Studio Change Log

All notable changes to Rasa Studio will be documented in this page.

[1.12.0.dev2] - 2025-02-20

Bugfixes

• Fixed an issue around flow version validation where conditions linked to a slot now apply only to draft flows, preventing validation issues caused by historical flow versions.

[1.12.0.dev1] - 2025-02-17

Features

• Inspector mode: Studio now includes Inspector Mode—a detailed debugging view that displays flow events, slot updates, and response details in real time. Users can easily switch between a simplified view and Inspector Mode to track their assistant’s behavior and resolve issues without leaving Studio.

Learn more about Inspector mode

Improvements

• Support OR operator between logical conditions: Users can now use the OR operator in conditions throughout Studio. This enhancement brings Studio closer to Rasa Pro in functionality, giving you more flexibility in designing your flows.
• Improved navigation between invalid nodes in a flow: Users can now quickly identify and fix validation errors, even in large flows.
• System Flow Reset Improvement: Resetting system flows no longer clears their change log, ensuring the flow history remains intact.
• Improved Flow List Ordering: The Flow list is now ordered based on the last update made by users, ensuring it reflects direct user actions rather than system-driven changes.
• Introduction of Superuser Group: A new Superuser user group has been introduced. Members of this group have full access to all features provided by Studio.

[1.11.1] - 2025-02-05

Bugfixes

• Fixed an issue where the slots list was not updating after creating new slots.
• Corrected the behavior to ensure landing on the correct response in the responses page when creating or editing responses.
• Other minor bug fixes.

[1.11.0] - 2025-01-15

Features

• Responses in CMS: We’re introducing a centralized hub for content managers and flow builders to manage all the responses used in an assistant project in one place.

Learn more about Responses in CMS

• Conditional Response Variations: Conditional responses let your assistant choose specific response variations based on one or more slot values, enhancing personalization and flexibility in replies. This feature allows users to easily create multi-lingual assistants or segment responses based on different criteria.

Learn more about Conditional Response Variations

Bugfixes

• Fixed an issue on import of categorical slot type.
• Aligned Validation Rules between Rasa Studio and Rasa Pro for configuring conditions where values are surrounded by double quotes.
• Other minor bug fixes.

[1.10.1] - 2024-12-19

Bugfixes

• Remove unwanted model-service database creation step from db migration job during Studio deployment.

[1.10.0] - 2024-12-17

Features

• Simpler Training in Studio: Enhanced Studio’s training architecture and performance to improve reliability, reduce failures, and simplify Deployment
  • The old setup involved multiple services for model running and training, leading to more points of failure. These have been consolidated into a single “Simple Model Service”, reducing moving parts and making training more reliable. This brings down the number of container images in Studio to 5.
  • The new architecture allows for faster training times, with a 85% improvement in training time.
  • We’ve replaced NFS with new storage options to provide greater flexibility for storing trained models:
    • Disk Storage: Simplified local storage for using Kubernetes Persistent Volume.
    • Cloud Integration: Seamless integration with your preferred cloud storage solution.

Old training architecture.

New training architecture.

• Error UX Improvements: Various training and deployment errors are now handled directly in the UI, allowing users to view errors in the popover without needing to download the training log.

Error during training.

Error during Deployment, which goes right after successful training.

• Model Download: Model input data which was used to train a model can be downloaded from UI in YAML format.

  

Improvements

• Performance Improvements: Improved performance of the Studio application to support large assistants and complex flows. Performance improved for following features:
  • Assistant pre-training validation
  • Assistant export and training
  • Toggling “ready for training” for flows linked from other flows
  • Creating a Logic node
  • System flows page
  • Small performance improvements in a few other areas
• Model Download API for CI Integration: Updates to the Model Download API to support CI integration to support the new model service architecture.

