add photograph

app-Copy1.py

@@ -1,387 +0,0 @@
-import gradio as gr
-import numpy as np
-from PIL import Image
-import io
-import os
-import requests
-import json
-from dotenv import load_dotenv
-import openai
-import base64
-import csv
-import tempfile
-import datetime
-
-# Load environment variables from .env file if it exists (for local development)
-# On Hugging Face Spaces, the secrets are automatically available as environment variables
-if os.path.exists(".env"):
-    load_dotenv()
-
-from io import BytesIO
-import numpy as np
-import requests
-from PIL import Image
-
-# import libraries
-from library.utils_model import *
-from library.utils_html import *
-from library.utils_prompt import *
-
-OR = OpenRouterAPI()
-gemini = OpenRouterAPI(api_key = os.getenv("GEMINI_API_KEY"),base_url="https://generativelanguage.googleapis.com/v1beta/openai/")
-
-# Path for storing user preferences
-PREFERENCES_FILE = "data/user_preferences.csv"
-
-# Ensure directory exists
-os.makedirs(os.path.dirname(PREFERENCES_FILE), exist_ok=True)
-
-def get_sys_prompt(length="medium"):
-    if length == "short":
-        dev_prompt = """You are a museum curator tasked with generating alt-text (as defined by W3C) of museum objects for visually impaired and blind users from images. Use British English and follow museum accessibility best practices. Do not start with phrases like 'The image shows' or 'This is an image of'. Be precise, concise and avoid filler and subjective statements. Repsonses should be a maximum of 130 characters."""
-    elif length == "medium":
-        dev_prompt = """You are a museum curator tasked with generating long descriptions (as defined in W3C) of museum objects for visually impaired and blind users from images. Use British English and follow museum accessibility best practices. Do not start with phrases like 'The image shows' or 'This is an image of'. Be precise, concise and avoid filler and subjective statements. Repsonses should be between 250-300 characters in length."""
-    else:
-        dev_prompt = """You are a museum curator tasked with generating long descriptions (as defined in W3C) of museum objects for visually impaired and blind users from images. Use British English and follow museum accessibility best practices. Do not start with phrases like 'The image shows' or 'This is an image of'. Be precise, concise and avoid filler and subjective statements. Repsonses should be a maxium of 450 characters."""
-    return dev_prompt
-
-def create_csv_file_simple(results):
-    """Create a CSV file from the results and return the path"""
-    # Create a temporary file
-    fd, path = tempfile.mkstemp(suffix='.csv')
-    
-    with os.fdopen(fd, 'w', newline='') as f:
-        writer = csv.writer(f)
-        # Write header
-        writer.writerow(['image_id', 'content'])
-        # Write data
-        for result in results:
-            writer.writerow([
-                result.get('image_id', ''),
-                result.get('content', '')
-            ])
-    
-    return path
-
-# Extract original filename without path or extension
-def get_base_filename(filepath):
-    if not filepath:
-        return ""
-    # Get the basename (filename with extension)
-    basename = os.path.basename(filepath)
-    # Remove extension
-    filename = os.path.splitext(basename)[0]
-    return filename
-
-custom_css = """
-.image-container img {
-    object-fit: contain;
-    width: 100%;
-    height: 100%;
-}
-"""
-
-# Define the Gradio interface
-def create_demo():
-    with gr.Blocks(theme=gr.themes.Monochrome(),css=custom_css) as demo:
-        # Replace the existing logo code section:
-        with gr.Row():
-            with gr.Column(scale=3):
-                gr.Markdown("# MATCHA: Museum Alt-Text for Cultural Heritage with AI 🍵 🌿")
-                gr.Markdown("Upload one or more images to generate accessible alternative text (designed to meet WCAG Guidelines)")
-                gr.Markdown("Developed by the Natural History Museum in Partnership with National Museums Liverpool. Funded by the DCMS Pilot Scheme")
-            with gr.Column(scale=1):
-                with gr.Row():
-                    # Use gr.Image with all interactive features disabled
-                    gr.Image("images/nhm_logo.png", show_label=False, height=120, 
-                            interactive=False, show_download_button=False, 
-                            show_share_button=False, show_fullscreen_button=False,
-                            container=False)
-                    gr.Image("images/nml_logo.png", show_label=False, height=120, 
-                            interactive=False, show_download_button=False,
-                            show_share_button=False, show_fullscreen_button=False,
-                            container=False)
-        
-        with gr.Row():
-            # Left column: Controls and uploads
-            with gr.Column(scale=1):
-                # Upload interface
-                upload_button = gr.UploadButton(
-                    "Click to Upload Images", 
-                    file_types=["image"], 
-                    file_count="multiple"
-                )
-                
-                # Define choices as a list of tuples: (Display Name, Internal Value)
-                model_choices = [
-                    # Gemini
-                    ("Gemini 2.0 Flash (default)", "google/gemini-2.0-flash-001"),
-                    # GPT-4.1 Series
-                    ("GPT-4.1 Nano", "gpt-4.1-nano"),
-                    ("GPT-4.1 Mini", "gpt-4.1-mini"),
-                    ("GPT-4.1", "gpt-4.1"),
-                    ("ChatGPT Latest", "openai/chatgpt-4o-latest"),
-                    # Other Models
