v1.3.3

Various fixes and improvements. Notable ones include:

• lmpanel script up now includes detached as default. No more -d needed. But if you don't want detached, you can use -nd flag (short for not detached, very creative, I know).
• Device and model ordering in the web interface is now handled via up/down buttons rather than drag and drop. Drag and drop looked nice but didn't work well on some devices.
• Docker containers don't auto start on system boot by default. You can adjust this behavior as needed but the new defaults ensure that you don't end up in a cycle of constant crash - reboot - crash - reboot if you are having stability issues. Although the following list item should make it a non-issue, on certain configurations, this can still be helpful:
• Watchdog behavior improvements that prevent OOM and potential system hang in multi-GPU environments when RAM constrained.

Ensure that with each update, especially this one, you update your env file. You can use a command like cp .env.example .env to do this. But, you'll have to move any adjustments to the new env file after doing so.

v1.3.2

Various minor stability and reliability improvements.

v1.3.1

This release has some stability improvements, with a focus on AMD hardware.

v1.3.0

Various fixes and improvements, but the major one is that instead of manually using profiles to determine which GPUs are supported when creating your Docker containers, we have combined everything into a single inference container and created a script based approach.

This automates the GPU selection process every time you run LmPanel. No more profile tags or remembering to use the matching profile tag when shutting down your containers.

Also, instead of using CUDA for NVIDIA, we're now just using Vulkan for all GPUs. After extensive testing across both single GPU and multi-GPU hosts, the performance difference is minimal. However, it vastly simplifies the code base.

If you have been using a previous version of LmPanel, please review the README and other documentation.

v1.2.2

Various fixes and improvements. Some notable ones include:
-Reimagined how pools and pooled devices are displayed on the Devices page.
-Added request timeout settings. This prevents "stuck" requests from wasting resources for extended periods of time.
-Fixed terms & policies markdown display so it's actually readable.

v1.2.1

Just some UI/UX improvements:
-Devices page: Layout is now consistent with Models settings where the settings are in a modal not directly on the page.
-Devices page: Device type now displays in all caps to match how device identifiers are displayed
-Packages page: Since Default package can not be renamed or deleted, the Delete button for that package is now not displayed, and the Name input field is now disabled. Before, attempting these actions would correctly result in an error, but the front end UI elements to attempt those changes were still present.

v1.2.0

Many changes and improvements. A few notable ones include:
-New feature: Export chat feature that generates a txt, json, or html file of that particular chat
-New feature: Brute force protection for failed logins
-New feature: Email notification for chosen events
-Fixed: Inaccurate counting of web search tool usage
-Fixed: Difficult to read text on file upload/processing toasts

v1.1.0

Many changes and improvements. A few notable ones include:
-Completely overhauled color scheme of user interface
-Adjustments to status page to improve visibility when using 2+ devices
-Fixed user IP in logs showing Docker container IP instead of the user's actual IP

v1.0.3

Many changes and improvements. A few notable ones include:
-Multi file sharded model support.
-Uploading model files takes half as long now because we write directly to the disk instead of writing to a temporary directory then copying them to the models folder. Not only is this awesome because I'm impatient but it will improve SSD lifespan.
-Everything in the web UI is [square] now!

v1.0.2

Many changes and improvements. A few notable ones include:

• New settings page layout and structure
• New model settings layout
• Dynamic favicon
• Improved handling of models