ollamadiffuser 1.1.6__tar.gz → 1.2.1__tar.gz

ollamadiffuser-1.2.1/CHANGELOG.md

@@ -0,0 +1,292 @@
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [1.2.0] - 2025-06-02
+
+### 🚀 Major Features Added
+
+#### ⚡ GGUF Model Support
+- **Quantized Models**: Full support for GGUF (GPT-Generated Unified Format) quantized models
+- **Massive VRAM Reduction**: Run FLUX.1-dev with 3GB VRAM instead of 20GB+ 
+- **7 Quantization Levels**: From q2k (3GB) to f16 (16GB) for different hardware capabilities
+- **Hardware Optimization**: Native CUDA and Metal acceleration support
+- **CPU Fallback**: Automatic CPU inference when VRAM is insufficient
+
+#### 🎛️ GGUF Model Variants
+- **flux.1-dev-gguf-q2k**: Ultra-low VRAM (3GB) for testing and low-end hardware
+- **flux.1-dev-gguf-q3ks**: Balanced option (4GB) for mobile GPUs
+- **flux.1-dev-gguf-q4ks**: **Recommended** (6GB) - best quality/performance balance
+- **flux.1-dev-gguf-q5ks**: High quality (8GB) for mid-range GPUs
+- **flux.1-dev-gguf-q6k**: Near-original quality (10GB) 
+- **flux.1-dev-gguf-q8**: Minimal quality loss (12GB)
+- **flux.1-dev-gguf-f16**: Full precision (16GB)
+
+### 🛠️ Technical Implementation
+
+#### GGUF Engine Integration
+- **Backend**: stable-diffusion.cpp with Python bindings integration
+- **Automatic Detection**: Seamless GGUF model recognition and loading
+- **Memory Management**: Intelligent VRAM usage and CPU offloading
+- **Hardware Acceleration**: CMAKE-based CUDA and Metal compilation support
+
+#### CLI Enhancements
+- **GGUF Check**: `ollamadiffuser registry check-gguf` command for compatibility verification
+- **Model Pull**: Seamless GGUF model downloading with progress tracking
+- **Status Monitoring**: Real-time GGUF support and model status checking
+
+### 🎯 Performance Optimizations
+
+#### Generation Parameters
+- **Optimized Settings**: 4-step generation (FLUX-optimized)
+- **CFG Scale**: guidance_scale=1.0 for best FLUX results
+- **Euler Sampler**: Recommended sampler for GGUF models
+- **Hardware Adaptation**: Automatic parameter adjustment based on available VRAM
+
+#### Memory Efficiency
+- **Smart Loading**: Load only required model components
+- **Progressive Quantization**: Automatic fallback to lower quantization when needed
+- **Resource Management**: Intelligent GPU memory allocation and cleanup
+
+### 📚 Documentation & Guides
+
+#### Comprehensive GGUF Guide
+- **GGUF_GUIDE.md**: Complete 160+ line guide with installation, usage, and troubleshooting
+- **Hardware Recommendations**: Specific guidance for different GPU tiers
+- **Performance Comparisons**: Quality vs speed vs VRAM usage tables
+- **Troubleshooting**: Common issues and solutions for GGUF models
+
+#### Usage Examples
+- **CLI Workflows**: Step-by-step GGUF model usage examples
+- **Python API**: Code examples for programmatic GGUF model usage
+- **Web UI Integration**: Browser-based GGUF model selection and generation
+
+### 🔧 Dependencies & Requirements
+
+#### New Dependencies
+- **stable-diffusion-cpp-python**: Core GGUF inference engine
+- **gguf**: Model format handling and validation
+- **Enhanced OpenCV**: Updated to >=4.8.0 for improved compatibility
+
+#### Hardware Support
+- **NVIDIA CUDA**: CMAKE_ARGS="-DSD_CUDA=ON" installation
+- **Apple Metal**: CMAKE_ARGS="-DSD_METAL=ON" for M1/M2 Macs
+- **CPU Inference**: Full CPU fallback support for any modern processor
+
+### 🎨 User Experience Improvements
+
+#### Accessibility
+- **Low-End Hardware**: Enable FLUX.1-dev on 3GB GPUs (previously impossible)
+- **Faster Downloads**: Reduced model sizes from ~24GB to 3-16GB
+- **Quick Testing**: Instant model switching between quantization levels
+
+#### Web UI Enhancements
+- **GGUF Model Selection**: Dropdown menu with GGUF model variants
+- **VRAM Monitoring**: Real-time memory usage display
+- **Quality Preview**: Visual quality indicators for each quantization level
+
+### 🐛 Bug Fixes & Improvements
+- **Memory Leaks**: Improved GGUF model cleanup and resource management
+- **Error Handling**: Better error messages for GGUF-specific issues
+- **Compatibility**: Enhanced hardware detection and fallback mechanisms
+
+### ⚠️ Breaking Changes
+- **Dependency Requirements**: New GGUF dependencies required for full functionality
+- **Model Loading**: GGUF models use different loading mechanisms than regular models
+
+### 🔄 Migration Guide
+For users upgrading to v1.2.0:
+
+1. **Install GGUF Dependencies**: `pip install stable-diffusion-cpp-python gguf`
+2. **Check Compatibility**: `ollamadiffuser registry check-gguf`
+3. **Download GGUF Model**: `ollamadiffuser pull flux.1-dev-gguf-q4ks`
+4. **Update Hardware Acceleration**: Reinstall with CUDA/Metal support if needed
+
+### 📊 Performance Metrics
+- **VRAM Reduction**: Up to 85% reduction (20GB → 3GB)
+- **File Size**: Up to 87% smaller downloads (24GB → 3GB)
+- **Generation Speed**: Comparable or faster due to optimized quantization
+- **Quality Retention**: 90%+ quality retention with q4ks quantization
+
+## [1.1.6] - 2025-5-30
+
+### 🎨 New Features
+
+#### ControlNet Sample Images
+- **New CLI Command**: `ollamadiffuser create-samples` for creating ControlNet demonstration images
+- **Built-in Samples**: Pre-made control images for Canny, Depth, OpenPose, and Scribble controls
+- **Web UI Integration**: Sample images automatically available in the web interface for easy testing
+- **Force Recreation**: `--force` flag to recreate all samples even if they exist
+
+#### Installation Helper
+- **New Script**: `install_helper.py` for platform-specific installation guidance
+- **Shell Detection**: Automatically detects user's shell (bash, zsh, fish) and provides correct install syntax
+- **Multiple Installation Options**: Basic, Full, and Development installation commands
+- **Interactive Installation**: Option to install directly from the helper script
+
+### 🛠️ Improvements
+
+#### CLI Enhancements
+- **Progress Tracking**: Enhanced download progress display with Ollama-style formatting
+- **Better Error Handling**: Improved error messages and graceful failure modes
+- **Warning Fixes**: Resolved various CLI warnings and edge cases
+
+#### Web UI Enhancements
+- **Sample Image Gallery**: Built-in ControlNet samples with 3 images per control type
+- **Automatic Sample Creation**: Samples generated automatically when needed
+- **Better UX**: Visual samples make ControlNet testing more intuitive
+
+### 🐛 Bug Fixes
+- **Version Inconsistencies**: Fixed version numbering across different components
+- **Installation Issues**: Resolved shell-specific installation syntax problems
+- **CLI Warnings**: Fixed various warning messages and edge cases
