v2.0.0

AutoGGUF v2.0.0

Changelog:

• Clipboard support for save/load preset functionality with shift-click option
• Support for shift-clicking to get quantization command
• AUTOGGUF_BACKEND_REPO environment variable for custom GitHub repository fetching
• New HF to GGUF conversion types: tq1_0 and tq2_0
• Updated multiple dependencies:
  • PySide6, PyTorch, Transformers, FastAPI, uvicorn, and other core libraries to their latest compatible versions
• Adjusted monitoring intervals from 0.2s to 0.5s
• Updated copyright year to 2025
• Bundled llama.cpp licensing text in About menu
• Removed x86 build matrix from CI
• Removed Import Model confirmation dialog
• Resolved PySide6 segfault issue
• Fixed error when deleting models from list
• Corrected incorrect menu bar name for Load Preset

First release of the New Year! Check the Planned Features section in the README to learn more about what's happening going forward.

v1.9.1

AutoGGUF v1.9.1

Changelog:

• Support for specifying log directory name using AUTOGGUF_LOG_DIR_NAME environment variable
• Work in progress GGUF merge window
• Support for repository types in HF Transfer utility
• New dequantize_gguf.py script
• Support for MiniCPM3, RWKVv6, OLMoE, IBM Granite, and Jamba in llama.cpp convert scripts (conversion only)
• Add Nuitka build script for Linux
• Updated Finnish and Russian localizations using Claude 3 Opus
• Improved layout of HF Upload window
• Updated gguf library from upstream
• Refactored code to use localizations for menubar
• Renamed imports_and_globals.py to globals.py
• Moved general functions verify_gguf and process_args to globals.py
• Created Plugins class for extensibility
• Updated dependencies:
  • huggingface-hub
  • fastapi (~=0.115.0)
  • setuptools (~=75.1.0)
  • pyside6 (~=6.7.3)
  • uvicorn (~=0.31.0)
• Corrected localization strings and file select types for GGUF merging
• Fix minor errors in build scripts

Quick progress update:

• working on a configuration update next, to allow for more customizability
• there are no binaries or builds yet in the release, working on a fix for segfault (PySide6 issue, resolved)

Full Changelog: v1.9.0...v1.9.1

v1.9.0

AutoGGUF v1.9.0

Changelog:

• Implemented Hugging Face (HF) upload functionality with GUI definitions
• Added RAM and CPU usage graphs to UI
• Input validation using wraps added to UI
• Right-click context menu added to the models list in UI
• Support for iMatrix generation tracking
• GGUF splitting feature added
• Japanese and German localizations updated
• Refactored to move functions out of AutoGGUF to reduce bloat
• Optimized GGUF imports and renamed related modules
• Removed old HFTransfer class
• Adjusted logging strings and updated French and Dutch localizations
• Moved functions into existing classes and files
• Adjusted indeterminate progress bar behavior
• Improved startup time by optimizing default configuration, disabling network fetches for backends/updates
• Removed requests and python-dotenv to reduce size
• Removed comments in requirements.txt and updated its formatting
• Updated fastapi requirement from ~=0.112.2 to ~=0.114.2
• Updated torch requirement from ~=2.4.0 to ~=2.4.1
• Updated setuptools requirement from ~=74.0.0 to ~=74.1.2
• Updated safetensors requirement from ~=0.4.4 to ~=0.4.5
• Updated huggingface-hub requirement from ~=0.24.6 to ~=0.24.7

Full Changelog: v1.8.1...v1.9.0

v1.8.1

AutoGGUF v1.8.1

Changelog:

• AutoFP8 quantization classes and window (currently WIP)
• Minimize/maximize buttons to title bar
• API key authentication support for the local server
• HuggingFace upload/download class
• OpenAPI docs for endpoints
• Added new showcase image
• Replaced Flask with FastAPI and Uvicorn for improved performance
• Moved functions out of AutoGGUF.py into utils.py and TaskListItem.py
• Updated llama.cpp convert scripts
• Improved LoRA conversion process:
  • Allow specifying output path in arguments
  • Removed shutil.move operation
  • Increased max number of LoRA layers
• Changed default port to 7001
• Now binding to localhost (127.0.0.1) instead of 0.0.0.0
• Upadted Spanish localizations
• Updated setuptools requirement from ~=68.2.0 to ~=74.0.0
• Updated .env.example with new configuration parameters
• Fix web page not found error
• Use of proper status in TaskListItem
• Passing quant_threads and Logger parameters to TaskListItem
• Improved window moving smoothness
• Prevent moving window below taskbar
• Optimized imports in various files
• Remove aliased quant types

v1.8.0

AutoGGUF v1.8.0

Changelog:

