Max.ai Analytics Assistant Release Notes - Version 24.19
Release Date: October 16th, 2024

Introduction

Version 24.19 brings several enhancements designed to optimize search functionality and export workflows within the Max.ai platform. This update introduces an improved fuzzy search algorithm, the ability to create skill snapshots during editing, and enhanced export capabilities for datasets and skill resources. These improvements focus on streamlining daily tasks and providing more control and flexibility for users working with assistants and skills.

Key Improvements

Export Datasets, Connections, and Resources

When exporting an assistant, any related datasets, connections, and resources (such as file attachments for assistants or skills) are now automatically included in the transport package. If resources are present, the export will be bundled into a zip file, with resources neatly organized into folders. This ensures that all critical files are maintained and easily accessible when transferring assistants between environments.

When importing, the system automatically extracts and applies all the relevant resources and connections, ensuring a smooth transition and full functionality in the target environment.

Export/Import Functionality for Prompts

It is now possible to export a prompt and then import it into a target environment. The "Import Prompt" submenu has been added to Add Prompt button. Users can now upload prompts via a transport package, ensuring any dependencies are handled during the import process. When importing skills or assistant, the system checks if required LLM prompts exist in the target environment. If a prompt is missing or has been altered, the user will receive warnings to address these dependencies.

Additional Enhancements

Enhanced Fuzzy Search Algorithm

The dimension value search within the question box has been enhanced with a more intuitive fuzzy search method. Results are now prioritized as follows: values starting with the search text appear first, followed by values containing the search text, and finally, those found through fuzzy matches. This ensures more relevant and efficient search results. This search technique is applied when searching for dimension values on a dataset, using the @ symbol in chat and by the dimension value grounding as a part of the chat pipeline.

Tested Models

This release has been tested with the following models to ensure compatibility and stability:

OpenAI:

• Chat: gpt-4o-2024-08-06
• Narrative: gpt-4o-2024-08-06
• Evaluations: gpt-4o-2024-08-06
• Embedding: text-embedding-3-small

Gemini:

• Chat: gemini-1.5-pro-latest
• Narrative: gemini-1.5-pro-latest
• Evaluations: gemini-1.5-pro-latest
• Embedding: models/embedding-001

Anthropic:

• Chat: claude-3-5-sonnet-20240620
• Narrative: claude-3-5-sonnet-20240620
• Evaluations: claude-3-5-sonnet-20240620
• Embedding: Not Applicable