Fix image generation to properly handle diffusers vs GGUF models

- Added check in sd.cpp fallback to skip HF model IDs that are likely diffusers models
- Prevents sd.cpp from trying to download non-GGUF files like .gitattributes for diffusers models
- Tongyi-MAI/Z-Image-Turbo and similar diffusers models now handled correctly by diffusers library
- GGUF models still work with sd.cpp as before

codai/api/images.py

@@ -671,8 +671,19 @@ async def create_image_generation(request: ImageGenerationRequest, http_request:
             elif os.path.isfile(model_to_use):
                 model_path = model_to_use
             else:
-                # Try to resolve as HuggingFace model ID
-                print(f"Trying to resolve as HuggingFace model ID: {model_to_use}")
+                # Check if this is a likely diffusers model (not GGUF)
+                # If it doesn't contain 'gguf' and we're in image generation context,
+                # it's probably a diffusers model that should be handled by the diffusers library
+                is_likely_diffusers = ('gguf' not in model_to_use.lower() and
+                                     not model_to_use.endswith('.gguf'))
+
+                if is_likely_diffusers:
+                    print(f"Model '{model_to_use}' appears to be a diffusers model (not GGUF), skipping sd.cpp resolution")
+                    print("This model should be handled by the diffusers library above")
+                    model_path = None
+                else:
+                    # Try to resolve as HuggingFace GGUF model ID
+                    print(f"Trying to resolve as HuggingFace GGUF model ID: {model_to_use}")
                     try:
                         from huggingface_hub import hf_hub_download, list_repo_files
 
@@ -719,13 +730,12 @@ async def create_image_generation(request: ImageGenerationRequest, http_request:
                                         model_path = hf_hub_download(repo_id=repo_id, filename=gguf_files[0])
                                         print(f"Downloaded to: {model_path}")
                                 else:
-                                # No GGUF files found, try to download the first file as fallback
-                                if files:
-                                    print(f"No GGUF files found, trying to download first file: {files[0]}")
-                                    model_path = hf_hub_download(repo_id=repo_id, filename=files[0])
-                                    print(f"Downloaded to: {model_path}")
+                                    # No GGUF files found - this is not a GGUF model
+                                    print(f"No GGUF files found in {repo_id} - this is not a GGUF model")
+                                    model_path = None
                     except Exception as e:
-                    print(f"Could not resolve as HuggingFace model: {e}")
+                        print(f"Could not resolve as HuggingFace GGUF model: {e}")
+                        model_path = None
             
             if model_path is None:
                 print("Warning: Could not resolve sd.cpp model path via HuggingFace GGUF resolution")