Subsystem Update Log

Midori AI photo Midori AI photo

5/10/2024

  • Update: Planned changes for LocalAi’s Gallery API
  • Bug Fix: Fixed a loading bug with how we get carly loaded
  • Update: Moved Carly’s loading to the carly help file
  • Update: Updated the news page
  • Update: added invokeAI model support
  • Update: added docker to invokeai install
  • Update: Few more text changes and a action rename
  • Update: Cleans up after itself and deletes the installer / old files
  • Update: more text clean up for the backends menu
  • Update: added better error code for invoke.ai system runner
  • Update: added support for running InvokeAI on the system
  • Bug Fix: Fixed the news menu
  • Update: Added a new “run InvokeAI” menu for running the InvokeAI program
  • Bug Fix: Did some bug fixes

5/7/2024

  • Update: Added a way for “other os” type to auto-update
  • Update: Added a yay or nay to purging the venv at the end of other os
  • Update: Added a new UI/UX menu
  • Bug Fix: Fixed the news menu
  • Bug Fix: Fixed naming on the GitHub actions
  • Update: Added a way to get the local IP address
  • Update: Fully redid some actions that make the docker images
  • Update: Reworked the subsystem docker files and the new news post

5/5/2024

  • Update: Fixed some of Ollama’s support
  • Update: Action updates
  • Bug Fix: Fixed some server ver bugs
  • Bug Fix: Fixed a few more bugs
  • Update: Removed verlocking
  • Update: More fixes
  • Update: Added a new way to deal with python env
  • Update: Code clean up and fixed a socket error

4/22/2024

  • Update: Fully reworked how we pack the exec for all os
  • Update: Fully redid our linting actions on github to run better
  • Update: Mac OS Support should be “working”
  • Bug Fix: Fixed a odd bug with VER
  • Bug Fix: Fixed a bug with WSL purging docker for no reason

4/20/2024

  • Update: Added new “WSL Docker Data” backend program (in testing)
  • Update: Added more GPU checks to make sure we know for sure if you have a GPU
  • Update: Better logging for debugging
  • Bug Fix: Fixed a few bugs and made the subsystem docker 200mbs smaller
  • Update: Removed some outdated code
  • Update: Added new git actions thanks to - Cryptk
  • Update: Subsystem Manager builds are now on github actions, check them out - Actions

4/13/2024

  • Known Bug: Upstream changes to LocalAI is making API Keys not work, I am working on a temp fix, please use a outdated image for now.

4/13/2024

  • Update: Added InvokeAI Backend Program (Host installer)
  • Update: Added InvokeAI Backend Program (Subsystem installer)
  • Update: Site wide updates, added Big-AGI
  • Update: Updated LocalAI Page
  • Update: Updated InvokeAI Page
  • Update: Fixed Port on Big-AGI (server side, was 3000 now 33000)
  • Update: Removed Home Assistant links
  • Update: Removed Oobabooga links
  • Update: Removed Ollama link
  • Update: Full remake of the Subsystem index page to have better working links

4/12/2024

  • Bug Fix: Fixed the GPU question to only show up if you have a gpu installed
  • Update: Getting ready for InvokeAI backend program to install on host

4/10/2024

  • Bug Fix: Fixed a bug that was making the user hit enter 3 times after a update
  • Bug Fix: Fixed the system message on the 14b ai that helps in the program (she can now help uninstall the subsystem if needed)
  • Update: Added new functions to the server for new function based commands for the helper ai
  • Update: Updated Invoke AI installer (if its bugged let Luna or Carly know)

4/9/2024

  • Bug Fix: Fixed a loop in the help context
  • Bug Fix: Fixed the Huggingface downloader (Now runs as root and is its own program)
  • Bug Fix: Fixed LocalAI image being out of date
  • Bug Fix: Fixed LocalAI AIO image looping endlessly
  • Update: Added LocalAI x Midori AI AIO images to github actions
  • Update: Added more context to the 14B model used for the help menu

4/7/2024

  • Bug Fix: AnythingLLM docker image is now fixed server side. Thank you for your help testers!

4/6/2024

  • Bug Fix: Removed alot of old text
  • Bug Fix: Fixed alot of outdated text
  • Bug Fix: Removed Github heartbeat check ||(why were we checking if github was up??)||
  • Known Bug Update: Huggingface Downloader seems be bugged on LocalAI master… will be working on a fix
  • Known Bug Update: AnythingLLM docker image seems to be bugged, will be remaking its download / setup from scratch

4/3/2024

  • New Backend: Added Big-AGI to the subsystem!
  • Update: Added better huggingface downloader commands server side
  • Update: Redid how the server sends models to the subsystem
  • Bug Fix: Fixed a bug with ollama not starting with the subsystem
  • Bug Fix: Fixed a bug with endlessly installing backends

4/2/2024

  • Update: Added a menu to fork into nonsubsystem images for installing models
  • Update: Added a way to install Huggingface based models into LocalAI using Midori AI’s model repo
  • Bug Fix: Fixed some type o and bad text in a few places that was confusing users
  • Bug Fix: Fixed a bug when some links were used with Huggingface
  • Update: Server upgrades to our model repo api

4/1/2024

  • Update 1: Added a new safety check to make sure the subsystem manager is not in the Windows folder or in system32
  • Update 2: Added more prompting for the baked in Carly model for if you are asking about GPU or not with cuda

3/30/2024

  • Update 1: Fixed a bug with the subsystem ver not matching the manager ver and endlessly updating the subsystem

3/29/2024

  • Update 1: Fixed a big bug if the user put the subsystem manager in a folder not named “midoriai”
  • Update 2: Fixed the new LocalAI image to only download the models one time
  • Update 3: Added server side checks to make sure models are ready for packing to end user
  • Update 4: Better logging added to help debug the manager, thank you all for your help!

3/27/2024

  • Update 1: Fixed a bug that let the user use the subsystem manager with out installing the subsystem (oops)
  • Update 2: LocalAI images are now from the Midori AI repo and are update to date with LocalAI’s master images*
  • Update 3: Added the start for “auto update of docker images” to the subsystem using hashes