# Changelog ## 2026 ## 8/2026 'Fugue'
Release notes #### Features and improvements: * **Enhanced workflow builder in Automations:** The side menu inside workflows has been redesigned for a cleaner, more intuitive experience. Nodes now show more helpful details at a glance, like which AI model is in use, and you can click these details to jump straight to their settings. * **Dixa channel integration improvements:** We’ve deepened our Dixa channel integration, allowing you to sync bot messages as internal notes, apply bot tags to conversations, and check agent availability for specific queues. Additionally, customer ratings collected in Ebbot can now be sent directly to Dixa as CSAT scores for the specific agent involved. * **AI insights updates:** We've added a dedicated column for language in your insights dashboard and improved the data collection for closed chats to give you even more accurate reporting. * **New platform language:** The Ebbot Chat platform is now available in Italian. #### Bug fixes: * **Chat widget and Live chat:** We’ve polished the chat experience by fixing source numbering in the widget, correcting the timing logic for inactive chats, and resolving a limitation that prevented mixing headers and bullet points in live chat messages. * **EbbotGPT Knowledge:** Fixed issues where Markdown files could fail to upload as a knowledge source, and where selected fields in TOPdesk knowledge base scrapes were not always included in the fetched content. * **AI insights:** Fixed an issue with incorrect error message and pie chart readability in dark mode. * **Conversation starters:** Fixed an issue where opening the same chat in multiple tabs could trigger conversation starters multiple times. * **Dixa channel integration:** We’ve resolved issues regarding queue availability, ensured chats transferred back to the bot close correctly, and removed incorrect notifications. * **Chat history:** Chat history date filters are now working properly. * **Zendesk channel integration naming:** Resolved naming bugs in Dashboards and Chat history related to Zendesk channel integration. * **Workflow nodes in Automations:** Fixed a bug in the workflow builder where deleting a node would occasionally leave stale connections, which could break sorting and affect published flows.
## 7/2026 'Interval'
Release notes #### Features and improvements: * **Persona markdown support:** We’ve upgraded the Persona editor in EbbotGPT Configurations to support Markdown. This makes it significantly easier to visualize and style the instructions that define your chat agent’s persona. * **Smarter Conversation starters:** You can now prevent your chat widget's Conversation starters from appearing if an end user is already in an active chat or speaking with a human agent, ensuring a seamless experience without interruptions. * **Improved UI for advanced model settings:** We’ve overhauled the model settings UI in EbbotGPT Configurations by removing obsolete parameters and adjusting the valid ranges for top\_p and top\_k to ensure optimal performance. * **Reliable branching:** We've improved how data moves through complex workflows in Automations, ensuring that automated steps remain stable even when a conversation follows a specific branch or logic path. #### Bug fixes: * **Automations:** Resolved template publishing errors, fixed "node addition" issues in the builder, and improved the template modal UI. * **Dixa integration:** Fixed bot interference in live chats, improved agent status accuracy, and stabilized queue routing and assignments. * **AI Insights:** Corrected data display for languages and topics, and ensured chat counts are now consistent across all dashboards. * **Chat widget:** Improved citation mapping accuracy, enabled instant styling previews, and resolved issues where previous chat states could "bleed" into new sessions.
## 6/2026 'Mezzo'
Release notes #### Features and improvements: * **Automations – Comment nodes in workflow:** Stay organized with the newly added Comment functionality when building flows in Automations. You can now add comments directly into your workflows to document logic or leave reminders. Find Comment under Logic within a workflow, along with LLM, Condition and Transformer. * **Automations – Improved log readability:** We’ve made some changes to make the workflow logs easier to digest. * **Expandable logs:** Log entries now have a maximum line limit to prevent clutter. You can use the new manual "expand" toggle to dive into the details when needed. * **Role display:** The LLM message role (User, Assistant, System) has been moved to appear directly above the message instead of off to the side, improving the view when building flows. * **Safeguards – UI enhancements:** We’ve refined the Safeguard interface to be more intuitive. This includes better categorization and indented dropdowns, making it easier to visualize and manage your specific safeguard settings. * **AI Insights – Refined "Unsolved" categorization:** We have tuned the prompt for our Unsolved metric. This ensures that responses such as "I don't have the answer to that" are now accurately categorized as unsolved, giving you more reliable performance data. #### Bug fixes: * **Automations fixes:** We’ve resolved friction points in Automations to ensure a smoother experience. Including a fix for the intermittent builder freeze on load, and an issue where workflows would appear off-center. We also polished the UI by fixing variable overflows in nodes, standardizing token usage number formatting, and correcting scrollbar styling for light mode. * **Dashboards fix:** Fixed an issue where platform users with access to more than 20 bots were unable to access their Dashboards in the Chat platform. * **Suggested topics fix:** Fix to ensure that metrics for the same topic are correctly merged regardless of capitalization. * **Chat history UI bug:** Fixed an issue in the Chat history UI – the Search dataset tool no longer requires horizontal scrolling. * **EbbotGPT API fix:** Resolved a "400 Invalid JSON payload" error specifically affecting Google models when used over the API.
