Release Notes
Learn about our product's newest updates and enhancements that will make your conversational applications better.
Last updated
Learn about our product's newest updates and enhancements that will make your conversational applications better.
Last updated
The first update for this month is a new sidebar item "Tutorials" featuring various how-to guides. We've added this enhancement to improve your learning experience, allowing you to access our learning videos, documentation links, and additional resources all in one convenient place.
The second one is a new API endpoint you can use for pre-populating user context for a conversation in full-page webchat. You send us initial attribute values for a given user, and in return, you'll get a link for a full-page webchat for this user with this data ready to go. More information is available in our docs.
Fixed carousel slider styles;
Fixed language switching in webchat via 3-dots menu;
Removed redundant condition operator when there is only one condition;
Added better feedback for webhook action failure so you can debug more effectively;
Added links to language services in interpreter dropdown;
Fixed default styles for tables and numbered lists in markdown messages;
This month, we have released a significant update to the preview page and made several general improvements to the system.
We reflected on our preview page and redesigned it to make testing your conversations more efficient.
The preview console now takes up the full right side of the screen, featuring two tabs: "Conversation Logs" and "Attributes". This layout allows you to see incoming and outgoing intents more clearly, with the considered and selected paths easily accessible.
In the "Attributes" tab, you'll find an enhanced search function and filters. Additionally, custom attributes are clickable and will take you directly to their management page.
Setting custom attributes to test specific paths of your conversation has also become a lot easier, as you can now set multiple attributes.
For more information, please see the docs.
Fixed issue when Chatwoot integration is inactive by default
Improved spacing for code blocks in markdown messages
Made conversation intents clickable within considered and selected path panels
Resolved bug with conditions during initial scene creation
Moved the banner for quickstart AI agent to the Dashboard.
Happy New Year! To start the year off with a bang, we wanted to give our messages even more flexibility. Which is why, in this release, we are happy to introduce markdown support for our text messages. In addition, we have been busy making improvements throughout the product, to make the usability of our product even smoother.
Our text messages now support markdown. You can create bullet lists, bold/italic text, headings, links, images, code blocks, and tables using standard markdown syntax, and they will be rendered in webchat. For more information, please see the docs.
Fixed the double-scroll bar issue on the preview page;
Fixed leading and trailing whitespaces in webhook action headers;
Fixed sidebar overlapping with incoming intents in preview;
Improved messaging for the semantic classifier test panel when no intents matched.
We are wrapping up this year with some user experience improvements and clearing some small flaws from our platform.
We are delighted to present a tracker for your vectorisation usage restrictions. Each Topic page now features a progress bar that shows the number of jobs currently in progress and updates each time you add new items to the queue until you reach the quota. This should help you better understand how many jobs you can have per topic at any given time.
Order list of available actions in conversation designer side panel alphabetically;
Fix visibility issue for existing conditions on message card in message editor;
Fix reCAPTCHA resubmission issue;
Improve performance for dashboard page.
After all the significant changes we made in October, we've taken some time to further improve our existing features and improve usability across the platform, for you and your users.
Does it sometimes feel like it takes ages for the AI Agent to reply and you are worried that your users might feel something went wrong, or simply give up? We've got you covered!
Introducing...OpenDialog's interactive wait time indicator.
Using our interactive wait time indicator, will :
Keep users engaged without causing anxiety that something might be broken
Enhance the sense of immediacy of the AI Agent's reply
Provide agent ‘presence’ while the user is waiting
Improve interaction quality and trust in your AI Agent
Give it a try in your scenario's Interface Settings and immediately notice the difference between occupied and unoccupied time!
Improved indication in the navigation sidebar
We have made improvements to the navigation sidebar for language services and workspaces, as well as indicative labels for the different language service types.
Ability to duplicate actions
In the three-dot menu for any action, you can now find a new option to duplicate this action so you don't need to start from scratch every time.
Condition Operators
So far, conditions in OpenDialog inherently were considered as using the 'AND' operator. We have now added the possibility for you to combine conditions differently, by letting you choose the operator that controls them : AND or OR.
Performance improvements for webchat
Fixed date picker message broken in some cases
Improved messaging for when the vectorization queue is busy
Fixed appearance of the page with aliases
Ability to expand the semantic classifier prompt field
Fixed long strings breaking the formatting of a list message
Fixed data mismatch between the analysis page and API data export
Fixed embed code button visibility
This month, we've got a LOT for you! Both things that please the eye and improve the quality of life on our platform!
