Changelog#

Lit#

User Experience:

Boosted candlestick chart performance for traders historical adapters by 200x
Titlebar improvements: Made titlebar consistent across all windows with maximize/restore, close buttons.
Added support to Discovery for image & json datasources

Administrative Improvements:

Sudo Access for Platform-Created Accounts: Accounts created within the platform can now be granted sudo access on the underlying machine. This new option provides administrators with greater flexibility in managing user permissions and system access.
Created an downloadable upgrade path for existing users

SDK improvements:

Discovery:

Full-Screen Discovery View: The Discovery view can now be expanded to full-screen mode, allowing for a more detailed and immersive exploration of your data.
Search Tool in Discovery View: We've added a search tool to the Discovery view. When used, all packets matching that string will be highlighted on the chart, making it easier to identify specific data conditions. Model Deployment
Custom REST API Endpoint: AI models can now be deployed to a custom REST API endpoint managed by the platform. This new feature provides enhanced flexibility for integrating models into different applications. Access to this endpoint is fully logged and secured with token management, which users can control to ensure data security and compliance.

Improvements to Experiments:

Server-Side Paging for Experiments Grid-View: We have implemented server-side paging to speed up the experiments grid-view. This enhancement is especially beneficial when live-monitoring multiple training sessions concurrently, each with tens of thousands of epochs, as it significantly reduces loading times.
Stop, Continue buttons in Grid-View: The grid-view on the experiments screen now includes the ability to stop or connect to running experiments, as well as resume experiments that were previously stopped. This enhancement streamlines experiment management, giving users greater control and flexibility directly from the grid interface.
Model Weights Format Change: Model weights are now stored as .keras files instead of .h5 files. This change optimizes compatibility and performance with modern TensorFlow versions.
Bug Fix for Archiving Experiments: A race condition that occurred when archiving large batches of experiments has been identified and resolved, ensuring smoother and more reliable data management.

Improvements to Assets:

Optimized Table Sizing: The asset tables are now automatically sized for the best fit on your screen, improving readability and user experience.
Sample Count in List Views: We have added a 'sample count' column to the asset list views. This allows users to quickly compare the total number of samples for each asset of multiple assets matching a search criteria in a single view.