## 5/2026 'Scale'
Release notes #### Features and improvements * **Automations – Email agent launch:** As a first use case in our new agentic workflow builder, we’re introducing automated handling of email/ticket cases. This includes drafting and sending replies, categorizing, and routing them correctly. Navigate to Automations in the platform to learn more (Note: Available only on [ebbot.eu](http://ebbot.eu)). You can also explore [Email agent](https://www.ebbot.com/platform/email-agent) on our website or why not deep dive into the [documentation](https://docs.ebbot.ai/ebbot-docs/automations/automations) straight away. * **AI Insights – language tracking:** Introducing language tracking, giving a clear breakdown of the different languages used in conversations. * **AI Insights – improved view:** The Resolution Rate charts are updated and metrics like "Incomplete" and "No data" are now grouped under "Unsolved" (displayed in blue) to give you a cleaner, more accurate view of your performance. * **AI Insights – Suggested Topics:** Support for Suggested Topics, improving the categorization logic in AI Insights. #### Bugfixes * **Promptguard scenarios:** Fixed a bug where Promptguard would trigger a catch-all scenario instead of the defined one when using "next" pointers. * **PDF import fix:** Resolved an issue where certain PDFs failed to import due to font compatibility. We’ve overhauled our scraper to be much more resilient. * **Google Translate formatting:** Bot messages in Chat History will now maintain their formatting (bolding, bullets) even when translated.
## 4/2026 'Modulation'
Release notes #### Features and improvements * **Search-and-replace functionality in EbbotGPT:** We’ve added Simple Transformation Settings capability to source filtering. This allows you to standardize terminology across your sources. For example, you can replace “chatbot” with “AI agent” so that the chat agent (chatbot 🙂) consistently uses your preferred naming in generated responses. Read more [here](https://docs.ebbot.ai/ebbot-docs/release-notes/broken-reference). * **Live chat translation improvements:** Live chat translations have been improved so that all messages are now translated across chat history, live chat, and the widget. * **Gemini 3 Flash now available:** Compared to version 2.5, this release delivers improved accuracy and reasoning, allowing customer inquiries to be handled with greater precision. You will also notice faster response times, helping you resolve issues more efficiently. To learn more about the specific improvements, refer to the [official documentation](https://blog.google/products-and-platforms/products/gemini/gemini-3-flash/). If you would like to try it out, feel free to reach out! #### Bugfixes * **Clicked URL buttons in chats:** We fixed the “Solved” metric logic in AI Insights. Chats are no longer marked as “Unsolved” simply because a user clicked a URL button and stopped interacting with the chat agent. * **AI Insights download modal:** Fixed a UI issue where the download insights modal would persist and follow the user across different chat views until a restart. * **Easit GO knowledge source:** Fixed a bug that prevented certain content from being fetched from Easit GO due to an unexpected API response format. * **Markdown knowledge source:** Fixed an issue that prevented markdown files from being uploaded correctly.
## 3/2026 'Crescendo'
Release notes #### Features and improvements * **Accessibility improvements in widget:** We’ve updated the widget to improve accessibility (WCAG) and structural SEO. These changes ensure a smoother experience for users navigating via screen readers or keyboards.\ \ \- **Semantic headings:** Markdown now renders using real HTML heading tags ($h1$–$h6$) for better document structure and SEO.\ \- **Enhanced navigation:** Interactive cards now function as direct links, and the "Powered By" section is fully accessible via keyboard.\ \- **Screen reader optimization:** Switched from title attributes to aria-label for clearer element descriptions. Removed redundant alt text on the start page to reduce repetitive audio announcements.
* **AI Tools – HTTP Request Tool:** Connect your chat agents to your existing tech stack. This feature allows you to define preconfigured requests, such as product search, that your agent can trigger to fetch live data from your own APIs or third-party services.\ \ \- **Real-time integration:** Seamlessly pull data like product availability, pricing, etc. into the chat.\ \- **Dynamic placeholders:** Define custom arguments that the chat agent understands and fills in automatically based on the end user's request.\ \- **Customized responses:** Use "response paths" to filter out technical noise and ensure the AI only receives the specific information it needs to display.\ \- **Example**: If a customer asks for "black jeans”, the chat agent can query your product catalog in real-time and present a curated list of available items with images and "Buy Now" links directly in the widget. * **Content Guard improvements:** We’ve upgraded the underlying engine for the Safe Guard, Content Guard. The new model significantly outperforms previous benchmarks, delivering higher accuracy in detecting harmful content while reducing false positives. #### Bugfixes With this release, we have resolved bugs within the platform related to chat widget styling, scenario management, and bot list navigation.
## 2/2026 'Motif'
Release notes #### Features and improvements * **Filtering TOPdesk Knowledge Base using EbbotGPT API** – When using the EbbotGPT API with the TOPdesk Knowledge Base as a knowledge source, you can now filter on all available query parameters from the TOPdesk API, such as language, status, and more. * **AI Insights loads faster** – AI Insights have been optimized, resulting in faster data generation. * **AI Insights Resolution metrics update** - Based on user feedback, we have streamlined these metrics to include only Solved and Unsolved statuses. #### Bugfixes * **Fix for Get Website tool** – Fixed an issue where fetching an empty page could return an empty string, causing gateway validation to fail and interrupt the workflow. The tool now handles empty pages properly without crashing. * **Fix for Search Google tool** – Improved reliability and increased the search limit, fixing occasional failures when fetching company content. * **Prompt Guard triggered when clicking button** – Fixed an issue where button clicks in the widget (sent as XML tags) were incorrectly triggering Prompt Guard. Button presses will no longer be blocked.
