Skip to content
LIVOI
DE Dashboard

Changelog

All platform updates in chronological order, including screenshots and technical improvements.

  1. 1.1.0

    Concurrent chats, a redesigned dashboard and more flexible custom tools

    Dashboard

    The dashboard shows conversations, completed agent responses, active users and token usage, each with changes from the previous period. Switch the usage chart between responses and tokens, or view and copy its data as a table. Breakdowns by channel, agent and model show where usage comes from.

    “Agent responses” replaces the previous message count, which included user messages. Date boundaries, failed responses and comparisons across daylight-saving changes are handled correctly.

    LIVOI dashboard in version 1.1.0 with agent responses, token usage and breakdowns by channel, agent and model LIVOI dashboard before the update with message trends and platform statistics
    Before1.1.0

    Chat

    New chats start with the greeting, agent description and input together in the center. During a conversation, the input moves to the bottom and becomes more compact in longer histories. A desktop timeline lets you preview and jump to individual turns.

    Empty LIVOI chat in version 1.1.0 with the greeting, agent description and input together in the center Empty LIVOI chat before the update with the greeting at the top and the input at the bottom
    Before1.1.0
    • Concurrent conversations: Responses can run in several chats at once. Drafts, attachments and ongoing responses stay with their conversation; a background chat finishing does not change the current view.
    • Prepare follow-ups: Queue more messages and attachments while a response is running. Remove queued messages, or pause and resume the queue. After failures, unsent follow-ups can be recovered as editable drafts. Queuing and recovery apply to the current session.
    • Longer input: Messages support up to 65,536 characters. Large pasted texts become text-file attachments automatically; Ctrl/Cmd+Shift+V keeps them inline within the message limit.
    • History and attachments: Open original files for every attachment type. Exports include the complete conversation; structured responses, attachments and tool details survive reloads and reconnects. Duplicate responses and incorrectly empty histories have been fixed.

    Inbox

    The conversation list shows agent avatars and names alongside chat titles. Label changes also appear in other open sessions, including removal of the last label. Large selections and reloaded histories are handled more reliably.

    LIVOI inbox in version 1.1.0 with agent avatars, agent names and smaller WhatsApp icons in the conversation list LIVOI inbox before the update with WhatsApp icons and chat titles in the conversation list
    Before1.1.0

    Agent settings

    Profile, model, instructions and assignments are together on one page with expandable sections. Edit the name and model options there and save them together. Files and calendars appear according to the enabled tools.

    Create new agents on a dedicated page. A confirmation helps protect unsaved input when leaving; agent switching and large selection lists respond faster.

    Agent settings in version 1.1.0 with model, instructions and assignments in one view Agent settings before the update with separate tabs for instructions, tools and files
    Before1.1.0

    Custom tools

    Custom HTTP tools can map agent inputs into fixed JSON requests, including nested structures. Input types and defaults are preserved.

    • Preview requests: Inspect the URL, headers and body using example arguments without contacting the external API. Credentials are masked.
    • Send real tests: After explicit confirmation, send a request to the external API. This can create, change or delete data there. Previewing and testing do not save changes to the tool configuration.
    • Inspect tool calls: Request information, status and error details are available on tool calls in chat and remain in the history. Repeated calls to the same tool are grouped compactly and can be expanded.
    • Discover tools when needed: Enable on-demand discovery for each custom tool. The agent then receives its description and input schema when it discovers the tool. Existing tools remain directly available by default.
    Tool request preview with example arguments, validated arguments, request URL, masked headers and request body

