Add QwenVL/Qwen3 models, PromptForge vocabulary, and progress bars
Local LLM Provider:
- Add 10 QwenVL models (FP8 variants + 32B Instruct/Thinking)
- Add 5 Qwen3 text models (0.6B, 1.7B, 4B, 8B, 4B-2507)
- Add num_beams parameter for beam search
- Add VRAM enforcement with auto-quantization downgrade
- Add custom_models.json support for user-defined models
- Add preset prompts system (Tags, Descriptions, Cinematic, etc.)
Z-Image Vocabulary (from PromptForge):
- Add 96 new terms across 7 new categories
- Camera angles (12): low/high/dutch angle, bird's eye, drone shot
- Portrait lighting (10): loop, clamshell, cross, beauty dish
- Lighting effects (10): god rays, caustics, volumetric fog
- Lens effects (12): bokeh, anamorphic flare, chromatic aberration
- Art styles (16): Renaissance, Baroque, anime, Studio Ghibli
- Themes (20): cyberpunk, cottagecore, dark academia, steampunk
- Material effects (16): holographic, stained glass, liquid chrome
UI Improvements:
- Add visual progress bars to Prompt Enhancer node
- Add visual progress bars to Photography Prompts node
Infrastructure:
- Add GitHub Actions workflow for ComfyUI registry publishing
- Update pyproject.toml with new dependencies
- Update dependency checks in __init__.py
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