## 1/2026 'Ensemble'
Release notes #### Features and improvements * **Improved document splitting for knowledge sources**\ Knowledge sources now correctly split content on Heading 5 (H5) and Heading 6 (H6), in addition to Headings 1-4. This ensures that all content is captured and structured properly, even when deeper heading levels are used. This improves data quality and completeness. * **GPT Evaluation now available to all**\ GPT Evaluation is now available to everyone with EbbotGPT enabled on their account. This feature lets you evaluate and improve the quality of AI-generated responses. To try it out, navigate to EbbotGPT > GPT Evaluation in the platform. You can read more about how it works in our [documentation](https://docs.ebbot.ai/ebbot-docs/ebbotgpt/gpt-evaluation). #### Bugfixes * **TOPdesk forms now update in real time** We fixed an issue where pre-populated powerup and incident attribute forms for TOPdesk users were no longer updating dynamically. The forms now reflect changes in real time again, making it easier to validate data and flows. * **Multi-file upload in chat widget stabilized** An intermittent issue causing the multi-file upload component in the chat widget to fail with an “Invalid token” error has been resolved. Uploads now work reliably across files and formats. * **Chat export with customer info no longer fails** Exporting chats with the “Include customer info” option could previously fail in certain edge cases. This has been fixed, and chat exports now complete successfully in all scenarios. * **Identity provider domains now generated correctly** We resolved an issue where configuring an identity provider did not generate the expected login domain. Domains are now created correctly, ensuring SSO login works as intended.
## 2025 ## 17/2025 'Salt'
Release notes #### New features and improvements * **Ability to scrape PDFs via the SharePoint API**: The platform can now retrieve and process PDF files directly from SharePoint via the API. This enables bots to automatically access and scrape content from multiple PDFs without requiring manual uploads, significantly reducing maintenance overhead for customers. * **New language instructions in prompt**: To address issues with LLMs responding in the wrong language, a new prompt configuration ensures that the model always answers in the correct language. This applies both to the initial user message and to any mid-conversation language switches. Key updates include: The LLM now follows strict language instructions defined in the system prompt, using either a specified language or the language of the last user message. * **Support for rating in Teams**: Teams conversations can now include support ratings directly in the chat flow. * **Knowledge source filter based on keyword in document**: A new filtering feature has been added to EbbotGPT knowledge sources to improve content relevance and control. Users can now filter documents or columns based on specific keywords using conditions like “contains,” “doesn’t contain,” or “starts with.” You can remove individual columns or the entire document if they match the filter criteria. #### Bugfixes * **Exporting files failing on ebbot.eu**: Resolved an issue preventing large file exports for certain customers. Exports were failing with the error: “InvalidAccessKeyId: The AWS Access Key Id you provided does not exist in our records”. * **Chats on V2 not closing**: Fixed an issue where chats marked with type close\_chat were not being processed by the cron job in V2, preventing automatic chat cleanup. The end\_conversation scenario was also not being triggered correctly. * **Get website tool allowed urls**: You no longer need to add an asterisk (\*) to the end of a URL to capture trailing queries or subdirectories. * **Dashboard performance on ebbot.eu degraded**: Resolved a performance issue affecting dashboards on the platform. Customers experienced excessively long data load times, often over one minute compared to the usual 10–15 seconds. * **Bad request – Assert issue on widget SSO**: Fixed an issue causing errors for users logging in via the SSO in widget. After the fix, SSO logins through the widget are now working correctly. * **Dashboard issue on V2, different results based on groupings**: Fixed an issue in V2 dashboards where duration metrics displayed inconsistent values depending on the selected grouping (e.g., day, week, month, or no grouping). * **Insights – no automatic reconnect to database:** Fixed an issue where the Insights service did not automatically reconnect to the database after a connection failure. * **Users have to edit the URL after creating an account**: Fixed an issue where newly created accounts occasionally encountered an “Access Denied” page immediately after setting their password. The problem was caused by an unnecessary portion of the auth callback remaining in the URL. * **Users not created in Ebbot app after SAML login**: Fixed an issue where users logging in via SAML were successfully authenticated but were not automatically created in the Ebbot platform. * **MFA toggle missing for Company Owner**: Fixed an issue on OVH and V2 platforms where the MFA (Two-Factor Authentication) toggle was not visible for Company Owner users in the company settings panel. * **.WebP format not supported for uploading in widget settings**: Fixed an issue where uploading .webp files as logos or avatars in Widget Styling would fail. * **Automatic update in Knowledge sources and datasets not working**: Fixed an issue where automatic updates for sources or datasets for EbbotGPT using the list option would trigger a 400 Bad Request from the API. * **Domains not generated for identity providers configured**: Fixed an issue affecting all platforms where configuring an identity provider in Ebbot did not generate the expected login domain (e.g., customer.ebbot.eu). After this fix, domains are now correctly generated for newly configured identity providers, enabling SAML login to function properly.
## 16/2025 'Hunt'
Release notes #### New features and improvements * **Escalation metric in AI Insights** – Now you can easily understand how many chats have been escalated. The Escalation Rate metric within AI Insights quantifies the percentage of AI-handled conversations that are transferred to a human support agent or result in ticket creation. * **Export data from AI Insights** – It’s now possible to export your AI Insights as a CSV. This feature is beneficial when you want to import your data from AI Insights to your Business Intelligence tools (e.g., Tableau, Power BI) or if you would like to perform your own custom AI analyses. Want to use this feature? Reach out to and we will walk you through it. * **Use EbbotGPT to create a general-purpose assistant** – the new configuration option lets you set up an agent that can support a wide range of needs. Our finetuned models are built for customer service and rely on knowledge, persona settings, and guardrails to ensure a safe and relevant experience for your users. Still, we know that your use cases may vary, even within service. With this update, you can also configure EbbotGPT for more general tasks. A common example is using it internally, just like you would with other digital assistants in your daily work. Let it summarize text or use it to bounce ideas, always with the same level of safety and control you expect from Ebbot. * **Improved filtering on TOPdesk API source** – improving the filtering options for the TOPdesk knowledge source, making it easier to filter knowledge items and select the most relevant ones to use as EbbotGPT knowledge. [Click here](https://docs.ebbot.ai/ebbot-docs/ebbotgpt/knowledge/source-types/topdesk-api#filter-articles-using-query-filters) to read more about how to filter, and if you need any help – reach out to us!