    Other improvements

    • Guided introduction: Short product tours explain the interface using isolated demo data. They leave real records untouched and return you to the previous view when finished.
    • Files and calendars: Files already assigned to an agent are hidden in its file picker. Large file, calendar and group-member lists load completely. When some assignments fail, successful changes are preserved and errors are reported.
    • WhatsApp: Newly connected numbers can be registered and verified more reliably. All website entries can be removed from a business profile. Templates support media headers, and the notice about Meta’s required payment method is clearer.
    • Settings and privacy: Member, group and role pages reuse loaded data when switching between them. Download the data processing agreement as a PDF under Settings → Data and privacy.
    • Loading and recovery: Previously loaded content remains visible where possible during temporary failures, with a notice and retry option. Sign-in and sign-up are simpler; new accounts wait for setup to finish. Navigation, permission updates and tenant switching are more reliable.
    • Interface and languages: More compact headers and updated P-CATION colors make the interface more consistent. German, English, Italian and Polish translations have been completed.
    Technical improvements and API notes
    • Real-time updates and history: Typed tool results and structured responses are preserved in WebSocket updates. Deletion events also reach authorized owners of their own resources; recovery after connection loss has been improved.
    • Chat streaming: Streams send blank NDJSON lines as heartbeats during longer pauses. Custom clients must skip these lines when parsing. Streaming and regular chat requests accept up to 65,536 characters.
    • Whitelist events: Entries in agent.whitelist are now a list of objects with id, match_type and match_value, replacing the previous grouped-value structure. Custom event consumers need to be updated.
    • Custom tools: Defaults are validated against their declared input type when saving. The custom-tool HTTP client no longer stores or sends cookies. Error diagnostics retain useful status information with configured credentials masked.
    • Models, access and operations: New agents default to gpt-6-luna; existing model selections are preserved. Access checks for deactivated accounts and administrative impersonation have been tightened. Reused database connections reduce connection overhead; new readiness checks account for dependent services.
  2. Backend v1.0.0 · Frontend v1.0.3

    Interactive dashboard analysis, structured responses, and clearer chats and file management.

    • Dashboard
    • Agents
    • Chat
    • Knowledge search
    • Files
    • Inbox
    • WhatsApp

    What’s new in October:

    • Explore message trends in detail: Compare daily messages with the previous period in the dashboard and switch between area, bar, and table views. Drag across the chart to zoom into a date range, or use the + and − keys. Reset the zoom at any time. Copy the visible table values, including comparison data when enabled, and paste them into a spreadsheet. Tooltips show values for each day; the platform chart shows conversation counts and each channel’s share.

    • Structured agent responses: agents with an output schema configured through the API return structured results, displayed in chat as tables with nested fields. Data types are preserved when copying, exporting, and reloading results. When required information is missing, the agent can ask follow-up questions. Output schemas are currently configured through the API.

    • More targeted knowledge search: agents with knowledge search can list their assigned files and open relevant sources without knowing the filename beforehand. Text passages are kept complete. A notice shows when some results cannot be included. In the settings, you can choose how much text the search may return. Found files and processing problems are easier to recognize in chat.

    • More reliable chats: messages and available tool results are preserved more reliably when responses fail. Attachments remain selected when switching agents. Chat titles refresh after every five user messages; until the first title is generated, the user’s message serves as a temporary title.

    • A clearer inbox: compact entries show the agent, contact, and labels with colored dots. Filters and exclusions are easier to recognize. Loaded pages remain visible even when filtered lists stay open for longer. Use Shift to select a range and Ctrl or Cmd to add or remove individual selections, including in other lists such as files, labels, and Office cases.

      LIVOI inbox with example chats and German labels for Sales, Follow-up, and Order

    • Easier file management: sort file lists across all pages and filter by multiple agents. Previews load on demand, and large image attachments display more smoothly. Generated file titles, tags, and knowledge-processing status update in real time. Large-file processing and recovery after worker failures have been improved. When an upload is rejected because of duplicate files, a notice names the affected files and explains how to correct the selection.

    • Clearer deletion feedback: Chats, files, labels, prompts, and API keys stay visible but dimmed and disabled while deletion is pending. They disappear after server confirmation; if deletion fails, they become available again. Compact status messages at the bottom right remain visible above the file-processing panel and match the dark theme.

    • More flexible WhatsApp messaging: Select announcement recipients from existing contacts or add phone numbers directly, including German numbers with a leading zero and international numbers with a country code. Templates appear as messages in the chat, with send status and delivery checkmarks making the history easier to follow.

    • Edit roles and groups directly: Change a user’s roles and groups directly in the member list. Use the three-dot menu next to a member to assign or remove roles and groups without leaving the page.