+- **Sample Generation**: Improved reliability of sample image creation
+
+### 📦 Technical Changes
+- **MANIFEST.in**: Updated to include sample images and static files
+- **Dependencies**: Refined dependency management for better compatibility
+- **Shell Compatibility**: Better support for zsh, fish, and bash shells
+
+### 🎯 Sample Images Created
+- **Canny Control**: Geometric shapes, house outline, portrait silhouette (3 samples)
+- **Depth Control**: Depth map variations for different scene types (3 samples)
+- **OpenPose Control**: Human pose variations for different positions (3 samples)
+- **Scribble Control**: Hand-drawn style sketches and outlines (3 samples)
+
+## [1.1.0] - 2025-5-29
+
+### 🚀 Major Features Added
+
+#### ⚡ Lazy Loading Architecture
+- **Instant Startup**: Application now starts immediately without downloading ControlNet models
+- **On-Demand Loading**: ControlNet preprocessors initialize only when actually needed
+- **Performance Boost**: `ollamadiffuser --help` runs in milliseconds instead of 30+ seconds
+- **Memory Efficient**: No unnecessary model downloads for users who don't use ControlNet
+
+#### 🎛️ Complete ControlNet Integration
+- **6 ControlNet Models**: SD 1.5 and SDXL variants (canny, depth, openpose, scribble)
+- **10 Control Types**: canny, depth, openpose, hed, mlsd, normal, lineart, lineart_anime, shuffle, scribble
+- **Advanced Preprocessors**: Full controlnet-aux integration with graceful fallbacks
+- **Web UI Integration**: File upload, preprocessing, and side-by-side result display
+- **REST API Support**: Complete API endpoints for ControlNet generation and preprocessing
+
+#### 🔄 Enhanced LoRA Management
+- **Web UI Integration**: Download LoRAs directly from Hugging Face in the browser
+- **Alias Support**: Create custom names for your LoRAs
+- **Strength Control**: Adjust LoRA influence with intuitive sliders
+- **Real-time Loading**: Load/unload LoRAs without restarting the application
+
+### 🛠️ Technical Improvements
+
+#### ControlNet Preprocessor Manager
+- **Lazy Initialization**: `ControlNetPreprocessorManager` with `is_initialized()`, `is_available()`, `initialize()` methods
+- **Automatic Fallback**: Basic OpenCV processors when advanced ones fail
+- **Error Handling**: Robust validation and graceful degradation
+- **Status Tracking**: Real-time initialization and availability status
+
+#### Web UI Enhancements
+- **ControlNet Section**: Dedicated controls with status indicators
+- **Initialization Button**: Manual preprocessor initialization for faster processing
+- **File Upload**: Drag-and-drop control image upload with validation
+- **Responsive Design**: Mobile-friendly interface with adaptive layouts
+- **Real-time Status**: Live model, LoRA, and ControlNet status indicators
+
+#### API Improvements
+- **New Endpoints**: `/api/controlnet/initialize`, `/api/controlnet/preprocessors`, `/api/controlnet/preprocess`
+- **File Upload Support**: Multipart form data handling for control images
+- **Status Endpoints**: Check ControlNet availability and initialization status
+- **Error Handling**: Comprehensive error responses with helpful messages
+
+### 📦 Dependencies Updated
+- **controlnet-aux**: Added `>=0.0.7` for advanced preprocessing capabilities
+- **opencv-python**: Added `>=4.8.0` for basic image processing fallbacks
+- **diffusers**: Updated to `>=0.26.0` for ControlNet compatibility
+
+### 🎨 User Experience Improvements
+
+#### Startup Performance
+- **Before**: 30+ seconds startup time, 1GB+ automatic downloads
+- **After**: Instant startup, downloads only when needed
+- **User Control**: Choose when to initialize ControlNet preprocessors
+
+#### Web UI Experience
+- **Status Indicators**: Clear visual feedback for all system states
+- **Progressive Loading**: Initialize components as needed
+- **Error Messages**: Helpful guidance for common issues
+- **Mobile Support**: Responsive design works on all devices
+
+#### CLI Experience
+- **Fast Commands**: All CLI commands run instantly
+- **Lazy Loading**: ControlNet models load only when generating
+- **Status Commands**: Check system state without triggering downloads
+
+### 🔧 Configuration Changes
+- **setup.py**: Added ControlNet dependencies
+- **pyproject.toml**: Updated dependency specifications
+- **Model Registry**: Enhanced with ControlNet model definitions
+
+### 📚 Documentation Updates
+- **CONTROLNET_GUIDE.md**: Comprehensive 400+ line guide with examples
+- **README.md**: Updated with lazy loading features and ControlNet quick start
+- **API Documentation**: Complete endpoint reference with examples
+
+### 🐛 Bug Fixes
+- **Startup Crashes**: Fixed 404 errors from non-existent model repositories
+- **Memory Leaks**: Improved cleanup of ControlNet preprocessors
+- **Device Compatibility**: Better handling of CPU/GPU device switching
+- **Error Handling**: More graceful failure modes with helpful messages
+
+### ⚠️ Breaking Changes
+- **Import Behavior**: `controlnet_preprocessors` module no longer auto-initializes
+- **API Changes**: Some ControlNet endpoints require explicit initialization
+
+### 🔄 Migration Guide
+For users upgrading from v1.0.x:
+
+1. **No Action Required**: Lazy loading is automatic and transparent
+2. **Web UI**: ControlNet preprocessors initialize automatically when uploading images
+3. **API Users**: Call `/api/controlnet/initialize` for faster subsequent processing
+4. **Python API**: Use `controlnet_preprocessor.initialize()` for batch processing
+
+### 🎯 Performance Metrics
+- **Startup Time**: Reduced from 30+ seconds to <1 second
+- **Memory Usage**: Reduced baseline memory footprint by ~2GB
+- **First Generation**: Slightly slower due to lazy loading, then normal speed
+- **Subsequent Generations**: Same performance as before
+
+## [1.0.0] - 2025-5-28
+
+### Added
+- Initial release with core functionality
+- Support for Stable Diffusion 1.5, SDXL, SD3, and FLUX models
+- Basic LoRA support
+- CLI interface
+- REST API server
+- Web UI interface
+- Model management system
+
+### Features
+- Model downloading and management
+- Image generation with various parameters
+- Multiple interface options (CLI, API, Web UI)
+- Hardware optimization (CUDA, MPS, CPU)
+- Safety checker bypass for creative freedom
+
+---
+
+## Development Notes
+
+### Version Numbering
+- **Major** (X.0.0): Breaking changes, major feature additions
+- **Minor** (1.X.0): New features, significant improvements
+- **Patch** (1.1.X): Bug fixes, minor improvements
+
+### Release Process
+1. Update version in `__init__.py`
+2. Update CHANGELOG.md with new features
+3. Update documentation
+4. Create release tag
+5. Deploy to package repositories 