• .env.example file added
• Sha256 generation support added to build.yml
• Allow importing models from any directory on the system
• Added manual model import functionality
• Verification for manual imports and support for concatenated files
• Implemented plugins feature using importlib
• Configuration options for AUTOGGUF_MODEL_DIR_NAME, AUTOGGUF_OUTPUT_DIR_NAME, and AUTOGGUF_RESIZE_FACTOR added
• Moved get helper functions to utils.py
• Added type hints
• Reformat TaskListItem.py for better readability
• Separate macOS and Linux runs in CI/CD
• Updated .gitignore
• Updated numpy requirement from <2.0.0 to <3.0.0
• Fixed sha256 file format and avoided overwriting
• Updated regex for progress tracking
• Arabic and French localizations fixed
• Only count valid backends instead of total backend combos
• Import missing modules

Full Changelog: v1.7.2...v1.8.0

v1.7.2

AutoGGUF v1.7.2

Changelog:

• Update checking support (controlled by AUTOGGUF_CHECK_UPDATE environment variable)
• Live update support for GPU monitor graphs
• Smoother usage bar changes in monitor
• Unicode X button in KV Overrides box
• PyPI setup script
• Inno Setup build file
• Missing requirements and dotenv file loading
• Moved functions out of AutoGGUF.py
• Relocated CustomTitleBar to separate file
• Updated torch requirement from ~=2.2.0 to ~=2.4.0
• Updated showcase image
• setup.py issues

Full Changelog: v1.7.1...v1.7.2

v1.7.1

AutoGGUF v1.7.1

Changelog:

• Modern UI with seamless title bar (preview of v2.0.0)
• Window resizing shortcuts (Ctrl+, Ctrl-, Ctrl+0)
• Theming support
• CPU usage bar
• Save Preset and Load Preset options in File menu
• Support for EXAONE model type
• Window size configuration through environment variables
• Refactored window to be scrollable
• Moved save/load preset logic to presets.py
• Updated docstrings for AutoGGUF.py, lora_conversion.py, and Logger.py
• Adapted gguf library to project standards
• Updated version to v1.7.0
• Fixed IDE-detected code typos and errors
• Hotfix for quantization and import errors

A setup program created with Inno Setup has been attached below with the -setup suffix.

Full Changelog: v1.7.0...v1.7.1

v1.7.0

AutoGGUF v1.7.0

Changelog:

• Added menu bar with Close and About options
• Introduced program version in localizations.py
• Implemented support for 32-bit builds
• Updated torch requirement from ~=1.13.1 to ~=2.4.0
• Updated psutil requirement from ~=5.9.8 to ~=6.0.0
• Refactored functions out of AutoGGUF.py and moved to ui_update.py
• Changed filenames to follow PEP 8 conventions
• Optimized imports in AutoGGUF.py
• Updated README with new version and styled screenshot
• Fixed image blur in documentation
• Added build workflow to CI/CD
• Build size optimization using GitHub Actions
• Disabled .md and .txt CodeQL analysis
• Added dependency audit to CI/CD
• Implemented radon, dependabot, and pre-commit workflows
• Updated SECURITY.md
• Updated CONTRIBUTING.md with PR instructions

Full Changelog: v1.6.2...v1.7.0

v1.6.2

AutoGGUF v1.6.2

Changelog:

• Server functionality with new endpoints:
  • /v1/backends: Lists all backends and their paths
  • /v1/health: Heartbeat endpoint
  • /v1/tasks: Provides current task info (name, status, progress, log file)
  • /v1/models: Retrieves model details (name, type, path, shard status)
• Environment variable support for server configuration:
  • AUTOGGUF_SERVER: Enable/disable server (true/false)
  • AUTOGGUF_SERVER_PORT: Set server port (integer)
• Updated AutoGGUF docstrings
• Refactored build scripts
• Set GGML types to lowercase in command builder

Full Changelog: v1.6.1...v1.6.2

v1.6.1

AutoGGUF v1.6.1

Changelog:

• Add optimized build scripts
• Use Nutika for building
• Update .gitignore
• Fixes bug where deletion while a task is running crashes the program

Full Changelog: v1.6.0...v1.6.1

To use the build files below, please note that the fast build has a higher unzipped size (97MB) but smaller download (38MB), and the standard build was created with PyInstaller which has a medium download and unzipped size (50MB), although it may be slower.