues between database operations
• Database Operations: Improved handling of database connection issues for more stable operation
• System cleanup: Improved cleanup processes to prevent resource buildup over time

h2oGPTe v1.6.36​

Improvements​

User interface and experience​

• Prompt templates: Added prompt template group sharing functionality
• Document management: Fixed manual document tag creation
• Document browsing: Resolved pagination limit issues when browsing documents
• Desktop filters: Adjusted document filters minimum width to prevent layout overflow
• File upload: Improved ingestion dialog appearance and user experience
• Banner support: Added configurable banner functionality for system administrators

Security and authentication​

• User permissions: Fixed issue where users could assign shared keys from other users, now properly limited to admin shared keys
• Guest authentication: Improved guest user fingerprint handling as subject identifier
• Token exchange: Enhanced token exchange for JavaScript RPC client with performance optimizations

System administration and configuration​

• Settings management: Improved settings with automatic updates and better performance

Performance and scalability​

Processing optimizations​

• Concurrent processing: Improved handling of concurrent requests for better throughput
• Image processing: Added support for non-A4 format pictures with proper fitting

Backend improvements​

• MCP enhancements: Upgraded Model Context Protocol (MCP) with various improvements
• Browser automation: Updated Puppeteer for improved browser-use support

Bug fixes​

Agent functionality​

• Environment variables: Improved agent environment variable handling for better reliability

User interface​

• Document tags: Resolved manual document tag creation problems
• Collection sharing: Enhanced collection sharing functionality with proper access controls

Documentation improvements​

• System administration: Enhanced documentation for system administration and configuration features

Model and integration updates​

h2ogpt updates: Multiple h2ogpt version updates with various improvements and fixes

h2oGPTe v1.6.33​

Improvements​

User interface and mobile experience​

• Collection management: Fixed collection count display when filtering through search, ensuring accurate result counts
• Error messaging: Enhanced error messages for administrators who lose access to private collections
• Card layouts: Improved username display in card views with cleaner text formatting and responsive design for mobile devices
• Mobile navigation: Better menu handling and responsive design for smaller screens
• Code rendering: Enhanced prompt code block rendering with syntax highlighting headers and improved formatting

API and developer experience​

• Documentation: Added comprehensive async API examples covering authentication, document management, and chat operations
• Code execution: Enhanced agent tool detection for Python and shell code blocks, expanding supported programming languages

Bug fixes​

System stability​

• Error handling: Improved UI reliability and stability
• Agent sessions: Enhanced error handling for agent chat sessions to reduce failure rates when restarting messages

User interface​

• Fixed cursor navigation issues in code blocks for keyboard users
• Resolved mobile button list rendering errors affecting interface stability
• Improved error handling for collection sharing with clearer permission feedback

Backend and API​

• Enhanced persistence of collection configuration settings, eliminating page refresh requirements
• Fixed non-streamed REST API response handling for better reliability

Performance improvements​

• Enhanced PII (Personally Identifiable Information) detection and redaction with improved accuracy and expanded data type coverage for document processing workflows
• Improved arXiv document downloader reliability for consistent document processing

New model support​

New model support in this version:

• Claude 3.5 (Bedrock)
• OpenAI o1 (Azure)
• OpenAI o1-mini (Azure)
• Gemini 2.0 Flash
• Gemini 2.0 Flash Thinking
• DeepSeek V3
• DeepSeek R1
• Qwen/Qwen2.5
• Qwen/Qwen2-VL
• Qwen/QwQ

h2oGPTe v1.6.32​

Improvements​

Document and collection management​

• File management: Added delete option to documents grid view for easier file management
• Collections layout: Enhanced space management and visual layout on Collections page
• Processing speed: Introduced Lite ingest mode for faster, streamlined document processing

User interface and experience​

• Mobile usability: Enhanced filter bar on Documents page for improved mobile device usability
• Private mode: Improved UI appearance in private mode
• Notifications: Moved model failure notifications to sidebar alerts to reduce user experience interruptions
• Session management: Enhanced auto-logout logic to prevent premature session terminations
• Localization: Added prompt template for British English to support international users

System administration and configuration​

• License flexibility: Added support for older H2O.ai public keys during license checks
• Network configuration: Exposed configurable WebSocket ping timeouts for administrators to fine-tune chat session connections
• Storage optimization: Exposed S3 connection limits for tuning during high-load document ingestion periods
• Feature control: Made web crawl functionality optional, giving administrators control over external website crawling as a RAG (Retrieval-Augmented Generation) ingestion method

Performance and scalability​

Processing optimizations​

• PDF handling: Resolved slow ingestion speeds for certain PDF documents
• Chat performance: Optimized chat queries for faster, more efficient performance
• Processing reliability: Ensured unique processing paths for per-page PDFs to improve ingestion reliability
• Browser updates: Updated to latest Chromium version for improved performance and security

AI and agent enhancements​

• RAG optimization: Enhanced prompt queries to ensure LLMs recognize RAG task performance
• Metadata efficiency: Deduplicated document metadata sent to LLMs for improved efficiency and response quality
• PII detection: Updated PII (Personally Identifiable Information) model and detection threshold

System reliability​

Stability improvements​

• Service coordination: Enabled sub-services to refresh and register state independently for better system awareness

API and developer enhancements​

• Performance optimization: Optimized automatic chat naming feature logic by checking settings before pulling chat history

Bug fixes​

User interface and application​

• Agent messaging: Fixed bug where new agent chat messages would fail after stopping previous messages in the same session
• Collection integration: Fixed agent chat with collections failing to include RAG context
• Configuration persistence: Fixed collection configuration settings not applying to new chats until page refresh
• API responses: Ensured final responses for non-streamed REST API calls are always returned
• Template sharing: Fixed issue preventing non-owners from sharing prompt templates
• Document references: Ensured most up-to-date document names are used in references
• Page synchronization: Fixed page resynchronization issues after redaction
• Navigation consistency: Fixed back button alignment with page headings across all relevant pages
• UI consistency: Aligned thumbnail picker container appearance with thumbnail card containers
• Message highlighting: Applied correct parsing algorithm for highlighting evaluated chat messages
• User interface: Fixed visibility condition for user pairing link

Backend and API​

• API reliability: Improved file type detection for better document processing