Migration Guide

• Make sure to use the latest Helm chart version 2.0.1 & above for deploying this release.
• Make sure to provision a Persistent Volume for storing trained models in the new architecture. Alternatively you can use cloud storage for storing trained models. More details can be found in the deployment guide.
• If you are migrating from an older version of Studio, you can safely delete the below pods and services from your Kubernetes cluster after deploying the latest version of Studio:
  • Pods starting with dsj- for example dsj-0f436992-895c-4fb3-92cb-045ea06c41df-4q7tx
  • Pods starting with tsj- for example tsj-0f436992-895c-4fb3-92cb-045ea06c41df-4q7tx
  • Services starting with dss- for example dss-048671c6-3086-49c0-afa7-a17043df2287

[1.9.0] - 2024-11-08

Features

• Conversation Tagging API: Expose API for tagging conversations for our clients. Key updates include:

  • Keycloak client: Introduce studio-external Keycloak client ID to authenticate the API calls
  • Simplified conversation query: Fetch conversations filtered by time range
  • Delete conversations: A new API to delete conversations by ID
  • Tag management: Tag IDs will be visible in the Studio UI to support integrations that use conversation tagging
  • Assistant ID: Assistant UUID id is now visible for customers using API to leverage it

  Learn more about conversation tagging API

• API for CI integration: API for CI integration enables Studio users to automate the download of their assistant data via an API endpoint. Instead of manually exporting and pushing data to their infrastructure, technical users can now pull Studio's data directly into any CI platform. This is accomplished by exposing a query that returns raw YAML files and the trained model file (model.tar.gz). It is possible to download latest and all trained assistant versions.

  Learn more about API for CI integration

[1.8.1] - 2024-10-30

Bugfixes

• Fixed an issue to prevent steps from being wrongly referenced in flow during export/download

[1.8.0] - 2024-10-23

Features

• Flow Version Control: Flow Version Control allows you to save the flow version at any point, which will be logged in the flow history. From the flow history, you can restore previously saved versions, allowing you to revert any breaking or unwanted changes, giving you further control over the changes that you or your teammates have made to flows.

  Learn more about flow version control

  

• Button Payloads: Button payloads allow you to assign commands to buttons, so when a user clicks on them, your assistant knows exactly what to do next. This can include:

  • Setting slot values
  • Triggering intents
  • Passing entities to the assistant
  • Sending predefined messages to guide the conversation

  While Rasa Pro users have been enjoying this feature, we’re bringing this capability to Studio, making it even more powerful and flexible for building your conversational experiences.

  Learn more about buttons in Studio

  

• Download your assistant project You can now download your entire assistant project, including all flows marked as "Ready for training," along with the NLU data and configuration files, all in one click.

  

[1.7.2] - 2024-10-04

Bugfixes

• Fixed a couple of issues with conditions of if in logic node.
• Fixed an issue where empty page was rendered if slot used in flow guard condition was deleted

[1.7.1] - 2024-10-01

Bugfixes

• Fixed an issue where assistant couldn't be trained when a flow uses flow guards.

Improvements

• Added support for jinga templates.

[1.7.0] - 2024-09-20

Features

• Customizing patterns in Studio: System flows, also known as patterns, are pre-built flows available out of the box, designed to handle conversations that go off track. For example, they help when:

  • The assistant asks for information (like an amount of money), but the user responds with something else.
  • The user interrupts the current flow and changes the topic.
  • The user changes their mind about something they said earlier.

  You can now fully customize these system flows in Studio in the “System flows” tab.

  Learn more about patterns in Studio

  

• Custom actions in the Collect step: Studio now supports collecting slot values via custom actions, in addition to the existing method of using responses. This feature, available from Rasa Pro 3.8, enhances the flexibility of the Collect step.

  Instead of solely relying on templated responses, you can now employ a custom action to collect slot values. This is particularly useful when you need to display dynamic options fetched from a database, such as presenting available values as buttons.

  Learn more about different ways to collect information

  

Improvements

• You can now export your flow as an image, making it easier to share and visualize your conversations.

  

[1.6.1] - 2024-09-10

Bugfixes

• Fixed issue in upload/download assistant with entity slot mapping with roles.
• Prevent slot creation and update if name contains the dot character.
• Support rasa studio import to work with flows starting with condition step.

[1.6.0] - 2024-08-30

Features

• Assistant Settings: The new "Assistant settings" page in Studio is designed to streamline how you manage and configure your assistant. This new page replaces the "Manage assistants" modal and features two key areas:

  1. General settings: View and manage basic settings such as the assistant's name and mode.

  2. Configuration: Allows users in developer role to directly edit the config.yml and endpoints.yml files, enabling real-time adjustments without leaving Studio.