New designs for login and registration screens
We've got a much better and slicker design for some of our pages, such as Login, Registration, and Reset Password.
New sidebar
Our new navigation pattern makes it easier for you to navigate the platform and accomplish your tasks more quickly. We've also uncovered some hidden gems like user management and attribute dashboard in case you knew it was there but were going to them via direct URL 👀
New way of creating scenarios
Say goodbye to the snake animation (my favorite) on the scenario creation screen and behold a brand new one with an updated "Start from scratch" template, the ability to import scenarios, and visual updates to this one
So many improvements await you in this release, worth highlighting these ones:
Allowed for different date/time formats via attribute filters.
Updated the interface settings page to the latest relevant state.
Added a setting in URL-source vectorization to omit HTML tags.
New pop-up for prompt editing.
Fixed the preview dropdown z-index.
Fixed the "add scene/turn" button always showing the wrong sidebar on the first click.
Fixed LLM Action output attributes not enforcing unique attribute names.
Fixed the Card component being cut in half when the screen resolution is low.
Fixed Activation toggles sitting outside of the Card component in LLM Actions.
Fixed the confirmation popup when deleting an action.
Activation toggles are now on by default on all system components.
Unified the behavior of description fields.
A significant milestone in OpenDialog's evolution, bringing you an even more powerful way of building AI Agents!
Semantic Intent Classifier
Remember our intent classifiers? Imagine a world where you can define a simple list of intents and a short description for each, and thats it. No example phrases, nothing! Any semantically matching user utterances will be classified as you want. Guess what? We did it! With our brand new semantic classifier you can do just this - specify an intent name, give it a description, choose an LLM engine which will be used to perform classification and use it in your scenario.
For more details check out our documentation here.
RAG services
We now have our very own support for Retrieval-Augmented Generation (RAG) which gives you the ability to generate contextually relevant responses, based on a semantic search over a vectorised knowledge base. You can now use your data to feed into your AI agents and use this knowledge to drive conversations. Want your users to be able to ask questions about where to find a parking spot at your facility? No problem. You have complicated policy document and you want your AI agent to answer questions about its content? In OpenDialog, it has become as easy as ABC. With our RAG service you can use any custom URL, pdf, .csv, docx. xlsx file or just simply drop raw text and our RAG service will take it from there.
More in our docs here.
Fixed issue with processing composite JSON object an output attribute from wehbook action
Fixed unexpected behaviour when no-matching / escalation
Fixed override input field in new webchat causes blocking error
Removed character limit for webhook action URL
Removed ability to create outdated interpreters such as LUIS, QnA and RASA. Existing scenarios which uses them will still work
Added soft wrap for long component names
Fixed stretched cards on dashbaord
Added new "step" property to number input
Added ability to map attributes from another attributes in "Set Attributes" action
LLM Actions
We've been working hard to bring you an exciting new way to interact with LLMs, called LLM Actions! It's like a usual action which you can attach to an intent but which will leverage LLMs to perform a task. Give it a system prompt, specify output attributes, configure moderation - and make your AI agent respond with dynamic messages, extract specific information from user utterance, format request to upstream webhook and much more. Check out documentation for detailed usage.
We didn't forget to do some housekeeping in our product so we did a few fixes, such as:
Fixed akward minimize/maximize webchat behaviour
Fixed download chat link with iphone not working
Fixed lower options are hidden in dropdown menu
New gpt-4o-mini model support
We've been busy tidying up our product and making a lot of improvements and bugfixes, including:
Fix for message history is not consistently ordered
Allow free input for standart autocomplete
Support for cancel functionality for different message types
Added address autocomplete message
Added new location attribute
Added support for GPT-4o model
Message Design: Introducing further support via the Message Editor to use the different message types that were introduced during the Proxima release. You can learn more on how to use these new message types here:
We also made some improvements & bug fixes throughout the core product, as well as webchat.
Some of the updates include :
We have added GPT-4o to the managed available Language Services model
We have added links in the Message Editor to the different message types for easy access
We have updated the scrolling positioning when a message is served in webchat
Webchat is now compliant with WCAG 2.2 standard 🎉 It consist of many little pieces like:
Changed color contrast
Highlighted focused elements
Improved visual validation feedback on forms
Font sizes using Em instead of pixels
Keyboard navigation
Screenreaders can read webchat
Semantic changes to webchat HTML
and many more...
