Devlogs
AI Runner 4.1.2 now available
Posted April 14, 2025 by Capsize Games
#ai #llm #stable diffusion #art #ai art #artificial intelligence #chatbot #chat #local llm #privacy #private #offline #local #desktop app #local ai #offline ai #uncensored ai
Iβm excited to tell you that AI Runner v4.1.2 is now available to download. Itβs been 3 months since the last release and much has changed. Letβs dive right in to the changelog:
π§ AI & Model Improvements
- π£οΈ New TTS & STT Support: Added OpenVoice, SpeechT5 voice selection, and fallback to eSpeak when others fail.
- π§ Mistral Support: Integrated Mistral 8B agent with enhanced memory and conversation handling.
- π§ Better RAG Handling: Improved performance, caching, and reloading of RAG (Retrieval-Augmented Generation).
- ποΈ Agent Tools Upgraded: ReActAgentTool, SQLite-backed memory, dynamic UI rendering, and better agent abstraction.
- π¨ Improved SDXL & Turbo Handling: Fixes for inpainting, model scanning, and better support for SDXL Turbo.
π₯οΈ GUI & UX Enhancements
- ποΈ System Tray Support: Minimize AI Runner to the system tray with smooth show/hide transitions.
- πͺ Improved Load & Start Time: Faster GUI loading and startup experience.
- π¨ Canvas Overhaul: Smooth scrolling, dragging, and better grid alignment for a more fluid visual interface.
- π±οΈ Icon & UI Polish: Fixed cursor switching, better tab handling, and collapsible panels behaving properly.
- π New Settings Views: Moved GUI elements to
gui/
folder, cleaner styles, and dynamic sliders.
π³ Docker & Packaging Overhaul
- π¦ Massive Docker Rework: Rebuilt the Docker system for cross-platform builds with reduced image sizes.
- π§ Linux Build Support: Official Linux builds now working with correct user permissions and paths.
- π§ New Dev Tools: Added developer-friendly scripts, setup automation, and better local install instructions.
π οΈ Refactoring & Code Quality
- π§Ή LLM, TTS, STT Refactors: Modularized handlers, improved type hints, and added full test coverage.
- π§ͺ Expanded Tests: Increased test coverage across key modules including agents, filters, and widgets.
- π§Ό General Cleanup: Removed hundreds of unused imports, dead files, and unnecessary variables.
- π Improved Documentation: Updated README, CONTRIBUTING guide, and added license info for new models.
π Bug Fixes
- π§ Fixed double responses from agents and LLMs.
- πΌοΈ Fixed image prompt bugs and stable diffusion crashes.
- π Resolved issues with TTS not syncing with STT.
- πͺͺ Corrected handling of usernames, memory overrides, and system prompts.
- π§± Fixed startup crashes related to permissions, paths, and missing packages.
π Infrastructure & Build
- π CI/CD Revamp: Moved all builds to GitHub Actions and added self-hosted runner support.
- π Env Management: Added better handling of
.env
files, removed old secrets, and cleaned up var logic.
- π οΈ Improved Logging & Debugging: Added structured logs, error messages, and log level overrides.
𧬠Experimental & Advanced Features
- π§ͺ Toggle to disable database chat store & setup wizard
- 𧬠Summarization with Summy for conversations
- π§ Dynamic window generation via agents
- π§ͺ ControlNet and Stable Diffusion enhancements
- πͺ Plugin support and new abstractions for LLM tasks
As always, thank you for using AI Runner β and if you havenβt tried it yet, grab the latest version here from Github or the packaged version here on Itch.