  This update makes it easier to see and adjust your assistant’s settings directly within Studio, enhancing control and flexibility.

  Learn more about Config in Studio

  

• Scalable Command Generator: The flow-retrieval feature, introduced in Rasa Pro 3.8, is now implemented in Studio. It ensures that only the flows relevant to the conversation context are included in the LLM prompt. This optimizes performance by managing the number of flows more efficiently and reduces LLM-related costs.

  Enabled by default in the config.yml file accessible via the new Assistant settings page, this feature can be adjusted or disabled as required. It also allows users to designate specific flows as "always included" in the prompt, ideal for general flows like "welcome" and "chitchat" to ensure they are consistently retrievable.

  

• Slot Mappings: Slot mappings are global settings that define how slot values should be extracted.

  By default, in Rasa, the slot value is extracted from the user's message using an LLM. However, if users prefer to extract slots from a specific custom action, intent, entity, or the last user message instead of relying on LLM, they can now do so in Studio. Learn more about slot mappings

  

• Flow Change Log: To enhance collaboration and increase transparency in assistant development, we've introduced the flow change log. This feature provides a detailed history of each flow's changes, including details on who made the change and when. It aims to improve project management by allowing users to track modifications and eventually, revert to previous versions if necessary [coming soon].

  

[1.5.6] - 2024-10-31

Bugfixes

• Fixed an issue to prevent steps from being wrongly referenced in flow during export/download

[1.5.5] - 2024-09-26

Bugfixes

• Increased database transaction timeout and made performance improvements to prevent errors during flow node creation.

[1.5.4] - 2024-07-31

Bugfixes

• Fixed issue where in-development feature flags were not being respected

[1.5.3] - 2024-07-30

Bugfixes

• Fixed performance issues related to flow validation

[1.5.2] - 2024-07-26

Bugfixes

• Fixed validation issues related to linked flows
• Fixed issue where a new flow's ready for training button was disabled even after adding a valid node for the first time
• Other minor bug fixes

Improvements

• Update studio-database-migration container image and studio-backend image to support deployments in restrictive environments

[1.5.1] - 2024-07-23

Improvements

• Update studio-web-client container image to use nginx-unprivileged base image to support deployments in restrictive environments

[1.5.0] - 2024-07-19

Features

• Flow Validation: Users are provided with real-time error feedback during the flow creation process. This ensures that errors are caught before training starts and allows for explicit error messages to be displayed in the user interface.

  The feature includes:

  Mapping errors to specific fields where they occurred — allowing users to easily return to the flow and correct them at any time.

  

  Error status on the assistant level - after clicking the "Train" button and before actual training begins, Studio performs a quick validation of all the flows marked as “Ready for training”, indicates if there are errors in any particular flows, and shows the list. This helps users navigate through all the errors until they’re all fixed before attempting to train the model again.

  

• Call a flow and return: This step enables control to jump from a parent flow to a child flow and then return to the original parent flow. The child flow is treated as part of the parent flow, which results in the following behaviors out of the box:

  • Slots in the child flow can be filled upfront, without needing to reach the collect step of that slot.
  • Slot values in the child flow can be corrected after they have been filled and control has returned to the parent flow.
  • Slots in the child flow are not reset until the parent flow is completed.

  

• Data retention policy: Users can define a custom period for how long Studio should retain conversation data.

Bugfixes

• Fixed issue where the import of a CALM assistant fails when nlu.yml is missing or when no entities are present in the nlu.yml file
• Fixed import failure of a CALM assistant when slots are of boolean type
• Conversation view filter are now reset when user changes the assistant

Improvements

When logging in, the user is redirected to a specific page based on their role:

In CALM-based assistant:

• Flow Builder, NLU Editor, leadAnnotator, Developer and Business User are redirected to Flows page.

In Classic assistant:

• Lead Annotator is redirected to Annotation Dashboard.
• Annotator is redirected to Annotation Inbox.
• Business User, NLU Editor, Flow Builder, Developer are redirected to NLU

Conversation Analyst is redirected to Conversation View, regardless of assistant mode.

