Advanced settings

app.py

@@ -39,17 +39,36 @@ os.makedirs(os.path.dirname(PREFERENCES_FILE), exist_ok=True)
 # Define model pricing information (approximate costs per 100 image API calls)
 # These are placeholder values - replace with actual pricing
 MODEL_PRICING = {
-    "google/gemini-2.0-flash-001": {"cost": "$0.50", "speed": "Fast"},
-    "gpt-4.1-mini": {"cost": "$2.00", "speed": "Medium"},
-    "gpt-4.1": {"cost": "$5.00", "speed": "Medium-Slow"},
-    "anthropic/claude-3.7-sonnet": {"cost": "$4.00", "speed": "Medium"},
-    "google/gemini-2.5-pro-preview-03-25": {"cost": "$1.50", "speed": "Medium"},
-    "google/gemini-2.5-flash-preview:thinking": {"cost": "$0.75", "speed": "Fast"},
-    "gpt-4.1-nano": {"cost": "$1.00", "speed": "Fast"},
-    "openai/chatgpt-4o-latest": {"cost": "$4.50", "speed": "Medium-Slow"},
-    "meta-llama/llama-4-maverick": {"cost": "$0.25", "speed": "Medium"}
+    "google/gemini-2.0-flash-001": {"cost": "$0.50"},
+    "gpt-4.1-mini": {"cost": "$2.00"},
+    "gpt-4.1": {"cost": "$5.00"},
+    "anthropic/claude-3.7-sonnet": {"cost": "$4.00"},
+    "google/gemini-2.5-pro-preview-03-25": {"cost": "$1.50"},
+    "google/gemini-2.5-flash-preview:thinking": {"cost": "$0.75"},
+    "gpt-4.1-nano": {"cost": "$1.00"},
+    "openai/chatgpt-4o-latest": {"cost": "$4.50"},
+    "meta-llama/llama-4-maverick": {"cost": "$0.25"}
 }
 
+# Define model lists globally to avoid scope issues
+PREFERRED_MODELS = [
+    ("Gemini 2.0 Flash (default)", "google/gemini-2.0-flash-001"),
+    ("GPT-4.1 Mini", "gpt-4.1-mini"),
+    ("GPT-4.1", "gpt-4.1"),
+    ("Claude 3.7 Sonnet", "anthropic/claude-3.7-sonnet"),
+    ("Gemini 2.5 Pro", "google/gemini-2.5-pro-preview-03-25"),
+    ("Gemini 2.5 Flash Thinking", "google/gemini-2.5-flash-preview:thinking")
+]
+
+ADDITIONAL_MODELS = [
+    ("GPT-4.1 Nano", "gpt-4.1-nano"),
+    ("ChatGPT Latest", "openai/chatgpt-4o-latest"),
+    ("Llama 4 Maverick", "meta-llama/llama-4-maverick")
+]
+
+# Default model
+DEFAULT_MODEL = "google/gemini-2.0-flash-001"
+
 def get_sys_prompt(length="medium", photograph=False):
 
     extra_prompt = ""
@@ -95,6 +114,36 @@ def get_base_filename(filepath):
     filename = os.path.splitext(basename)[0]
     return filename
 
+# Functions for model selection outside the create_demo scope to avoid errors
+def update_model_dropdown(show_all, current_model):
+    """Update the model dropdown based on the checkbox state"""
+    if show_all:
+        # Combine preferred and additional models
+        all_models = PREFERRED_MODELS + ADDITIONAL_MODELS
+        return gr.Dropdown.update(choices=all_models, value=current_model)
+    else:
+        # Only show preferred models
+        # If current model is not in preferred, switch to default
+        current_model_in_preferred = any(model[1] == current_model for model in PREFERRED_MODELS)
+        if not current_model_in_preferred:
+            return gr.Dropdown.update(choices=PREFERRED_MODELS, value=DEFAULT_MODEL)
+        return gr.Dropdown.update(choices=PREFERRED_MODELS, value=current_model)
+
+def update_model_info(model_id):
+    """Update the model information display"""
+    # Find the display name for the model ID
+    all_models = PREFERRED_MODELS + ADDITIONAL_MODELS
+    model_name = next((name for name, id in all_models if id == model_id), "Unknown Model")
+    
+    # Get pricing info
+    pricing_info = MODEL_PRICING.get(model_id, {"cost": "Unknown"})
+    
+    # Create markdown text - removed speed information
+    info_text = f"""**Current Model**: {model_name}  
+    **Cost per 100 Images**: {pricing_info['cost']}"""
+    
+    return info_text
+
 # Define the Gradio interface
 def create_demo():
     # --- Reintroduce CSS ---
@@ -146,30 +195,11 @@ def create_demo():
                     file_count="multiple"
                 )
                 
