Release notes for watsonx Assistant for IBM Software Hub
IBM Software Hub
Use these release notes to learn about the latest updates to watsonx Assistant for IBM Software Hub.
Web chat versions
When you install watsonx Assistant for IBM Software Hub, the latest available version of the web chat integration is included. See the following table for details about the latest available web chat version for each watsonx Assistant for IBM Software Hub release. If your web chat version is not locked, then the web chat integration is upgraded to the latest available version when you upgrade watsonx Assistant for IBM Software Hub.
The following table shows the latest version of the web chat integration that is included with each release of watsonx Assistant for IBM Software Hub. IBM Software Hub supports web chat versions 8.3.2 or later. To customize and change version numbers, see Controlling the web chat version.
watsonx Assistant for IBM Software Hub version | Latest web chat version available |
---|---|
5.1.3 | 8.7.1 |
5.1.2 | 8.6.0 |
5.1.1 | 8.5.1 |
5.1.0 | 8.3.2 |
5.1.3 release, 30 April 2025
Updates
- Support for HTTP proxy configuration
- watsonx Assistant now supports HTTP proxy configuration. By configuring cluster HTTP proxy settings in IBM Software Hub, you can ensure that your cluster communicates with external repositories and access the required resources to install and update the software. For more information, see Applying cluster HTTP proxy settings to IBM Software Hub.
- Update on the search confidence for conversational search
- When you make small modifications to text processing, it impacts the values of retrieval confidence, response confidence, and extractiveness. For more information on the scores, see Conversational search analytics.
5.1.2 release, 26 March 2025
New features
- Use LLMs for your watsonx Assistant
- Large language models (LLMs) are now available for you to configure to improve the capabilities of your assistants. From the Base LLM section of the new Generative AI page, you can enable LLM features for the existing actions in your assistants to improve the conversation capabilities of your assistants. You can choose the foundation model that best interacts with your assistants for conversational search. For more information, see Supported foundation models for GPU features.
Updates
Evaluate testing message limits You no longer have a limit on the number of tests in a test run to evaluate and analyze your watsonx Assistant's performance but each test can include a maximum of 250 messages. For more information, see Testing limits for the set of data.
5.1.1 release, 27 February 2025
New features
- Hide or display the skill output
- You can now choose to hide or display the skill output that is stored as an assistant variable. For more information, see Passing values to a subaction.
- Integrate your watsonx Assistant with Slack
- You can integrate your watsonx Assistant with Slack. Then you can configure the assistant to support certain events so that your assistant can respond to questions that are asked in Slack direct messages or in Slack channels where the assistant is directly mentioned. For more information, see Integrating with Slack.
- Evaluate assistant's performance
- You can now evaluate the performance of your watsonx Assistant by uploading a comprehensive, relevant collection of utterances and sending the utterances to your assistant in a test run. . For more information, see Evaluating the assistant.
- Post-process the conversational search responses
- You can now save your conversational search responses in action variables of your watsonx Assistant for post-processing. For more information, see Search for the answer.
Updates
- Updated conversational search user query
- To improve integration with Elasticsearch, the question mark (?) is no longer automatically added to the user query in conversational search. In rare cases, this change might alter the conversational search responses.
- Limit on number of entries for translation download
- You can enable language data files to be downloaded for translation, only if the total number of entries in your assistant is 400000 or less. For more information, see Using multilingual downloads for translation.
- Updated LLM models for Conversational search
- As granite-13b-chat-v2 model is deprecated, you can now use the updated foundation model granite-3-8b-instruct for Conversational search. For more information, see Mirroring images directly to the private container registry.
5.1.0 release, 11 December 2024
For the list of watsonx Assistant known issues, see Limitations and Known issues in watsonx Assistant.
For a list of new features and fixes, see What's new and changed in watsonx Assistant.