-                    ("Claude 3.7 Sonnet", "anthropic/claude-3.7-sonnet"),
-                    ("Llama 4 Maverick", "meta-llama/llama-4-maverick"),
-                    # Experimental Models
-                    ("Gemini 2.5 Pro (Experimental, limited)", "gemini-2.5-pro-exp-03-25"),
-                    ("Gemini 2.0 Flash Thinking (Experimental, limited)", "gemini-2.0-flash-thinking-exp-01-21")
-                ]
-                
-                # Find the internal value of the default choice
-                default_model_internal_value = "google/gemini-2.0-flash-001"
-                
-                # Add model selection dropdown
-                model_choice = gr.Dropdown(
-                    choices=model_choices,
-                    label="Select Model",
-                    value=default_model_internal_value, # Use the internal value for the default
-                    # info="Choose the language model to use." # Optional: Add extra info tooltip
-                    visible=True
-                )
-
-                
-                # Add response length selection
-                length_choice = gr.Radio(
-                    choices=["short", "medium", "long"],
-                    label="Response Length",
-                    value="medium",
-                    info="Short: max 130 chars | Medium: 250-300 chars | Long: max 450 chars"
-                )
-                
-                # Preview gallery for uploaded images
-                gr.Markdown("### Uploaded Images")
-                input_gallery = gr.Gallery(
-                    label="",
-                    columns=3, 
-                    height=150,
-                    object_fit="contain"
-                )
-                
-                # Analysis button
-                analyze_button = gr.Button("Generate Alt-Text", variant="primary", size="lg")
-                
-                # Hidden state component to store image info
-                image_state = gr.State([])
-                filename_state = gr.State([])
-                
-                # CSV download component
-                csv_download = gr.File(label="CSV Results")
-            
-            # Right column: Display area
-            with gr.Column(scale=2):
-                with gr.Column(elem_classes="image-container"):
-                    current_image = gr.Image(
-                        label="Current Image",
-                        height=600,         # Set the maximum desired height
-                        width=1000,
-                        type="filepath",
-                        show_fullscreen_button=True,
-                        show_download_button=False,
-                        show_share_button=False,
-                        elem_classes="image-container"
-                    )
-                
-                # Navigation row
-                with gr.Row():
-                    prev_button = gr.Button("← Previous", size="sm")
-                    image_counter = gr.Markdown("", elem_id="image-counter")
-                    next_button = gr.Button("Next →", size="sm")
-                
-                # Alt-text heading and output
-                gr.Markdown("### Generated Alt-text")
-                
-                # Alt-text
-                analysis_text = gr.Textbox(
-                    label="",
-                    value="Upload images and select model to generate alt-text!",
-                    lines=6,
-                    max_lines=10,
-                    interactive=False,
-                    show_label=False
-                )
-                
-                # Hidden state for gallery navigation
-                current_index = gr.State(0)
-                all_images = gr.State([])
-                all_results = gr.State([])
-        
-        # Handle file uploads - store files for use during analysis
-        def handle_upload(files):
-            file_paths = []
-            file_names = []
-            for file in files:
-                file_paths.append(file.name)
-                # Extract filename without path or extension for later use
-                file_names.append(get_base_filename(file.name))
-            return file_paths, file_paths, file_names
-        
-        upload_button.upload(
-            fn=handle_upload,
-            inputs=[upload_button],
-            outputs=[input_gallery, image_state, filename_state]
-        )
-        
-        # Function to analyze images
-        # Modify the analyze_images function in your code:
-
-        def analyze_images(image_paths, model_choice, length_choice, filenames):
-            if not image_paths:
-                return [], [], 0, "", "No images", "", ""
-            
-            # Get system prompt based on length selection
-            sys_prompt = get_sys_prompt(length_choice)
-            
-            image_results = []
-            
-            for i, image_path in enumerate(image_paths):
-                # Use original filename as image_id if available
-                if i < len(filenames) and filenames[i]:
-                    image_id = filenames[i]
-                else:
-                    image_id = f"Image {i+1}"
-                
-                try:
-                    # Open the image file for analysis
-                    img = Image.open(image_path)
-                    prompt0 = prompt_new()  # Using the new prompt function
-                    
-                    # Extract the actual model name (remove any labels like "(default)")
-                    if " (" in model_choice:
-                        model_name = model_choice.split(" (")[0]
-                    else:
-                        model_name = model_choice
-                    
-                    # Check if this is one of the Gemini models that needs special handling
-                    is_gemini_model = "gemini-2.5-pro" in model_name or "gemini-2.0-flash-thinking" in model_name
-                    
-                    if is_gemini_model:
-                        try:
-                            # First try using the dedicated gemini client
-                            result = gemini.generate_caption(
-                                img, 
-                                model=model_name, 
-                                max_image_size=512,
-                                prompt=prompt0, 
-                                prompt_dev=sys_prompt, 
-                                temperature=1
-                            )
-                        except Exception as gemini_error:
-                            # If gemini client fails, fall back to standard OR client
-                            result = OR.generate_caption(
-                                img, 
-                                model=model_name, 
-                                max_image_size=512,
-                                prompt=prompt0, 
-                                prompt_dev=sys_prompt, 
-                                temperature=1
-                            )
-                    else:
-                        # For all other models, use OR client directly
-                        result = OR.generate_caption(
-                            img, 
-                            model=model_name, 
-                            max_image_size=512,
-                            prompt=prompt0, 
-                            prompt_dev=sys_prompt, 
-                            temperature=1
-                        )
-                    
-                    # Add to results
-                    image_results.append({
-                        "image_id": image_id,
-                        "content": result
-                    })
-                    
-                except Exception as e:
-                    error_message = f"Error: {str(e)}"
-                    image_results.append({
-                        "image_id": image_id,
-                        "content": error_message
-                    })
-            
-            # Create a CSV file for download
-            csv_path = create_csv_file_simple(image_results)
-            
-            # Set up initial display with first image
-            if len(image_paths) > 0:
-                initial_image = image_paths[0]
-                initial_counter = f"{1} of {len(image_paths)}"
-                initial_text = image_results[0]["content"]
-            else:
-                initial_image = ""
-                initial_text = "No images analyzed"
-                initial_counter = "0 of 0"
-            
-            return (image_paths, image_results, 0, initial_image, initial_counter, 
-                    initial_text, csv_path)
-
-        
-        # Function to navigate to previous image
-        def go_to_prev(current_idx, images, results):
-            if not images or len(images) == 0:
-                return current_idx, "", "0 of 0", ""
-            
-            new_idx = (current_idx - 1) % len(images) if current_idx > 0 else len(images) - 1
-            counter_html = f"{new_idx + 1} of {len(images)}"
-            
-            return (new_idx, images[new_idx], counter_html, results[new_idx]["content"])
-        
-        # Function to navigate to next image
-        def go_to_next(current_idx, images, results):
-            if not images or len(images) == 0:
-                return current_idx, "", "0 of 0", ""
-            
-            new_idx = (current_idx + 1) % len(images)
-            counter_html = f"{new_idx + 1} of {len(images)}"
-            
-            return (new_idx, images[new_idx], counter_html, results[new_idx]["content"])
-        
-        # Connect the analyze button
-        analyze_button.click(
-            fn=analyze_images,
-            inputs=[image_state, model_choice, length_choice, filename_state],
-            outputs=[
-                all_images, all_results, current_index, current_image, image_counter,
-                analysis_text, csv_download
-            ]
-        )
-        
-        # Connect navigation buttons
-        prev_button.click(
-            fn=go_to_prev,
-            inputs=[current_index, all_images, all_results],
-            outputs=[current_index, current_image, image_counter, analysis_text]
-        )
-        
-        next_button.click(
-            fn=go_to_next,
-            inputs=[current_index, all_images, all_results],
-            outputs=[current_index, current_image, image_counter, analysis_text]
-        )
-        
-        # Optional: Add additional information
-        with gr.Accordion("About", open=False):
-            gr.Markdown("""
-            ## About this demo
-            
-            This demo generates alternative text for images.
-            
-            - Upload one or more images using the upload button
-            - Choose a model and response length for generation
-            - Navigate through the images with the Previous and Next buttons
-            - Download CSV with all results
-            
-            Developed by the Natural History Museum in Partnership with National Museums Liverpool.
-            
-            If you find any bugs/have any problems/have any suggestions please feel free to get in touch:
-            chris.addis@nhm.ac.uk
-            """)
-    
-    return demo
-
-# Launch the app
-if __name__ == "__main__":
-    app = create_demo()
-    app.launch()