#### Bugfixes * **iFrame bug** – resolved an issue where links clicked within the embedded widget, when loaded inside an iFrame, would open inside the chat window rather than in a browser tab. * **Search dataset breaks if nr of docs is not set with +/- buttons** – fixed an issue where manually typing a value into the “number of results” input field caused the search dataset tool to stop working. * **Starting page showing even when disabled** – resolved a sporadic issue that caused the widget’s starting page to appear on the website despite the display setting being turned off in the platform. * **Custom widget font loaded unnecessarily** – Ebbot no longer proactively loads the Inter-VariableFont file on initial page load when the chat widget is not opened, improving page performance. * **“Show sources” placement on TOPdesk SSPs** – corrected improper rendering and placement of the source window when showing sources from TOPdesk SSP content within the widget. * **Rating conversation starter not working as intended** – fixed an issue where the rating conversation starter did not behave correctly. * **Prompt guard incorrectly detecting prompt injections** – corrected behavior where the prompt guard evaluated tool calls instead of user messages, leading to false detection of prompt injection attempts. * **Button clicks not sent to LLM** – fixed an issue where button click events were prevented from being sent to the LLM, resulting in a catch-all response instead of an intended GPT answer. * **WhatsApp channel: button cards breaking message flow** – resolved an issue where variable-type button cards caused the conversation flow to stop after a button was clicked.
## 15/2025 'Fisher'
Release notes #### New features and improvements * **Unique sessions for conversations through user channels:** Your chats within user channels (Teams, Slack) will no longer be stored as one continuous history. Each new conversation will automatically start as a unique, dedicated session, providing a cleaner separation between topics. * **AI tool ‘search dataset’ new functionality:** You can now customize the tool's search behavior. This includes defining one or more specific datasets for the tool to search and setting the number of documents it should retrieve. Additionally, a new input field allows you to provide a custom prompt that instructs the LLM on exactly how and when to use the tool, including which datasets to search for different topics. * **CSV column limit removed for GPT-source:** You can now import CSV files without worrying about column limits. This is particularly useful for importing complex data files such as product feeds. * **Support for user ratings from all channels (e.g. Teams and Slack):** Your users can now provide valuable feedback on conversations from any channel, such as Teams and Slack. This universal rating support ensures you capture user satisfaction data consistently, no matter where the conversation takes place. #### Bugfixes * Users were unable to save a scenario if it contained an image card. Saving now works correctly. * An internal issue caused by duplicate IDs was breaking the functionality of buttons used within chat or bot responses. Buttons now link and work reliably. * We resolved an issue where complex handoffs to a live agent were failing or not working correctly when using the Microsoft Teams integration. * The export of chat history previously contained unnecessary or placeholder data. The exported files are now clean and only contain relevant chat data. * A bug was causing administrators to randomly lose access or visibility to their managed bots. Admin access is now stable and consistent.
## 14/2025 'Smith'
Release notes #### New features and improvements * **Genesys Cloud integration** – You can now set up Genesys Cloud as an agent channel integration directly in the frontend. Navigate to Settings > Channels > Agent to find Genesys alongside Salesforce, Zendesk, Dixa, and (coming soon) Freshdesk. * **Triggered Content Guard category in Chat History** – Chat History now displays which Content Guard category was triggered in each conversation. This helps with troubleshooting and fine-tuning Content Guard settings to ensure optimal end-user interactions. * **Additional data source support** – You’re now able to upload .md (Markdown) files as data sources in EbbotGPT > Knowledge. * **Improved Conversation Starters view** – We’ve improved the layout for Conversation Starters. You can now see conditions, creation date, and URLs more clearly. This is of course specially useful for those of you who have a bunch of them configured under Chat widget > Conversation starters. #### Bugfixes * **Bot response issue in user channel integrations** – Fixed a bug causing bots to stop responding in user channel integrations (e.g., Teams and Messenger). * **Widget input field stability** – Resolved a bug where the input field in the widget disappeared, one situation being when end users initiated conversations from multiple browser tabs.
## 13/2025 'Gadget'
Release notes #### New features and improvements * **Support for Easit GO as a knowledge source** – Connect to your Easit GO instance and fetch relevant data over the API for your agent/chatbot to use as knowledge. * **Tool for math and calculations** – Introducing a new native AI tool to handle mathematical and calculation-based queries efficiently. #### Bugfixes * **Fix for “Welcome message” scenario** **being triggered when it shouldn’t** – Ensures that the Welcome message isn’t triggered when, for example, clicking a scenario link card on the widget’s starting page. * **Default icon/logo in widget styling settings now works** – Resolved an issue where selecting the default icon/logo had no effect after a custom one had been uploaded. * **Agent availability check on link cards** – In some situations, the live chat agent availability checker was delayed. This fix ensures that the functionality is more responsive, offering an immediate status update when no live chat agents are ready to take on chats.