      LIVOI member list with example users and an open menu for selecting roles and groups in German
    Technical improvements
    • Loading and navigation: The initial JavaScript payload has been reduced. Server-side prefetching makes data available earlier; menu routes are prefetched on hover or keyboard focus. The current page remains visible until the next page is ready.
    • Dashboard and tenant selection: The selected dashboard date range is persisted. Deep links can select the tenant context when opened.
    • API keys: Management supports paginated search, configurable page sizes, and expandable descriptions.
    • Whitelist synchronization: Agent whitelist changes are sent to connected clients through WebSocket events and applied without a manual reload.
    • Tool configuration: Resetting tool descriptions uses the current defaults from the backend. An answer-review notice explains that responses appear in full after the review finishes, and that additional model calls can increase waiting time and token costs.
    • Sign-in and platform attribution: Session tokens preserve the platform context so chats remain attributed to the correct channel.
    • Permissions and translations: Role-based access checks account for permissions scoped to a user’s own chats and personal resources. Missing translations in German, English, Italian, and Polish have been added.

  3. Stored files in chat, animated avatars, better previews, and Italian and Polish support.

    Key improvements since the end of July, including refinements through 15 September:

    • Use files directly in chat: select stored files from the attachment menu or with @. File and agent search has improved.
    • Clearer chats: animated agent avatars, compact tool calls, and more reliable conversation histories. Agent changes synchronize automatically.
    • Better file previews: view HTML attachments directly, open large CSV files more smoothly, and track file processing in real time.
    • Easier navigation: open submenus with the sidebar collapsed and switch tenants without a full reload. Charts adapt better to light and dark modes.
    • New languages: Italian and Polish are now available.
    • Further improvements: more stable uploads, calendar editing, and sign-in, plus improved tool configuration and error messages.

  4. Real-time synchronization, editable tool descriptions in expert mode, a bilingual dashboard, and everyday improvements.

    Faster synchronization. Clearer origins. Two languages.

    The inbox and chat increasingly update in real time, dashboard and API chats are clearly distinguished, and the interface is now available in German and English.

    Tool calls are more transparent, and expert mode lets you adjust tool descriptions to guide agent behavior. Label assignment is more reliable, and WhatsApp QR codes are available in the dashboard. The new connectome offers an early view of LIVOI’s connected future.

    Real-time inbox and chat

    Chat lists, histories, and responses in progress increasingly synchronize in real time. New messages, changes, and response status appear sooner without manually reloading. WebSocket support is gradually expanding to other platform areas.

    Distinguish dashboard and API chats

    New chats started through the LIVOI dashboard at app.livoi.de receive the origin “LIVOI Chat”, instead of being incorrectly classified as API chats. Existing chats remain unchanged. The “API” filter remains and continues to show chats actually created through the LIVOI API.

    Dashboard in German and English

    Switch the dashboard language between German and English in the profile menu at the bottom left.

    Language selection in the LIVOI dashboard profile menu

    Select German or English directly in the profile menu.

    Clearer tool calls

    Agent tool use is easier to follow in the chat. Call data and results are displayed separately. Custom tools now also show the supplied input or request data, rather than only the response.

    Edit tool descriptions in expert mode

    Edit a tool’s description directly in the agent configuration. The description tells the agent what the tool is for and when to use it. This is particularly useful when an agent does not reliably use an existing tool: explicitly describe the tasks or requests that should trigger it.

    Editable tool description in an agent’s knowledge search configuration

    Adapt the tool description to its intended use in expert mode.

    More reliable automatic labels

    Agents now check for suitable chat labels on every message, rather than periodically. The required system tools are included with every agent and currently cannot be removed or edited.

    WhatsApp QR codes in the dashboard

    Display and use QR codes for connected WhatsApp numbers directly in LIVOI.

    Action for showing a WhatsApp QR code in the dashboard

    Find the QR code in the actions for a connected WhatsApp number.

    A first look at the LIVOI connectome

    The new connectome illustrates possible relationships between agents and knowledge. The current view is a demo viewer offering an early look at LIVOI’s future development.

    LIVOI connectome demo with connected knowledge and agent nodes

    The demo viewer visualizes relationships between agents, information, and knowledge.