{ollamadiffuser-1.1.6/ollamadiffuser.egg-info → ollamadiffuser-1.2.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ollamadiffuser
-Version: 1.1.6
+Version: 1.2.1
 Summary: 🎨 Local AI Image Generation with Ollama-style CLI for Stable Diffusion, FLUX.1, and LoRA support
 Home-page: https://github.com/ollamadiffuser/ollamadiffuser
 Author: OllamaDiffuser Team
@@ -52,15 +52,16 @@ Requires-Dist: psutil>=5.9.0
 Requires-Dist: jinja2>=3.0.0
 Requires-Dist: peft>=0.10.0
 Requires-Dist: numpy>=1.21.0
+Requires-Dist: controlnet-aux>=0.0.7
+Requires-Dist: opencv-python>=4.8.0
+Requires-Dist: stable-diffusion-cpp-python>=0.1.0
+Requires-Dist: gguf>=0.1.0
 Provides-Extra: dev
 Requires-Dist: pytest>=7.0.0; extra == "dev"
 Requires-Dist: pytest-asyncio>=0.21.0; extra == "dev"
 Requires-Dist: black>=23.0.0; extra == "dev"
 Requires-Dist: isort>=5.12.0; extra == "dev"
 Requires-Dist: flake8>=6.0.0; extra == "dev"
-Provides-Extra: full
-Requires-Dist: controlnet-aux>=0.0.7; extra == "full"
-Requires-Dist: opencv-python>=4.8.0; extra == "full"
 Dynamic: author
 Dynamic: home-page
 Dynamic: license-file
@@ -81,11 +82,69 @@ Dynamic: requires-python
 
