Fix auto mode retry logic and improve error handling

- Fix retry logic bug: only run auto mode once (check for _auto_mode flag)
- Prevent infinite retry loops by preserving retry count across recursive calls
- Add better error handling for pipeline compatibility issues (FrozenDict, scale_factor errors)
- Add helpful troubleshooting messages for diffusers version incompatibilities
- Show retry exhaustion message when all alternative models fail

videogen

@@ -2866,7 +2866,8 @@ def main(args):
         parser.error("the following arguments are required: --prompt")
     
     # Handle auto mode with retry support
-    if getattr(args, 'auto', False):
+    # Only run auto mode if this is not a retry (retry count not set yet)
+    if getattr(args, 'auto', False) and not hasattr(args, '_auto_mode'):
         if not args.prompt:
             parser.error("--auto requires --prompt to analyze")
         args = run_auto_mode(args, MODELS)
@@ -3109,6 +3110,16 @@ def main(args):
                 print(f"   - Check if you can access: https://huggingface.co/{model_id_to_load}")
                 print(f"   - Try with a VPN if HuggingFace is blocked")
                 print(f"   - Check if HF_ENDPOINT is set (for China mirror): {os.environ.get('HF_ENDPOINT', 'not set')}")
+        elif "FrozenDict" in error_str or "scale_factor" in error_str or "has no attribute" in error_str:
+            print(f"❌ Pipeline compatibility error: {model_id_to_load}")
+            print(f"   This model uses a pipeline architecture incompatible with your diffusers version.")
+            print(f"   The model may require a specific diffusers version or different pipeline class.")
+            if debug:
+                print(f"\n   [DEBUG] Compatibility troubleshooting:")
+                print(f"   - Try updating diffusers: pip install --upgrade git+https://github.com/huggingface/diffusers.git")
+                print(f"   - Check the model's documentation for required versions")
+                print(f"   - The model may be incorrectly configured in models.json")
+            print(f"\n   💡 Try a different model with --model <name>")
         else:
             print(f"Model loading failed: {e}")
             if debug:
@@ -3116,6 +3127,16 @@ def main(args):
                 print(f"\n   [DEBUG] Full traceback:")
                 traceback.print_exc()
         
+        # If we've exhausted all retries, exit with error
+        if getattr(args, '_auto_mode', False):
+            retry_count = getattr(args, '_retry_count', 0)
+            max_retries = getattr(args, '_max_retries', 3)
+            alternative_models = getattr(args, '_auto_alternative_models', [])
+            
+            if retry_count >= max_retries or not alternative_models:
+                print(f"\n❌ All model retries exhausted ({retry_count}/{max_retries} attempts)")
+                print(f"   Try searching for alternative models: videogen --search-models <query>")
+        
         sys.exit(1)
     
     timing.end_step()  # model_loading