Releases: yukiarimo/yuna-ai
Releases · yukiarimo/yuna-ai
Release 6.5.4
What's Changed
- Improved overall client stability: The page no longer reloads for basic actions; instead, it dynamically updates content from the server
- Enhanced config loading process and added a new settings block
- All custom settings are now saved in LocalStorage for improved user experience
- Removed news toggle (temporarily) and replaced it with a custom config toggle
- Generalized and removed duplicated functions for better code organization
- Updated sendMessage() function to use custom config
New Features
- Implemented logic for regenerate and copy text buttons for each message
- Added last_n_tokens_size, use_mmap, and offload_kqv parameters to the config for more customization options
- Integrated official Huggingface embeddings model for improved RAG (Retrieval-Augmented Generation) implementation
Bug Fixes
- Fixed function calling in HTML
- Resolved issues with tabs in the Creator Studio
- Corrected variable names and grammar errors throughout the codebase
- Attempted to fix Kanojo-related issues (work in progress)
Release 6.5.3
What's Changed
- Update Dockerfiles for new location by @0xGingi in #102
- Release 6.5.3 - Docker Bug Fix and README update by @yukiarimo in #103
Full Changelog: v6.5.2...v6.5.3
Release 6.5.2
What's Changed
- Docker + Hard Coded Names Backend Removal by @0xGingi in #96
- docs: update README.md by @eltociear in #100
- Release 6.5.2 by @yukiarimo in #101
What's New:
- Fixed the bugs in audio route for audio processing
- Added more audio modes:
siri-pv
andnative
- Temporarily removed the emotional stuff (will be re-implemented with Himitsu)
- Improved the setup index.sh file
- Changed the
.gitignore
file to support new directories - Created a new function for general config fetching
get_config
and added config as param for functions where possible - Fixed the hard-coded dependencies across multiple modules
- Fixed file upload and added RAG and Direct Q&A for it
- Moved some garbage around
- Removed unnecessary prints and console logs
- Speeded up the initial launch by annihilating the unnecessary dependencies based on the user's configuration
- Added some new params to the
config.json
- Fixed the camera view: now it's mirrored nicely
- Changed the shortcuts for navigation to:
tab
-> open/close the sidebar andcommand+1-6
for the tabs - Improved the basic stuff
- Fixed bug with no enter in textarea
- Removed unnecessary code
- Full Docker Support for Windows and Linux
- Changed Creator Studio modes to: Naked (default), Presentation, and Article
- Added caching to speed up the front
- Used flex for messages (improves the location of the image on the PC)
- Stable call Q&A for images with voice input
- Removed flash messages support
- Added the notifications manager
New Contributors
- @eltociear made their first contribution in #100
Full Changelog: v6.5.1...v6.5.2
Release 6.5.1
What's new:
- New WebUI with a modern look
- Delay timeouts reduced to speed up the frontend
- CSS minified
- Stabilized JS scripts
Release 6.5
New Features:
- Rewritten Prompt Templates from scratch with some new ones
- Full Kanojo and Prompt template working system
- Text Streaming
- New model for image uploading with X5 speed-up
Beta Features:
- Creator Studio Freeform
Bug fixes:
- Audio route is not working
- History management and Saving logic fixed. Cookie loading is available as a request
- Faster full-stack loading, and all images are now in WebP
- Additional stability improvements
Removed Stuff:
- Removed support for Electron App
For more info, refer to our roadmap: #91
Full Changelog: v6.3.1...v6.5
Release 6.4
What's Changed
- Cleaned up test files
- Fixed routing
- Improved documentation with new data, updated images, and backstory
- Shell script returned to replace GUI for
menu.py
because it doesn't work - Updated images and README
- Added if else statements for stability and not mandatory dependencies based on the configuration
New Features
- AGI Worker
- Added search route for web searching
- Optimized workflow
- Added full support for native and fast audio modes
Bug Fixes
- Trash Picked Up
- Certificated issues fixed
- Fixed requirements
- Fixed TTS request form on the client
- Normalized model generation config params
Full Changelog: v6.3.1...v6.4
Release 6.3.1
What's Changed:
- v6.3.1 by @yukiarimo in #85
- Added WaifuCard
- Added BuyMeACoffee donation option
- Clean up the unnecessary dependencies
- Added Button for content creation
- PrompTemplates moved to the top bar
- Easy local kanojo management
- Removed WASM stuff
- General improvements
- Fixed "magic" bugs
Stil WIP:
- No config.json modification for the individual users from the client
- Book generation and text/img-to-video is not available for now
- Kanojo and history loader from the downloaded file
Full Changelog: v6.3...v6.3.1
Release 6.3
What's new:
- Stable support for video calls
- Moondream is back (directly imported, aiflow in the future)
- HTTPS fixes for non-local runs
- General bug fixes and optimization
- Removed unnecessary dependencies
- Added "some logic"
Still should work on:
- Model generation speed
- More AGI capabilities
- Server asynchronous support (Flask is fine; too early for the FastAPI)
- Deployment part
Release 6.2.1
What's new?:
- Stable Diffusion image generation full support
- Image recognition with updated Moondream 2 model in full support
- Improved history management
- X2 cold speed up
- Bug fixes
- Lighten WebUI stuff
- New background
- Character card management using Kokoro Connect
What's Changed
- Update README.md by @gitmanbloggus in #74
- Remove index.sh by @gitmanbloggus in #75
- v6.2.1 by @yukiarimo in #77
New Contributors
- @gitmanbloggus made their first contribution at #74
Full Changelog: v6.2...v6.2.1
Release 6.2
What's Changed:
- Create requirements-macos.txt by @yukiarimo in #51
- Modified: requirements.txt by @0xGingi in #50
- Add Metal support fully by @0xGingi in #52
- set to use CPU by default + add ctransformers to req by @0xGingi in #54
- Remove hardcoded names by @0xGingi in #55
- merge main into dev by @yukiarimo in #56
- Fix the wrong file name (it wasn't matching with huggingface) by @robonxt-ai in #58
- Config update by @yukiarimo in #59
- Final steps for a new library. v6.1.1 by @yukiarimo in #60
- llama-cpp-python GPU installation fixed + Added Donate by @0xGingi in #62
- v6.1.2 by @yukiarimo in #63
- v6.1.3 - Optimization performance by @yukiarimo in #67
- v6.2 by @yukiarimo in #71
New Features:
- First Kanojo implementation
- Prompt Templates are part of the Kanojo Connect
- Double speed due to the enhanced server stability
- Rewritten UI chat logic from scratch
- Fixed Bootstrap bugs
- Removed unnecessary dependencies
- Optimized images
- Complete PWA support for all platforms
- Auth system 2.0
- Offline mode (beta 1.0)
- Project directory refactor
- Vital configuration tweaks
- Easy installation via updated CLI and TUI
- Updated documentation
Not working:
- Both Himitsu and Himitsu Copilot
- Prompts without history
- Google Search
- Audio and Video calls
- Offline PWA load if using a server-side approach
- Bootstrap bug with toggles
- Light mode
New Contributors
- @robonxt-ai made their first contribution at #58
Full Changelog: v6.1...v6.2