New Features
Filter Test Results By Failures
Users can now filter the test results table to display only the questions that failed a specific evaluation. This feature enhances the usability of test evaluations by allowing for focused error analysis and correction.
Whitelabeling Admin Pages
Enhance your instance's visual identity with our new whitelabeling capabilities! Admins with branding permissions can now customize the look and feel of various features within Max. Access the "Branding" pages via the paint palette icon in the left navigation of the admin panel. From the favicon to the login and email screens, personalize your interface to align with your brand. Turn specific customizations on or off as needed.
Data Rights Tied to Copilot Rights
Data rights are now automatically assigned to users with copilot rights. Existing users with copilot access but without data rights need to manually update their access to include data.
Max Version Display in Query Browser
The query browser now includes a column that displays the version of Max used to run the query, ensuring better tracking and consistency across different versions.
Postgres Database Support
We now support Postgres as a database option in the connections page, expanding the flexibility and options for managing your data.
Database Connection Selection on Dataset
Change the database connection for an existing dataset easily by using the pencil icon next to the connection name. If the dataset's table exists in the newly selected database, it will be pre-selected for a seamless transition.
Feedback in Test Results
Test results now include user feedback for each evaluated question, providing clearer insights and aiding in continuous improvement.
Test Results as CSV
Test runs can now be downloaded as a CSV file, making it easier to analyze and share results outside of Max.
Configurable Max Response Prompt
Customize the prompt used in Max chat responses with Jinja templating. This new feature in the skill’s output node editor allows for more personalized and context-aware interactions.
Copilot Question Expected Skill Assertion
Users can now set an expected skill for Copilot Questions via the kebab menu in Skill Studio. After a test run, see a "Pass" or "Fail" status under the "Expected Skill Assertion" column for each assessed question.
Updates to Automated Nightly Evaluations
Our nightly evaluations now use a random selection of questions from the previous day, improving the diversity of our data analysis. These evaluations exclude questions from copilot test runs, focusing solely on user-generated queries.
Query Browser Chat Entry View Enhancements
The Chat Entry View now includes direct access to Debug and Evaluations tabs, making it easier to troubleshoot and assess queries right from the query browser.
Max Chat Response Renders Markdown
Max chat can now automatically render markdown, enhancing the readability and presentation of responses. This includes support for text formatting, code blocks, and styled lists.
Copilot Icons
Personalize your copilot by uploading custom icons, now visible in the copilot selector and at the beginning of chat threads.
Cancel Copilot Tests
Copilot tests can now be canceled if needed, providing more control over the testing process.
Edit Copilot Questions Modal
Copilot questions are now editable directly from the three-dot menu, facilitating quick adjustments and updates.
Table Name Change
You can now change a dataset’s table name with ease. Simply click the pencil icon under Source, select another table, and update your settings accordingly. Note that changes in table structure might affect dimension values and could lead to errors.
Dataset-wide Dim Value Searching
Explore and search dimension values across an entire dataset more efficiently with our new feature on the dataset page. Under the "Dim Values" tab, users can view all dimension values, see their mappings, original dimensions, and matched values. Utilize the context menu to either view the entire dimension or see the mappings for a specific dim value.
Updated