[1.4.0] - 2024-06-14

Features

• Flow Guards: Flow Guard enables users to define specific conditions for triggering a flow, ensuring that the flow is not activated solely by user request

  

Bugfixes

• Fixed issue related to primitives not being pre-selected in the Manage window
• The condition editor no longer disappears when a new slot is created
• Fixed issue where Slot is not being updated after a different slot is selected in the dropdown
• Other minor bug fixes

Improvements

• Updated label of condition nodes. We now provide a clear indication of the logic within the condition, allowing users to quickly understand the conditions at a glance
• Added validation to check that the slot name doesn't contain any special characters
• When creating a collect message for a categorical slot, only values of a categorical slot can be used to create buttons for the related message, so that wrong entries can be prevented
• In the Flows table long flow descriptions are now visible on hover over the description
• Users can share links to filtered views of the Conversation Table and to individual conversations
• Other minor application improvements

[1.3.2] - 2024-06-05

Bugfixes

• Fixed issue where slots on different logical nodes got merged when exporting a flow to yml

[1.2.1] - 2024-06-04

Bugfixes

• Fixed issue where slots on different logical nodes got merged when exporting a flow to yml

[1.3.1] - 2024-05-30

Bugfixes

• Fixed issue where error raised during an unsupported yml import of a CALM assistant had an incorrect error message

Improvements

• Added support for entity annotation to CALM assistant import

[1.3.0] - 2024-05-16

Features

• Assistant Import: CALM Assistant import allows Rasa Pro customers to import existing assistants into Rasa Studio. Learn more.

Bugfixes

• Fixed issue where adding a new categorical slot value doesn't cause the if condition to not display the previously defined value anymore.
• Initial Value selection is now saved correctly for new categorical slots.
• In Flows, when adding a Logic step, the nodes order is now correct.
• Other minor bug fixes.

Improvements

• Auto-format primitives: Validation for message and action names is simplified: they don't need prefixes utter_, utter_ask_ , utter_invalid_, or action_ anymore.

• For all primitives that don't allow spaces in names, whitespace is automatically replaced with _ during typing.

  

• Search option for managing slots is added.

• Behaviour of "Ready for training" checkbox is now consistent.

• Support for additional events in Conversation Review event stream: Users can now see the following additional events in the conversation stream: FORM, ACTIVE_LOOP, FOLLOWUP_ACTION, RESTARTED.

• Review multiple conversations in Conversation Review: Users can select multiple conversations and create a batch for review.

  

  

• Conversation Review RBAC: Admins can now assign a Conversation Analyst role to users. Only users with this role will be able to access and use the Conversation Review feature.

• Change default classifier from DIETClassifier to LogisticRegressionClassifier for assistants with NLU triggers.

• Other minor application improvements.

[1.2.0] - 2024-04-18

Features

• Conversation View: Helps users identify ways to improve their NLU and CALM assistants by reading real user conversations. With Conversation View, users can browse user conversations, apply filters to help surface conversations in need of analysis and then see a turn-by-turn breakdown of the conversation, along with relevant session and event data.

  

  

• NLU Triggers: Enable users to create intents in Modern assistants and use them as a method for triggering flows. Start step is used to provide a comprehensive view of what can initiate or hinder the start of a flow: CALM, NLU Triggers, or Links/Calls.

  

• NLU page in Modern mode: Bring NLU page into Modern assistants for managing intents, setting a precedent for future primitives management. Users can now create, edit, and delete intents in the NLU page.

  

Bugfixes

• Fixed issue where the users were unable to export else node when it is in the end of the flow

• Fixed issues related to inconsistent state of Manage Slots modal

• Fixed issue where navigating back from a flow takes the user to the 1st page in the list of flow

• Other minor bug fixes

Improvements

• Added an optional "Initial value" field to slots: Let users specify an initial value for any slot, just as they can already do in Rasa Pro.