    Technical improvements
    • Expanded real-time communication: chat lists, histories, changes, and streaming status synchronize through WebSockets.
    • Modernized API structure: most REST endpoints are registered, validated, and documented through a shared technical structure.
    • Consistent agent configuration: models, instructions, and runtime settings share a configuration foundation. Tools use a consistent runtime and cache structure.
    • Reworked chat history and file storage: message content, tool calls, attachments, and usage data are processed more consistently and structurally. Files and attachments share a storage foundation.
    • Optimized authentication: JWT authentication needs fewer database queries. WebSocket connections can authenticate securely with bearer tokens or existing JWT sessions.
    • Improved platform stability: database connections are limited according to the system process. Database session lifetimes and asynchronous background and worker processes have also been optimized.

  5. Chat file uploads, file processing status, and clearer RBAC management with groups and roles.

    Use files more easily. Manage access more clearly.

    Files can now be uploaded in the chat, and processing for the RAG pipeline is more visible. LIVOI also introduces role-based access control: roles and groups help organize access transparently.

    Read the RBAC technical deep dive →

    File processing status window

    After uploads, a compact status window shows whether processing succeeded or individual files encountered errors. Processing includes steps that prepare files for the RAG pipeline, such as chunking and embedding. The status window shows whether files have finished processing or need attention.

    Upload files directly in the chat

    Attach and upload files in the LIVOI frontend chat to bring documents into ongoing conversations without switching context. Add files directly from the chat.

    Groups in RBAC

    Groups collect users and can receive their own roles and permissions in the RBAC system. Represent team structures centrally instead of maintaining access separately for each person. Users added to a group receive its permissions within the relevant tenant context. Groups simplify permission management for multiple users.

    Roles in RBAC

    Roles combine specific permissions and define which actions are allowed, such as reading, creating, editing, or deleting resources. Assign roles to users or groups to separate administration, operational use, and restricted access. Roles define which permissions users or groups receive.

    RBAC at a glance

    RBAC evaluates permissions in the tenant context and can limit access to specific resources. Resource grants selectively share individual items without expanding global roles. API keys can also have their own permissions, constrained by their owner’s allowed rights.

  6. Prompt library, HTML chat rendering, inbox WhatsApp templates, and custom API integrations for agents.

    Reuse your work. Format answers flexibly. Connect your APIs.

    Prompts can now be reused in an organized library, chat responses can render HTML, WhatsApp templates are available in the inbox, and custom API integrations for agents can be created and edited in the frontend.

    Prompt library in the chat

    The chat window at /chat now has a prompt library at the bottom of the input area, keeping recurring prompts close at hand.

    • Create folders to organize prompts.
    • Prepare prompts with variables for later reuse.
    • Use recurring inputs faster and more consistently. The prompt library is available directly in the chat, with folders, variables, and reusable prompts.

    Create and edit custom API integrations for agents

    Create and edit custom tools under Agents → Tools in the frontend. These interfaces let LIVOI use external systems and APIs, for example to retrieve CRM data, start internal processes, or include business application data in an answer.

    Provide individual functions directly as agent tools and manage them more easily. Enable expert mode under Agent → Settings to edit them. Create and adjust custom API integrations in expert mode directly in the frontend.

    HTML rendering in chat

    The LIVOI frontend now supports HTML in chat responses, including tables, formatted sections, and embedded visual elements.

    • Clearer, more readable responses.
    • Complex content displayed within the conversation.
    • Fewer switches to external views for structured output.

    Send WhatsApp templates from the inbox

    Send templates directly to users from an inbox conversation.

    • Prepared templates available in context.
    • Faster sending of standardized messages.
    • Less switching between inbox and template management.
    Technical changes
    • gemini-embedding-2 support: the model is technically supported; use currently requires coordination with P-CATION.
    • Expanded RAG tool logs: additional information when using gemini-embedding-2.
    • Updated error messages: detailed errors now appear in dialogs instead of toast notifications.

  7. Profile pictures, file previews, and stability improvements for LIVOI.

    Recognize agents faster. Inspect files directly. Switch tools less.

    This release improves two everyday tasks: managing agents visually and reviewing files directly in LIVOI. Technical improvements also make recurring workflows more reliable.

    Custom agent profile pictures

    Agents can now have profile pictures, making them easier to distinguish in lists, detail views, and workflows.

    • Upload an image or choose one from the gallery.
    • Adjust the crop when setting the image.
    • Reset or delete existing pictures. Select, crop, zoom, and save a profile picture directly to an agent.