## 12/2025 'Bourne'
Release notes #### Improvements * **Fetch Website tool now supports JSON, XML and Plain text** in addition to HTML. * **Content guard can now be customized** at the configuration level, reducing overly restrictive blocking for relevant questions. * **GPT evaluation now displays average ratings** for runs, making it easier to compare performance. To rate specific replies in each run, go to GPT Evaluation, click your run and start rating: * **GPT evaluation now shows clear visual feedback when CSV import fails**, improving usability. * **Quick replies can now be navigated using the arrow keys**, speeding up the chat user's workflow. * **Chat history can now be filtered for deleted agents**, ensuring historical data remains accessible after e.g. employee changes. #### Bugfixes * **Image replacements for conversation starters now work** correctly — you can update them without glitches. * Fixed so that some **chat history exports no longer get stuck during execution.**
## 11/2025 'Strawberry'
Release notes #### New features & improvements * **Fresh look for the chat blob** – We've updated the Ebbot chat icon to be more consistent with our brand. We hope you like the new design! * **Keyboard shortcuts in Chat History** – You can now navigate through your Chat History using your keyboard for a faster, more efficient workflow. Note! This is something you can do in the Live chat view as well. * **Editable placeholder text in Conversation Starters** – The "Attention seeker with input" Conversation Starter now allows you to edit the placeholder text in the input field, which previously was uneditable. * **New 'Escalated' metric in AI Insights** – We've added a new metric under the Topic Overview in AI Insights that shows which conversations your Bot/AI Agent can handle on its own and which require a human agent. This will help you better optimize your service flows. We plan to expand this metric to all filters soon. * **End chat button** – There's now a button in the widget header, allowing a quicker way to close a conversation once it has started. #### UI fixes & general improvements * **Better display for Power-ups** – Long power-up names now display correctly in the Live chat view. * **Visible BankID titles** – The titles for BankID are now visible again when configuring. * **Correct notifications for Live chat agents** – When "Auto accept chats" is enabled, only the assigned live chat agent will receive the chat notification. * **Easily identify different Bots in Dashboards** – You can now assign a unique color to each of your Bots, making them easier to spot across your Dashboards. * **The cursor is back** – The cursor no longer disappears from the widget's input field. * **Restored widget message** – The message that informs a user that their chat is being transferred to another live chat agent has been restored, providing a clearer experience for your customers. * **Number input is now working** – We've fixed a bug where number inputs were not working correctly. Scenarios with a number input were broken in the visual view.
## 10/2025 'Solid'
Release notes *New Features & Improvements* * *Dashboards & Insights:* * Added skill-based KPIs (AHT, Waiting Time, Queue Abandonment Rate) and a 'Last 14 Days' view. * Topic-specific data is now available in the Insights dashboard. * A new cost-saving graph has been added to AI Insights. * *Agent Experience:* * Enabled on-demand translation for chat history. * Previous user interactions are now translated for agents at live chat handover. * Added the ability to save different agent resolution statuses for Zendesk conversations. * *EbbotGPT* *Knowledge :* * Introduced automatic, scheduled updates for data sources. [(Read more & request access)](https://docs.ebbot.ai/ebbot-docs/ebbotgpt/knowledge/automatic-update) * Added the ability to stop an active scrape. * *Exports & Data:* * Chat history exports can now include customer information (name, personal number, etc.). * Agent names are now included as a column in chat history exports. * *UI/UX & Enhancements:* * Improved visibility for code blocks in dark theme. * Added an option to mask sensitive data in input cards. * GPT source settings now keep login information visible for easier troubleshooting. * Added a clearer error message when offline configuration generation fails. * Improved how long URLs are displayed in the "show source" section. *Bug Fixes* * Fixed an issue where the Two-factor authentication toggle would not save its state. * Resolved an issue causing Gemini 2.5 to sometimes return empty answers. * Corrected skewed metrics in AI Insights when filtering by a specific agent. * Fixed a bug that showed empty or unknown chats in AI Insights.
## 9/2025 'Gibbs'
Release notes ## 9/Gibbs #### Improvements * Language detection now supports languages that do not use spaces between words, like Japanese. Stability and Logging * Made several stability and logging improvements across our AI services, including the embedder, job-queue, and injection detection services. * Enhanced the EGPT-API with improved request handling, ensuring better traceability and stability across the platform. #### Bug fixes Platform & UI * Addressed critical bugs that prevented changes from being saved correctly for scenarios and offline configurations, which in some cases resulted in a 500 internal server error. * Resolved bugs related to scenario management, ensuring that arranging scenarios no longer incorrectly unpublishes them and that numeric sorting works as expected. * Fixed a number of UI display bugs in dark and light modes, including unreadable text on dataset menus, chat scenario history, and buttons in the widget settings. * Fixed a bug where editing a custom document in EbbotGPT would create a new document instead of saving the changes to the existing one. * The preview for conversation starters is now visible again in widget settings. * Fixed a bug where only the first line of a multi-line placeholder value was rendered in a shared component. * Corrected translation and display issues on the data retention page for Swedish language users. Analytics & Dashboard * Fixed an issue with the "Leaving Scenario Count" metric, which was providing inaccurate data. * The real-time queue count metric now works correctly and no longer includes historical data. Integrations * Fixed a bug that prevented users from tagging conversations. * Addressed intermittent `NameResolutionError` exceptions that were affecting various shared and custom components. * Fixed a bug incorrectly displaying the Live Chat Translation options in the widget. * Resolved an issue where agents were not being created from Genesys Cloud, which resulted in messages not being saved in the Ebbot platform. Conversation Starters * Fixed an issue where the "open" action for a conversation starter would play the welcome message regardless of the widget settings, making it behave inconsistently. * A chat that is ended by a conversation starter will now correctly remain in the agent's view for post-chat processing.