 ---
 
+## 🔑 Hugging Face Authentication
+
+**Do you need a Hugging Face token?** It depends on which models you want to use!
+
+### 🟢 Models that DON'T require a token:
+- **FLUX.1-schnell** - Apache 2.0 license, ready to use ✅
+- **Stable Diffusion 1.5** - Basic model, no authentication needed ✅
+- **Most ControlNet models** - Generally public access ✅
+
+### 🟡 Models that DO require a token:
+- **FLUX.1-dev** - Requires HF token and license agreement ⚠️
+- **Stable Diffusion 3.5** - Requires HF token and license agreement ⚠️
+- **Some premium LoRAs** - Gated models from Hugging Face ⚠️
+
+### 🚀 Quick Setup
+
+**For basic usage** (no token needed):
+```bash
+# These work immediately without any setup:
+ollamadiffuser pull flux.1-schnell
+ollamadiffuser pull stable-diffusion-1.5
+```
+
+**For advanced models** (token required):
+```bash
+# 1. Set your token
+export HF_TOKEN=your_token_here
+
+# 2. Now you can access gated models
+ollamadiffuser pull flux.1-dev
+ollamadiffuser pull stable-diffusion-3.5-medium
+```
+
+### 🔧 How to get a Hugging Face token:
+
+1. **Create account**: Visit [huggingface.co](https://huggingface.co) and sign up
+2. **Generate token**: Go to Settings → Access Tokens → Create new token
+3. **Accept licenses**: Visit the model pages and accept license agreements:
+   - [FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev)
+   - [Stable Diffusion 3.5](https://huggingface.co/stabilityai/stable-diffusion-3.5-medium)
+4. **Set environment variable**:
+   ```bash
+   # Temporary (current session)
+   export HF_TOKEN=your_token_here
+   
+   # Permanent (add to ~/.bashrc or ~/.zshrc)
+   echo 'export HF_TOKEN=your_token_here' >> ~/.bashrc
+   ```
+
+### 💡 Pro Tips:
+- **Start simple**: Begin with FLUX.1-schnell (no token required, commercial use OK)
+- **Token scope**: Use "read" permissions for downloading models
+- **Privacy**: Your token stays local - never shared with OllamaDiffuser servers
+- **Troubleshooting**: If downloads fail, verify your token and model access permissions
+
+---
+
 ## ✨ Features
 