    View files directly in LIVOI

    Open files from the file list without downloading them or finding them in another application.

    • PDFs appear in a side preview.
    • Text and Markdown files can also be read directly.
    • The preview stays within file management, making uploads, agent assignments, and downloads easier to review. Preview files without leaving your workspace.
    Clearer, more visual agent management
    • Set and crop profile pictures from uploads or the gallery.
    • Changes are saved to the agent immediately and displayed in the interface.
    Review files without switching context
    • Open PDFs and text files directly in the platform.
    • The file list remains in the background, preserving your context.
    Technical stability improvements
    • Further improvements to file previews, upload processing, and presigned URL access.
    • More robust, consistent background processes and tool configurations.
    • More stable uploads, file processing, and error handling.

  8. Bulk Dynamic Query, WhatsApp templates, slash commands, and greater chat transparency.

    A faster backend. A clearer platform.

    Slash commands and additional chat details are now available directly in the platform.

    Create WhatsApp templates in LIVOI

    Create templates for recurring WhatsApp communication directly in LIVOI without external tools or preparatory API steps.

    WhatsApp templates in LIVOI

    Create and manage WhatsApp templates directly in the platform.

    Slash commands for faster workflows

    Start actions and enter information directly from your working context.

    • Trigger common actions with /.
    • Reach standardized inputs and workflows faster.
    • Shorter, more direct interactions in chats and agent workflows.
    Slash commands in LIVOI

    Open slash commands in the input field and use them within the chat context.

    More information directly in the chat

    Chats now display additional technical context, including model, agent, and token usage. This makes responses easier to trace and supports analysis, debugging, and quality assurance directly in the chat.

    Chat details showing model, agent, and token usage

    The new detail view shows the response source and token consumption.

    New agent list view

    A compact list view makes larger collections of agents easier to manage and compare. Names and entries are quicker to scan than in the tile view.

    New list view for managing agents

    A denser, easier-to-scan view of your agents.

    Reorganized sidebar

    Important areas are grouped more clearly and are easier to reach.

    Reorganized LIVOI sidebar
    Expanded and faster Dynamic Query

    Dynamic Query has received major functional and technical improvements:

    • Bulk Dynamic Query: process multiple dynamic queries in one request.
    • Better performance: faster queries reduce waits for complex data retrieval.
    • Better scaling: a more robust foundation for data-intensive workflows and higher query volumes.
    Expanded foundation for dynamic tools
    • Dynamic tool foundation: generate and validate input models from schemas at runtime.
    • Expanded tool execution: technical groundwork for dynamic HTTP requests and API key management.
    • Broader agent capabilities: new foundations for production tool and agent workflows.

  9. More stable chats, a faster inbox, and better controls on desktop and mobile.

    More stability. More speed. Less friction.

    This release focuses on the essentials: more stable chats, a faster inbox, and fewer usability obstacles. Technical improvements also provide a more robust basis for integrations and future features.

    Improved chat

    The chat responds more reliably and consistently.

    • Messages no longer reload or appear twice.
    • Audio input supports ogg, wav, and mpeg.
    • Improved mobile controls and shortcuts.
    • Responses automatically use the language of the latest input.
    Audio input in the chat

    Simpler agent creation

    The interface for creating and editing agents has been simplified for greater clarity and focus.

    Simplified agent creation interface

    Improved navigation and settings

    • Deep links after sign-in: chat links return you directly to the correct conversation after authentication.
    • Updated /settings: maintain employees’ first and last names.
    • Correct name updates: changes are processed through the user endpoint.
    • Consistent API keys: clear ADMIN/ADMINISTRATOR labeling.
    • Search fixes: corrected agent and file search errors.
    • Integrations: added logos.
    Performance, caching, and async processing
    • Standardized Redis cache reads, writes, and deletes.
    • Stabilized default cache logic.
    • Optimized Celery tasks and asynchronous processing.
    • Reduced sources of errors in transcription and data processing.
    Query logic and data model
    • Further stabilized Dynamic Query subqueries, aliases, and aggregations.
    • Reworked column and path resolution.
    • Moved statement caching and logging to precompiled statements.
    • Improved user management updates, including correct handling of None and empty values.