## 8/2025 'Warden'
Release notes **Marketplace in Ebbot** **Coming soon:** A marketplace view within Ebbot where you can browse and request access to Ebbot’s many integrations and extensions. #### **WCAG Updates** * Improved WCAG support * Improved screen reader support #### **Dishwasher Improvements** * The Dishwasher now sanitizes data from both AI-generated and static bot responses, as well as user inputs. * Added user documentation for the Dishwasher. * Improvements to the Dishwasher for more accurate detection. The methods used to detect personal data can be read about here: #### **Bugs & Improvements** * Implemented stability fixes for AI Insights. * **Dashboard:** "Last month" now represents the previous full calendar month instead of the last 30 days. * Fixed a bug where the integrations page would not save settings.
## 7/2025 'Genesis'
Release notes * Support for Genesys Cloud Open messaging due to endpoint deprecation
## 6/2025 'Felicity'
Release notes ### New Features **Platform rebrand & updated UI** * **Updated** logos, icons, typography, and color schemes across the app and chat widget. * **Improved** accessibility in both light and dark mode. * **Refreshed** onboarding elements, feature and product naming conventions and user experience. **EbbotGPT** * **Tool Calling Support:** The platform now supports tool/function calling via API for advanced automation and integration scenarios. * **MCP Server Integration:** Added initial support for MCP server \ \ \&#xNAN;*(MCP and tool calling are currently in beta — let us know if you would like early access or to provide feedback.)* #### Integrations & Channels * **Agent Channels Now at Company Level**: Agent channel connections are now managed at the company level. You can enable/disable connections for each bot and apply bot-specific settings. * **Setup and user experience:** Updated forms make it easy to see which channels are active and which bots are using them, while “Connected” badges and summary views provide a clear, intuitive overview for seamless management. #### Insights & Analytics * **Rating Chart & Chat Table Enhancements**: You can now filter chats by customer rating with a new bar chart and break down each rating category. Ratings and stars are displayed in chat tables and modals. * **Sentiment Analysis**: Now only analyzes user messages, excluding bot replies for more accurate sentiment metrics. * **"No Topic" Filtering**: Find chats without a set topic in the "Most Common Topics" section of AI Insights and filter dropdown. This helps identify gaps in customer inquiry classifications. * **AI Insights - Pre-filter by Current Bot**: AI Insights dashboards are initially filtered by the bot you’re currently working on. * **Chats per Agent AHT Integration**: The table for agent analytics in Dashboard. can now include both the number of chats and average handling time (AHT) in a single view. #### User Experience & Accessibility * **Widget Preview Always Visible**: The widget preview in "Chat Widget > Styling" now stays locked in view as you scroll the page, improving usability. * **WCAG Accessibility Updates**: Enhanced screen reader support and compliance with accessibility standards, addressing feedback from customers regarding message reading, notifications, and magnification. * **Button Metadata in Messages**: Metadata (like IDs and custom classes) can be added to individual messages for improved styling and customization. #### Platform * **Hybrid Search Embedder Default**: All new datasets default to the hybrid embedder for improved search relevance. * **Button History**: Added more information in the conversation object about which buttons a user pressed while in a chat. ### Improvements * **Display Percentages & Amounts**: All filterable charts now show percentages on the card, and detailed amounts on hover. * **Insights Performance**: Resolved slow loading issues with insights dashboards and replaced inefficient caching with a more scalable solution. ### Fixes * **Admins Cannot Edit Insights Config**: Fixed an issue where admin accounts received "403 Forbidden" errors when saving changes to the insight config. * **Charts with Complex Queries**: Improved support for dashboards with large, complex chart queries. * **File Attachments from Messenger**: Files (like PDFs) sent by users through Messenger are now visible and accessible for agents in the platform. * **Agent Names in Analytics**: Resolved missing agent names in analytics in some cases. * **Insights Save After Toggling Off Metrics**: Saving insights config after toggling off a metric now works correctly. * **Agent Channel Transfer Messages**: Transfer system messages are now consistently visible in the chat history. * **UI Crash When Saving Dataset**: Fixed a crash caused by saving datasets with incompatible sources. * **Custom Error Messages for Scrapes**: Should display more helpful error messages on failed scrapes. ***
## 5/2025 'Smart'
Release notes **New Features** * **Introduction of AI Insights:** We are now launching our powerful AI Insights feature for deeper analysis of chat conversations. This comprehensive update includes the deployment of necessary backend infrastructure and data processing, a new user interface via a dashboard, improved functionality around data handling and EGPT. [Click here](https://docs.ebbot.ai/ebbot-docs/release-notes/broken-reference) for more information. #### Improvements & Bug Fixes * **Integrations:** * Improved security to prevent accidental removal of Teams integration secrets. * **Accessibility (WCAG):** Implemented several widget improvements based on customer feedback to increase accessibility according to WCAG guidelines. * **Widget**: * Optimized the widget by removing unused data from the browser's local storage. * Addressed an issue where the widget could stop working after prolonged chatting. * Fixed a bug where welcome messages with multi-select options did not work correctly. * Fixed an issue where the widget could load before custom CSS was applied. * Corrected UI issues related to select components overflowing when many options were selected. * Fixed UI issues that could occur when loading chat style and settings. * **Dashboard & Export:** Resolved an issue preventing dashboard export ("workbook is empty"). * **Chat History Crash:** Fixed an error where the chat history page could crash when reviewing certain conversations. * **Web Scrape Crashing:** Improvements to the web scraper resolve an issue that previously caused crashes during very large scrapes.