We've made various improvements and bug fixes to the core product, including :
Support for regex validation in form messages
UI fixes to a set of webchat messages
Interpreter Orchestration: By integrating Interpreter Orchestration into your workflow, it will allow for you to set highest priorities and be able to create smoother, more efficient running conversations with OpenDialog.
Interpreter orchestration is a feature that is available ON DEMAND. If you would like to use this feature, please reach out to support@opendialog.ai.
We also made some improvements & bug fixes throughout the product.
Some of the updates:
Message Editor : Button message now defaults to not allow continued interactions
Message Editor : You can now change the order of your message blocks using drag and drop
Attribute management: By integrating Attribute Management into your workflow, you can streamline the process of personalising conversations. Whether you're modifying existing attributes or introducing new ones on the fly, this feature simplifies the customisation process, enhancing the efficiency of your interactions.
A significant number of improvements & bug fixes have been made throughout the product to improve the conversation design experience.
Notable updates:
Allow designers to add placeholder text to form messages
Cancel button is now configurable in the message editor for form messages
Fixed the OpenAI OD-managed language service to return results
Fix to alias embed code to support nu-webchat scenarios
Language Services: Thanks to OpenDialog Language Services, you can easily manage your language model directly within the tool. This feature is now generally available (previously in Beta)
Aliases : OpenDialog Aliases allows you to create a single point of reference for your Web Chat application. This functionality enables you to alternate different scenarios at any given time.
NEW WEBCHAT!
OpenDialog webchat becomes more intuitive and highly customizable to adapt your branding through advanced Interface settings, theming, and custom messages.
Thanks to the integration with Speech-to-Text, users can now input their information and queries by voice.
Existing messages have been given a new look and feel, and a new set of custom messages has been introduced in new webchat as part of this release, and will continue to be updated in upcoming releases.
OpenDialog Integrate has been enhanced with integrations that offer the possibility to send end-user requests through a ticketing system, to a live agent or trigger an e-mail.
Set Attributes to Value on Intents: You can now design intents to set a value against a certain attribute based on the fact that this intent gets triggered. As an example : you can add the value 'done' to an attribute 'welcome' based on whether or not the WelcomeIntent was triggered. This allows you to set up even more varied conversational flows determined by whether a user has 'visited' a certain intent, or segment a user by the fact that certain intents were triggered.
Autosaving intents: With this release, we are introducing an Autosave functionality to the Intent settings in the Conversation Designer. You no longer have to remember to Save or Update your settings with every edit you make to your intents, we will do it for you - saving you time and headspace!
We have fixed a persistent bug in the product related to the description boxes for the different components in the Conversation Designer. All updates to these descriptions will now get saved and updated seamlessly.
Language Services: OpenDialog introduces Language Services, a centralized no-code interface for integrating, managing, and training your language models directly in OpenDialog. Available language services currently include : OpenDialog default intent classification, OpenAI-powered intent classification. Further language services will become available soon, including embeddings-powered Q&A.
This feature is currently in beta release and available on demand. If you want to access Language Services, get in touch via support@opendialog.ai
Send to Email: You can now trigger an e-mail to be sent during a user's conversation with your application, personalizing its content with variables from the context.
Microsoft Conversation Analysis Interpreter: OpenDialog allows you to integrate with your own language models via our Interpreter feature. Our Interpreter feature now also supports integration with the Microsoft Conversation Analysis service.
Microsoft Question Answering Interpreter: OpenDialog allows you to integrate with your own language models via our Interpreter feature. Our Interpreter feature now also supports integration with the Microsoft Question Answering service.
Webhook interpreter: In July, we also added a webhook interpreter to integrate with external language models without direct OD integration through webhooks.
We have added two new actions to the Action Library: storing the state of the conversation in the user context and returning to that state during the conversation.
Persistence of the user context when using virtual intents
Performance fixes
Freshdesk integration: The Freshdesk action lets you set up a connection between OpenDialog and your Freshdesk ticketing system, which you can then use within your conversation design.
Conversation hand-off: OpenDialog now allows you to set up a seamless handover to a human agent without interrupting the conversation.
And much more
Plus various bugfixes and improvements