-                # Define preferred and additional models
-                preferred_models = [
-                    ("Gemini 2.0 Flash (default)", "google/gemini-2.0-flash-001"),
-                    ("GPT-4.1 Mini", "gpt-4.1-mini"),
-                    ("GPT-4.1", "gpt-4.1"),
-                    ("Claude 3.7 Sonnet", "anthropic/claude-3.7-sonnet"),
-                    ("Gemini 2.5 Pro", "google/gemini-2.5-pro-preview-03-25"),
-                    ("Gemini 2.5 Flash Thinking", "google/gemini-2.5-flash-preview:thinking")
-                ]
-                
-                additional_models = [
-                    ("GPT-4.1 Nano", "gpt-4.1-nano"),
-                    ("ChatGPT Latest", "openai/chatgpt-4o-latest"),
-                    ("Llama 4 Maverick", "meta-llama/llama-4-maverick")
-                ]
-                
-                # Default model
-                default_model_internal_value = "google/gemini-2.0-flash-001"
-                
                 # Model dropdown (will be updated dynamically)
                 model_choice = gr.Dropdown(
-                    choices=preferred_models, 
+                    choices=PREFERRED_MODELS, 
                     label="Select Model",
-                    value=default_model_internal_value, 
+                    value=DEFAULT_MODEL, 
                     visible=True
                 )
                 
@@ -194,11 +224,12 @@ def create_demo():
                         value="Museum Object"
                     )
                     
-                    # Display current model info
+                    # Display current model info - initial value
+                    initial_model_info = f"""**Current Model**: Gemini 2.0 Flash (default)  
+                    **Cost per 100 Images**: {MODEL_PRICING[DEFAULT_MODEL]["cost"]}"""
+                    
                     model_info = gr.Markdown(
-                        """**Current Model**: Gemini 2.0 Flash  
-                        **Cost per 100 Images**: $0.50  
-                        **Speed**: Fast""",
+                        initial_model_info,
                         label="Model Information"
                     )
                 
@@ -239,36 +270,6 @@ def create_demo():
                 all_images = gr.State([])
                 all_results = gr.State([])
 
-        # --- Functions for model selection and display ---
-        # Update dropdown based on checkbox
-        def update_model_dropdown(show_all, current_model):
-            if show_all:
-                # Combine preferred and additional models
-                all_models = preferred_models + additional_models
-                return gr.Dropdown.update(choices=all_models, value=current_model)
-            else:
-                # Only show preferred models
-                # If current model is not in preferred, switch to default
-                current_model_in_preferred = any(model[1] == current_model for model in preferred_models)
-                if not current_model_in_preferred:
-                    return gr.Dropdown.update(choices=preferred_models, value=default_model_internal_value)
-                return gr.Dropdown.update(choices=preferred_models, value=current_model)
-        
-        # Update model info display
-        def update_model_info(model_id):
-            # Find the display name for the model ID
-            model_name = next((name for name, id in preferred_models + additional_models if id == model_id), "Unknown Model")
-            
-            # Get pricing info
-            pricing_info = MODEL_PRICING.get(model_id, {"cost": "Unknown", "speed": "Unknown"})
-            
-            # Create markdown text
-            info_text = f"""**Current Model**: {model_name}  
-            **Cost per 100 Images**: {pricing_info['cost']}  
-            **Speed**: {pricing_info['speed']}"""
-            
-            return info_text
-        
         # Connect checkbox to update dropdown
         show_all_models.change(
             fn=update_model_dropdown,
@@ -316,7 +317,13 @@ def create_demo():
                     img = Image.open(image_path)
                     prompt0 = prompt_new()
                     model_name = model_choice
-                    result = OR.generate_caption(
+                    client_to_use = OR # Default client
+                    # Add logic here if you need to switch between OR and gemini clients based on model_name
+                    # Example:
+                    # if model_name.startswith("google/gemini") and gemini:
+                    #      client_to_use = gemini
+
+                    result = client_to_use.generate_caption(
                         img, model=model_name, max_image_size=512,
                         prompt=prompt0, prompt_dev=sys_prompt, temperature=1
                     )
@@ -396,6 +403,5 @@ def create_demo():
 
 # Launch the app
 if __name__ == "__main__":
-
     app = create_demo()
     app.launch()