    Foundation for future features

    Structural improvements to the data model and API processes lay the groundwork for more stable integrations, better multi-tenancy, and future extensions.

  10. A more focused inbox, new filters and labels, and utility agents.

    Targeted search. Better organization. Utility agents.

    This release turns the inbox into a precision tool with powerful filters, color-coded organization, and context menus that speed up your workflows.

    The redesigned LIVOI inbox

    Inbox: find what you need

    We have expanded filtering and improved the layout to help you stay on top of high message volumes.

    • Custom time periods: alongside Today, Yesterday, and This week, choose Last month, This year, Last year, or define a custom range with the date picker.
    • Combined filters: narrow results by combining platform, date, agents, and labels.
    New inbox filter options
    • WhatsApp identification: the WhatsApp display name now appears after the phone number in the chat list. It is also available under chat attributes in the right sidebar of an open conversation.
    WhatsApp display names in the inbox

    Labels: organized communication

    New label management supports organization-wide chat categories.

    • Full customization: create labels with names, titles, descriptions, and color codes to highlight topics.
    • Automatic classification: LIVOI checks labels every 5 messages and assigns matching labels, even if the chat already has others. The AI uses each label’s description to decide, so specify exactly when it should apply.
    • Flexible assignment: a chat can have multiple labels, and a label can apply to any number of chats.
    • Direct filtering: use labels in the inbox sidebar as quick entry points.
    Managing chat labels in LIVOI settings

    Faster workflows with right-click menus

    Actions previously requiring multiple clicks or a view change are now available through context menus.

    • In the inbox: right-click a chat to delete it, download its history, or assign labels.
    • In a chat: right-click while reading messages to access actions such as downloading the chat.
    Right-click context menu in a LIVOI chat

    Integrations and system

    Utility agents: background specialists. Choose the appropriate LLM for each routine task instead of using an expensive model for everything. Create dedicated utility agents for:

    • Chat summaries: automatic conversation summaries.
    • Chat titles: intelligent conversation naming.
    • File management: automatic titles and tags for uploaded documents.
    Creating utility agents in LIVOI settings

    DATEV integration: preparing the connection. The foundation for automated accounting is in place. You will soon be able to connect LIVOI to DATEV Unternehmen Online. Future updates will introduce agent tools that use this connection, such as AI-assisted electronic invoice creation.

    DATEV connection in LIVOI integrations

    Administration and analysis

    • Dashboard analysis: a manual date picker lets you analyze precisely defined periods beyond fixed intervals.
    • Documented API keys: add comments when creating a key so its purpose is visible in settings.

  11. Expanded dashboard visualizations, more WhatsApp controls, and everyday improvements.

    A better overview. More control. Less friction.

    This update focuses on visible frontend improvements. New chart types make dashboard data clearer, WhatsApp management is more flexible, and smaller usability and performance improvements make everyday work smoother.

    • Dashboard: flexible trends and distributions. We have reworked visualizations to make KPIs easier to understand. Switch message trends between an area chart and a bar chart depending on your analysis. A new pie chart shows each communication channel’s percentage of platform activity, such as WhatsApp versus the frontend.
    Message trends as an area chart and platform activity as a pie chart
    Message trends as a bar chart and platform activity as a pie chart
    • Re-register WhatsApp numbers. Re-register numbers directly in the frontend when technical or external issues have disrupted their connection.
    • Edit the Meta Business profile in the frontend. Update the entire profile without opening the Meta Business portal.
    Editing a Meta Business profile in LIVOI
    • Predefined summary templates. Select additional instructions for sales, development, or purchasing with one click. AI instructions are filled in automatically, giving your summary the right perspective without typing.
    Selecting predefined instructions for chat summaries
    • Usability improvements. Alongside a calmer interface, error notifications now include a copy button. Copy the technical error message and send it to P-CATION, giving support the context needed to solve issues faster.
    Copyable error notifications for faster support
    • Performance improvements. Improved pagination and optimized list views speed up the dashboard, especially with larger datasets.
    Backend: new features
    • More stable agent and HTTP communication: a new internal HTTP client improves timeouts, pooling, and logging.
    • Expanded Dynamic Query endpoint: stronger filters, subqueries, AND/OR logic, and significantly better performance.
    • Improved WhatsApp integration: more robust profile handling, stable connections, and flexible upload and media flows.
    Bug fixes and stability
    • Consistent error handling with clearer validation and API errors.
    • Fixed rare duplicate IDs in worker and RAG processes.
    • Stabilized agent caching, file handling, and database operations.

