Complete GGUF model support for ComfyUI with three inference modes:
1. GGUF mode - Local quantized models (llama-cpp-python)
2. Transformers mode - Full HuggingFace models
3. Nexa SDK mode - Remote/local GGUF via Nexa SDK service (NEW)
Major Features:
- 🔷 Nexa SDK integration with 3 new nodes
- 📁 ComfyUI /models/LLM directory integration
- ⚙️ Configurable API endpoints
- 🔄 Dual mode: Remote service + Local GGUF files
- 💭 Thinking mode support (DeepSeek-R1, Qwen3-Thinking)
- 📝 Unified output naming: all text outputs as 'context'
- 🖼️ Multi-image analysis (up to 6 images)
- 🎯 System prompt presets
New Nodes:
- Nexa Model Selector
- Nexa Text Generation
- Nexa Service Status
- Text Model Loader
- Text Generation
- Vision Model Loader (GGUF)
- Vision Description (GGUF)
- Vision Model Loader (Transformers)
- Vision Generation (Transformers)
- Multi-Image Analysis
- System Prompt Config
Documentation:
- Complete integration guide
- Usage examples
- API reference
- Troubleshooting guide
Author: weekii <weekii2024@gmail.com>
weekii
4145577458