 - **🚀 Fast Startup**: Instant application launch with lazy loading architecture
 - **🎛️ ControlNet Support**: Precise image generation control with 10+ control types
 - **🔄 LoRA Integration**: Dynamic LoRA loading and management
+- **📦 GGUF Support**: Memory-efficient quantized models (3GB VRAM minimum!)
 - **🌐 Multiple Interfaces**: CLI, Python API, Web UI, and REST API
 - **📦 Model Management**: Easy installation and switching between models
 - **⚡ Performance Optimized**: Memory-efficient with GPU acceleration
@@ -109,6 +168,18 @@ curl -X POST http://localhost:8000/api/generate \
   --output image.png
 ```
 
+### GGUF Quick Start (Low VRAM)
+```bash
+# For systems with limited VRAM (3GB+)
+pip install ollamadiffuser stable-diffusion-cpp-python gguf
+
+# Download memory-efficient GGUF model
+ollamadiffuser pull flux.1-dev-gguf-q4ks
+
+# Generate with reduced memory usage
+ollamadiffuser run flux.1-dev-gguf-q4ks
+```
+
 ### Option 2: Development Installation
 ```bash
 # Clone the repository
@@ -162,12 +233,26 @@ curl -X POST http://localhost:8000/api/generate/controlnet \
 
 Choose from a variety of state-of-the-art image generation models:
 
-| Model | License | Quality | Speed | Commercial Use |
-|-------|---------|---------|-------|----------------|
-| **FLUX.1-schnell** | Apache 2.0 | High | **4 steps** (12x faster) | ✅ Commercial OK |
-| **FLUX.1-dev** | Non-commercial | High | 50 steps | ❌ Non-commercial |
-| **Stable Diffusion 3.5** | CreativeML | Medium | 28 steps | ⚠️ Check License |
-| **Stable Diffusion 1.5** | CreativeML | Fast | Lightweight | ⚠️ Check License |
+| Model | License | Quality | Speed | Commercial Use | VRAM |
+|-------|---------|---------|-------|----------------|------|
+| **FLUX.1-schnell** | Apache 2.0 | High | **4 steps** (12x faster) | ✅ Commercial OK | 20GB+ |
+| **FLUX.1-dev** | Non-commercial | High | 50 steps | ❌ Non-commercial | 20GB+ |
+| **FLUX.1-dev-gguf** | Non-commercial | High | 4 steps | ❌ Non-commercial | **3-16GB** |
+| **Stable Diffusion 3.5** | CreativeML | Medium | 28 steps | ⚠️ Check License | 12GB+ |
+| **Stable Diffusion 1.5** | CreativeML | Fast | Lightweight | ⚠️ Check License | 6GB+ |
+
+### 💾 GGUF Models - Reduced Memory Requirements
+
+**NEW**: GGUF quantized models enable running FLUX.1-dev on budget hardware!
+
+| GGUF Variant | VRAM | Quality | Best For |
+|--------------|------|---------|----------|
+| `flux.1-dev-gguf-q4ks` | 6GB | ⭐⭐⭐⭐ | **Recommended** - RTX 3060/4060 |
+| `flux.1-dev-gguf-q3ks` | 4GB | ⭐⭐⭐ | Mobile GPUs, GTX 1660 Ti |
+| `flux.1-dev-gguf-q2k` | 3GB | ⭐⭐ | Entry-level hardware |
+| `flux.1-dev-gguf-q6k` | 10GB | ⭐⭐⭐⭐⭐ | RTX 3080/4070+ |
+
+📖 **[Complete GGUF Guide](GGUF_GUIDE.md)** - Hardware recommendations, installation, and optimization tips
 
 ### Why Choose FLUX.1-schnell?
 - **Apache 2.0 license** - Perfect for commercial use
@@ -305,6 +390,11 @@ else:
 - **Stable Diffusion 3**: Latest architecture
 - **FLUX.1**: State-of-the-art quality
 