## 4/2025 'Scully'
Release notes **New Features** * **Chat History Export**: You can now export chat transcripts directly from the chat history. This makes it easier to keep records of conversations. * **Gemini Integration (Web)**: We've added Gemini as a new option in the web interface, giving you more flexibility in how you interact with Ebbot. Note! Reach out to if you are interested as we will need to enable this model on your account. * **Optional Spam Filter**: The conversation flow now allows for an optional spam filter, giving you more control over your interactions. * **Web Scrape Performance Visualization**: We've added new tools to visualize web scrape performance, helping us to monitor and improve efficiency. * **Sitevision Web Scrape Updates**: Web scrapes now include articles and archives from Sitevision, expanding the range of information Ebbot can access. * **Scenario Builder Enhancement**: We've made it easier to see "Connected Scenarios" within the scenario builder, improving workflow visualization. * **Zendesk Ticket Solving Option (Frontend)**: We've added a frontend option to solve Zendesk tickets directly within Ebbot. * **SSO in Widget Update**: We've changed the location and name of "SAML SSO" to make it easier to find and use. #### Bug Fixes * **Chat Widget Scrolling**: We've fixed an issue where it wasn't possible to scroll to the top of messages in the chat widget's text input field. * **Prompt Date Handling**: We've resolved a bug where the model was confusing days and months in prompts. * **CSS Styling Deletion**: We've added a confirmation dialog when deleting CSS styling, preventing accidental deletions. * **Web Scraper Behavior**: We've improved the web scraper to prevent it from clearing other documents when it returns zero documents. * **Chat Buffering**: We've fixed a bug that caused the chat to buffer when expanding variables with very long strings. * **Welcome Trigger**: We've addressed an issue where the welcome message was being triggered incorrectly after a conversation starter. * **Chat Widget Button Display**: We've corrected an issue where send buttons in the chat widget would disappear with a large font and long message. * **Dataset Document View Bug**: We've resolved a bug in the dataset document view, where disabling a document sometimes disabled another one. * **BankID Card Display Issue**: We've addressed a bug where the BankID card was sometimes not displaying correctly.
## 3/2025 'Powers'
Release notes ### **Zendesk and Ebbot Integration - Update** **Enhanced Requester Management:** * Chats can now be linked to existing or newly created requesters in Zendesk by specifying the `email` field in the Ebbot data object before initiating a handover. * When the email matches an existing requester, the chat will automatically merge, providing a consolidated overview of all interactions with that requester across various support channels. * If no email is provided, Zendesk will create an anonymous requester with a randomly generated name. * To create a new requester with accurate details, ensure that `email` is set in the Ebbot data object. It's also recommended to include `first_name` and `last_name` in the Ebbot `customer` object for completeness. Additionally, the `phone` field can be included to update the requester's phone information. * When a user ends the conversation in Ebbot, the integration will automatically attempt to set the corresponding Zendesk ticket status to `Solved`. If Zendesk has required ticket fields preventing this action, the integration will instead send a notification message indicating that the chat was closed on the users' end. **Integration Setup Form Improvements:** * The integration setup form has been enhanced for better reliability and ease of use. * You can now seamlessly connect or disconnect multiple bots to Zendesk/Sunshine platforms using the updated form. **API Token Requirement:** * The integration requires valid API tokens for both: * Zendesk Conversations API (Sunshine) * Zendesk API **Existing Client Impact:** * Existing clients currently utilizing the integration will automatically receive these updates without needing to reconfigure their setup. **Important Notes:** * If an email provided during handover matches an existing Zendesk Agent’s email address, the system will create an anonymous requester instead, without setting the email field. * Chat naming convention has been updated from "Conversation with Random Name" to "Ebbot - Incoming chat" for improved clarity and consistency. *** ### Web scraper stability * Updated the web scraper to fix issues where the scrapes would time out in certain cases. This provides stability and consistency to the Web scrapes. *** ### Bug fixes * Fixed an issue where the Multi select card would not work in the Welcome scenario. * Replaced the date picker in the Dashboard custom date selector.