  12. More transparency, control, and precision.

    More insight. More control. More possibilities.

    This update introduces agent tool inspection and detailed agent configuration in expert mode. Interface improvements, a summary reliability fix, and smaller optimizations also make the dashboard more comfortable to use.

    • Inspect agent tools in the inbox and chat. Each agent response now shows which tools were used, including the parameters and steps performed. This makes agent behavior traceable and debugging easier.
    Tool activity in a LIVOI agent conversation
    • Agent configuration in expert mode. Configure each tool individually, from advanced knowledge search parameters to allowed domains for web search.

    Example: knowledge search. Fine-tune ranking strategies, context windows, result limits, fuzzy matching, MMR settings, and more. Configure precise or exploratory search strategies to suit the use case.

    Tool configuration in expert mode

    How to enable expert mode: use the switch at the bottom of the agent detail view.

    Expert mode switch in the agent detail view
    • Domain controls for web search. Restrict searches to specified domains for precise, controlled, industry-specific results.
    • Interface improvements. Visual refinements make the dashboard clearer, calmer, and easier to read.
    New features
    • Agent tool inspection: see exactly which tools an agent used.
    • Agent expert mode: configure each tool in detail.
    • Web search domain controls: define allowed domains for research.
    Bug fixes
    • Chat summaries now work more reliably.

  13. Small changes with a big impact.

    Easier navigation. Clearer work. Faster responses.

    • Custom instructions for chat summaries. Add your own instructions when summarizing a conversation.
    Custom instructions for a LIVOI chat summary
    • New sidebar. Modernized, organized, and easier to navigate.
    • Global search (Ctrl / Cmd + K). Find content, chats, and settings across the system.
    • Performance and interface refinements. Smoother workflows, more stable behavior, and visual improvements.
    New features
    • Custom instructions for chat summaries.
    • Global search using Ctrl / Cmd + K.
    • Redesigned sidebar.
    Improvements and fixes
    • Faster responses in several areas.
    • Various minor bugs fixed.
    • Small interface refinements.

  14. A new chapter for communication. Faster, clearer, and easier to use.

    A new experience. A clearer focus. A faster LIVOI.

    Experience the next generation of LIVOI with a clear design, intuitive controls, and improved performance. We have rethought the tools you use every day: inbox, chat, agents, and performance.

    • The new inbox. Clearer, more intuitive, and easier to navigate. Previously called “Messages,” it is now the center of your workflow, redesigned for clarity and speed.
    The redesigned LIVOI inbox

    • Chat directly in the dashboard. Immediate, personal communication with your team or support, without switching tabs or interrupting your work.

    • Chat summaries. LIVOI automatically summarizes conversations so you can understand them in seconds rather than minutes.

    LIVOI chat summary feature

    • One click for all your data. Export your complete tenant data with a single button.

    • Your agent goes online. The new Web search tool lets your agent research current information on the web independently.

    • Background processing. A new task engine powers faster uploads and smoother processes.

    • Small changes with a big impact. Dozens of fixes, shorter loading times, and improved stability make everyday use smoother.

    New features
    • Redesigned inbox: a fresh design for more focus and fewer clicks.
    • Dashboard chat: communicate directly where you work.
    • AI summaries: see a chat’s key information at a glance.
    • Tenant data export: easier access to your own data.
    • Agent web search: research the web and use current information in answers.
    Technical improvements
    • Background task queue: a faster, more responsive system.
    • Optimized API structure: stable, consistent, and ready for future development.
    • Asynchronous file processing: parallel uploads for better performance.
    • Improved error handling: more precise, clearer, and more robust in daily operation.
    Bug fixes and performance
    • Fixed errors in the message view.
    • More stable chat histories, including on weak connections.
    • Faster file uploads and shorter loading times.
    • Improved chat title generation.
    • General performance improvements for a smoother experience.