  

• Users no longer have to type utter_ prefix when creating a message node

• Users no longer have to type action_ prefix when a creating a custom action node

• Other minor application improvements

[1.1.2] - 2024-03-27

Bugfixes

• Fixed issue where users were unable to update a slot under certain conditions

[1.1.1] - 2024-03-26

Bugfixes

• Removed the alphabetical sort of categories in the "Manage slots" modal
• Fixed issue where "Open the flow in a new tab" did not open the selected flow
• "Show variations" button in a message node is only displayed if there are message variations present
• Fixed issue which caused the creation of orphan nodes in flow builder
• Else condition nodes now do not allow assigning conditions to them
• Fixed issue where the flow builder was not listing all the flows while creating a link node
• Fixed issue related to invalid training data with wrong reference ID
• Studio now displays the affected message name in the error toast when an incorrect slot name is used in a message node
• Other minor bug fixes

[1.1.0] - 2024-03-18

Features

• Support the logic operators from the Pypred library that are currently missing in Studio

  • not: Negates a condition
  • =: Equal to
  • !=: Not equal to
  • matches: Uses regular expressions to match strings
  • not matches: Uses regular expressions to negate strings.

  

• Allow flow builders to assign slot values to slots directly in the flow without using custom actions or collecting user input. By doing so they can dictate the logic independently from user input. Slots can either be precomputed with a value or reset.

  

Improvements

• Enable users to select or unselect flows for training directly from the flow list

  

• Other minor application improvements

Bugfixes

• Annotator user is now prohibited from viewing options for editing NLU
• Chat history is now not shared between all assistants and user logins
• Fixed issue related to a race condition when changing categorical slots
• Improved error message when training is initiated with an invalid slot value and invalid config yaml
• Other minor bug fixes

[1.0.4] - 2024-02-12

info

Make sure to use Rasa Studio Helm chart version 0.4.0 & above for deploying this release.

Features

• Try your assistant page now maintains chat history

  Please note that the chat history is maintained in the local storage of the browser. The chat history is not maintained across different browsers

• Users can now search flows by flow description

Improvements

• Improved security of Studio container images. Studio now uses non-root user for running the application
• Simplified deployment time variables in the latest 0.4.0 Helm chart release
• Other minor application improvements

Bugfixes

• Fixed issues connected to rasa studio download cli failure due to absence of slots and empty flows
• Other minor bug fixes

[1.0.3] - 2024-01-22

Features

• Users can now search for primitives by typing in the search bar

• Users can now use Azure OpenAI API with Studio to train their assistants. A new RASA_CONFIG_FILE environment variable can now be passed to the backend service to define the config.yaml file used for training the assistant

  Please note that the contents of the config.yaml file needs to be base64 encoded and passed to the new environment variable. The Studio backend service will decode the variable value and use it for training the assistant. Users can pass the Azure OpenAI API key to the existing OPENAI_API_KEY_SECRET_KEY environment variable

  The RASA_CONFIG_FILE values overrides the Advanced configuration options defined in the Create assistant Project model in the Studio web client

  Sample config.yaml file:

    recipe: default.v1
    language: en
    pipeline:
        - name: LLMCommandGenerator
        llm:
            model_name: gpt-3.5-turbo
            api_type: azure
            api_base: https://studio-testing.openai.azure.com
            api_version: "2023-07-01-preview"
            engine: gippity-35
  
    policies:
    - name: rasa.core.policies.flow_policy.FlowPolicy

  Corresponding base64 encoded RASA_CONFIG_FILE value:

  cmVjaXBlOiBkZWZhdWx0LnYxCmxhbmd1YWdlOiBlbgpwaXBlbGluZToKICAtIG5hbWU6IExMTUNvbW1hbmRHZW5lcmF0b3IKICAgIGxsbToKICAgICAgbW9kZWxfbmFtZTogZ3B0LTMuNS10dXJibwogICAgICBhcGlfdHlwZTogYXp1cmUKICAgICAgYXBpX2Jhc2U6IGh0dHBzOi8vc3R1ZGlvLXRlc3Rpbmcub3BlbmFpLmF6dXJlLmNvbQogICAgICBhcGlfdmVyc2lvbjogMjAyMy0wNy0wMS1wcmV2aWV3CiAgICAgIGVuZ2luZTogZ2lwcGl0eS0zNQoKcG9saWNpZXM6CiAgLSBuYW1lOiByYXNhLmNvcmUucG9saWNpZXMuZmxvd19wb2xpY3kuRmxvd1BvbGljeQ

Improvements

• Improvements related to order of condition steps in flow export
• Better user notifications when there is a connection issue with a deployed model in try your assistant page
• Other minor application improvements

Bugfixes

• Fixed issue related to flow training when first step is the Logic step
• Users can now delete unused slot
• Other minor bug fixes