## 2/2025 'Bond'
Release notes ### New Features * **Customer name displays in conversation:** The customer name is now displayed above chat messages in Live chat and Chat history conversations. If the end user is not identified, a random customer name is displayed instead of just "Customer." * **Hyperlink functionality for live chat agents:** Corrected agent link sending to display the link text instead of the full URL. * **Fixes for keeping conversation alive between subdomains:** Resolved issues with the subdomain manager, allowing conversations and chat widgets to follow end users when moving between subdomains. * **Source retrieval settings update:** Updated settings for source retrieval are clearer and easier to use, found under Configurations > Source Retrieval Settings. * **Support for Confluence as a source:** Added support for Confluence, a knowledge management system, by allowing data retrieval via API using user email, API token, and 'Space keys.' * **New default for retention settings:** Retention is now set to 1 year for all companies unless configured otherwise. * **Date and time picker available in Styling:** The Styling page for the chat widget now allows adding link cards and alert messages with specific display times. * **Dynamic blob text:** The Styling page for the widget allows adding multiple Blob texts, displaying them dynamically one after another. * **Improvements on the auto accept chat feature:** * Concurrent chat limits are applied conditionally based on "Auto Accept" settings and configurations. * Administrators have control over "Auto Accept" enforcement through a new setting * User settings require a minimum of one active concurrent chat for consistent availability. * **Expanded text area in Power up form:** The "textarea" in a Power up component now automatically expands for better UX when there is a lot of text. * **Enable triggering scenarios using shortcuts:** Live chat agents can trigger scenarios within a conversation using shortcuts. * **Flag messages that should not be sent to EbbotGPT:** Messages can now be flagged to prevent them from being sent to EbbotGPT. * **Widget design in mobile is improved:** The widget is now smaller by default when displayed on mobile. * **Add custom icons to link cards:** Users can now add custom icons to the widget starting page. * **New layout and setting groups on the Styling page:** Search functionality and several other UX improvements were made on the **Chat widget > Styling** page for easier navigation between settings. * **Hiding purposeless fields in bot response:** Fewer fields are now available in the Intermediate bot response as unnecessary ones have been removed. * Various UI & UX Improvements to the knowledge manager * **Limitation of characters in input:** The widget's input field allows up to 2000 characters by default, this limit can now be adjusted in Chat widget settings. * **Hide logo in widget styling:** It is now possible to easily hide the logo when styling the widget. ### Bug fixes * **Custom document display fix:** Resolved issue where custom documents in Knowledge were not displayed correctly. * **Fix for sending link with hash symbol:** Fixed issue where links containing '#' were not sent from agents in live chat. * **Fix for marking message content in live chat view:** Resolves the issue for agents not being able to mark message content in Live chat. * **Removed highlighting on start conversation button:** Resolved issue where the start conversation button was always highlighted. It should only be highlighted when using the keyboard to navigate, in accordance with WCAG standards. * **Role Management UI Fix:** Enabled custom role management, found under Administration > Account management. * **Typo fix:** Fixed typo in the Multi Select bot response. * **Fix for keyword trigger:** Changing a scenario trigger from NLP to keyword is now working properly again. * **Fix for widget design:** The widget CSS now better adapts depending on window size. * **Notification sound for end user:** Fixed issue with sound notifications for end users in the widget. Note: The end user needs to be in another tab for the widget to make the sound; it won't play if they are active in the widget. * **Fix for numbers in Live chat view:** The numbers displayed for "All closed sessions" and "Your closed sessions" in the Livechat view are now reliable again. * **Fix for message order:** Resolved issue with NLP messages displayed in the wrong order in the widget when triggering the same NLP scenario twice. * **End user using widget in several tabs:** Fixed issue with widget not working properly when end users use it from several tabs. * **API Key Pagination Fix:** Fixed pagination issue on API key management page. * **Fix for Chat history filtering:** Fixed issues with filtering on the Chat history page. * **Adjustment on attention seeker position:** Attention seekers in Conversation starters are now nicely positioned in relation to the chat blob. * **Notifications for livechat agents clean up:** Resolved issue for agents in livechat not getting their notifications cleared when accepting chats and working in several tabs. * **Removed outdated feedback feature:** The feedback feature in the Livechat view is removed as the support bot is a better channel for feedback. * **Ability to copy and paste from Microsoft system to widget input field:** It is now possible to copy and paste text from applications like Microsoft Word into the widget input field. The widget previously translated clipboard from Microsoft programs as images. * **Inactive bots should not load:** Fixed issue so that when a bot is set as inactive, the widget does not load. * **Dashboard metrics fix:** Resolved issue where device would display as empty instad of other-os which is the fallback on the Dashboard. * **Fix for visual bug in Dashboards**: Fixed an issue where the bar chart would duplicate data when hovering over different bars. * **Fix for deleting conversation tags:** It is now possible to delete conversation tags that includes an apostrophe.
## 1/2025 'Hammersmark'
Release notes **New Features** * **Web scraper now removes duplicate elements:** The web scraper now removes unnecessary duplicate elements resulting in multiple of the same documents. * **Live chat markdown support:** Live chat agents now have support for markdown formatted messages. * **Updated integration pages:** The integration setup pages has been improved functionally and graphically. * **Salesforce integration:** The Salesforce Live Chat integration is now available to set up from within the Ebbot platform. * **Visitor browsing:** Page history while engaging with ebbot is now available in the data object. * **UX when editing documents**: Added tooltips and icons displaying the lock status of documents as well as a way to re-enable disabled sources. * **Support for PDF scrapes:** It is now possible to upload PDF as source when creating new sources. * **New channel - WhatsApp:** Ebbot now supports WhatsApp as a channel integration. **Improvements** * The "starting page - page" metric is now displaying the full URL from which the visitor initialized a chat. * Various WCAG improvements. * Improvements to the Sharepoint source integration. * Added loading spinner to the chat history list. * Improved markdown rendering. * The Ready-toggle in live chat no longer changes visually if the status change fails. **Bug fixes** * Fixed an issue where the Sharepoint source integration would not find all documents of a site. * Fixed encoding issues in certain places. * Fixed an issue where the SAML card was displayed twice upon click. * Fixed an issue where the scenario audit history modal would get stuck after opening nad not load correctly. * Fixed an issue where the number of concurrent chats setting would not apply. * Fixed Presets permission issues for some users. * Fixed an issue where the agent availability status would not display for some users. * Fixed an issue where card setting dropdowns would not be deselected properly. * Fixed incoming webhooks not returning the conversation object. * Fixed the inactivity timer for agents which did not work properly. * Chat names on the live chat page should no longer be displayed as \`undefined\` on refresh. * Fixed an issue where the "leave conversation"-button needed extra clicks to function. * Fixed an issue where the page would crash which searching among documents. * Deleting a data set no longer prompts the confirm modal twice. * User count now displays the correct amount of users.
--- # 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.ebbot.ai/ebbot-docs/release-notes/changelog.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.