align local GGUF prompt construction with API behavior for Qwen Image/Edit and Wan T2V/I2V
add text-only handling for Qwen-Image when no image is provided and skip mmproj/chat handler in that path
strengthen zh output enforcement and add second-pass Simplified Chinese normalization
add quoted-text placeholder protection/restore during second-pass normalization
add dedicated zh_normalize local style for stronger Chinese normalization
fix forced text-only model loading so mmproj is not required in second-pass normalization
tighten Wan I2V output format to return prompt body only and avoid analysis/report-style responses
- Fixed issue where Qwen2.5-VL were always loaded in text-only mode even when a valid mmproj file was specified.
- Improved mmproj auto-detection logic.
Improved stability when switching between Qwen3-VL GGUF models
Fixed mmproj reuse issues in local vision models
Refined internal GGUF model lifecycle management
Documentation updates:
- Clarified project scope as a prompt generator for QwenImageEdit and Wan2.2
- Reorganized Credits and Dependencies for clearer attribution
- Updated llama-cpp-python installation notes to reference the JamePeng fork documentation
Internal structure preparation for future backend refactoring
No breaking changes to node interfaces
kantan-kanto