+### GGUF Quantized Models
+- **FLUX.1-dev GGUF**: 7 quantization levels (3GB-16GB VRAM)
+- **Memory Efficient**: Run high-quality models on budget hardware
+- **Same API**: Works seamlessly with existing commands
+
 ### ControlNet Models
 - **SD 1.5 ControlNet**: 4 control types (canny, depth, openpose, scribble)
 - **SDXL ControlNet**: 2 control types (canny, depth)
@@ -338,6 +428,19 @@ image = engine.generate_image(
 )
 ```
 
+### GGUF Model Usage
+```bash
+# Check GGUF support
+ollamadiffuser registry check-gguf
+
+# Download GGUF model for your hardware
+ollamadiffuser pull flux.1-dev-gguf-q4ks  # 6GB VRAM
+ollamadiffuser pull flux.1-dev-gguf-q3ks  # 4GB VRAM
+
+# Use with optimized settings
+ollamadiffuser run flux.1-dev-gguf-q4ks
+```
+
 ### Batch Processing
 ```python
 from ollamadiffuser.core.utils.controlnet_preprocessors import controlnet_preprocessor
@@ -375,6 +478,7 @@ with open("control.jpg", "rb") as f:
 
 ## 📚 Documentation & Guides
 
+- **[GGUF Models Guide](GGUF_GUIDE.md)**: Complete guide to memory-efficient GGUF models
 - **[ControlNet Guide](CONTROLNET_GUIDE.md)**: Comprehensive ControlNet usage and examples
 - **[Website Documentation](https://www.ollamadiffuser.com/)**: Complete tutorials and guides
 
@@ -386,10 +490,17 @@ with open("control.jpg", "rb") as f:
 - **Python**: 3.8+
 
 ### Recommended Hardware
+
+#### For Regular Models
 - **GPU**: 8GB+ VRAM (NVIDIA/AMD)
 - **RAM**: 16GB+ system RAM
 - **Storage**: SSD with 50GB+ free space
 
+#### For GGUF Models (Memory Efficient)
+- **GPU**: 3GB+ VRAM (or CPU only)
+- **RAM**: 8GB+ system RAM (16GB+ for CPU inference)
+- **Storage**: SSD with 20GB+ free space
+
 ### Supported Platforms
 - **CUDA**: NVIDIA GPUs (recommended)
 - **MPS**: Apple Silicon (M1/M2/M3)
@@ -420,6 +531,18 @@ pip install "ollamadiffuser[full]"
 pip install 'ollamadiffuser[full]'
 ```
 
+#### GGUF Support Issues
+```bash
+# Install GGUF dependencies
+pip install stable-diffusion-cpp-python gguf
+
+# Check GGUF support
+ollamadiffuser registry check-gguf
+
+# See full GGUF troubleshooting guide
+# Read GGUF_GUIDE.md for detailed troubleshooting
+```
+
 #### Complete Dependency Check
 ```bash
 # Run comprehensive system diagnostics
@@ -474,6 +597,10 @@ curl -X POST http://localhost:8000/api/controlnet/initialize
 
 #### Memory Issues
 ```bash
+# Use GGUF models for lower memory usage
+ollamadiffuser pull flux.1-dev-gguf-q4ks  # 6GB VRAM
+ollamadiffuser pull flux.1-dev-gguf-q3ks  # 4GB VRAM
+
 # Use smaller image sizes via API
 curl -X POST http://localhost:8000/api/generate \
   -H "Content-Type: application/json" \
@@ -492,12 +619,18 @@ curl -X POST http://localhost:8000/api/generate \
 # If you encounter OpenCV issues on Apple Silicon
 pip uninstall opencv-python
 pip install opencv-python-headless>=4.8.0
+
+# For GGUF Metal acceleration
+CMAKE_ARGS="-DSD_METAL=ON" pip install stable-diffusion-cpp-python
 ```
 
 #### Windows
 ```bash
 # If you encounter build errors
 pip install --only-binary=all opencv-python>=4.8.0
+
+# For GGUF CUDA acceleration
+CMAKE_ARGS="-DSD_CUDA=ON" pip install stable-diffusion-cpp-python
 ```
 
 #### Linux
@@ -540,6 +673,8 @@ This project is licensed under the MIT License - see the [LICENSE](LICENSE) file
 ## 🙏 Acknowledgments
 
 - **Stability AI**: For Stable Diffusion models
+- **Black Forest Labs**: For FLUX.1 models
+- **city96**: For FLUX.1-dev GGUF quantizations
 - **Hugging Face**: For model hosting and diffusers library
 - **ControlNet Team**: For ControlNet architecture
 - **Community**: For feedback and contributions