.ipynb_checkpoints/app-Copy1-checkpoint.py → app-Copy2.py

@@ -41,50 +41,61 @@ def get_sys_prompt(length="medium"):
         dev_prompt = """You are a museum curator tasked with generating alt-text (as defined by W3C) of museum objects for visually impaired and blind users from images. Use British English and follow museum accessibility best practices. Do not start with phrases like 'The image shows' or 'This is an image of'. Be precise, concise and avoid filler and subjective statements. Repsonses should be a maximum of 130 characters."""
     elif length == "medium":
         dev_prompt = """You are a museum curator tasked with generating long descriptions (as defined in W3C) of museum objects for visually impaired and blind users from images. Use British English and follow museum accessibility best practices. Do not start with phrases like 'The image shows' or 'This is an image of'. Be precise, concise and avoid filler and subjective statements. Repsonses should be between 250-300 characters in length."""
-    else:
+    else: # long
         dev_prompt = """You are a museum curator tasked with generating long descriptions (as defined in W3C) of museum objects for visually impaired and blind users from images. Use British English and follow museum accessibility best practices. Do not start with phrases like 'The image shows' or 'This is an image of'. Be precise, concise and avoid filler and subjective statements. Repsonses should be a maxium of 450 characters."""
     return dev_prompt
 
 def create_csv_file_simple(results):
     """Create a CSV file from the results and return the path"""
-    # Create a temporary file
-    fd, path = tempfile.mkstemp(suffix='.csv')
-    
-    with os.fdopen(fd, 'w', newline='') as f:
-        writer = csv.writer(f)
-        # Write header
-        writer.writerow(['image_id', 'content'])
-        # Write data
-        for result in results:
-            writer.writerow([
-                result.get('image_id', ''),
-                result.get('content', '')
-            ])
-    
-    return path
+    try:
+        with tempfile.NamedTemporaryFile(mode='w', suffix='.csv', delete=False, newline='', encoding='utf-8') as f:
+            path = f.name
+            writer = csv.writer(f)
+            writer.writerow(['image_id', 'content'])
+            for result in results:
+                writer.writerow([
+                    result.get('image_id', ''),
+                    result.get('content', '')
+                ])
+        return path
+    except Exception as e:
+        print(f"Error creating CSV: {e}")
+        return None
+
 
-# Extract original filename without path or extension
 def get_base_filename(filepath):
     if not filepath:
         return ""
-    # Get the basename (filename with extension)
     basename = os.path.basename(filepath)
-    # Remove extension
     filename = os.path.splitext(basename)[0]
     return filename
 
-custom_css = """
-.image-container img {
-    object-fit: contain;
-    width: 100%;
-    height: 100%;
-}
-"""
-
 # Define the Gradio interface
 def create_demo():
-    with gr.Blocks(theme=gr.themes.Monochrome(),css=custom_css) as demo:
-        # Replace the existing logo code section:
+    # --- Reintroduce CSS ---
+    custom_css = """
+    /* Container for the image component (#current-image-display is the elem_id of gr.Image) */
+        #current-image-display {
+            height: 600px;           /* Define container height */
+            width: 100%;             /* Define container width (takes column width) */
+            display: flex;           /* Use flexbox for alignment */
+            justify-content: center; /* Center content horizontally */
+            align-items: center;     /* Center content vertically */
+            overflow: hidden;        /* Hide any potential overflow from container */
+        }
+    
+        /* The actual <img> element inside the container */
+        #current-image-display img {
+            object-fit: contain !important;    /* Scale keeping aspect ratio, within bounds */
+            max-width: 100%;        /* Prevent image exceeding container width */
+            max-height: 600px !important;       /* Prevent image exceeding container height */
+            width: auto;            /* Use natural width unless constrained by max-width */
+            height: auto;           /* Use natural height unless constrained by max-height */
+            display: block;         /* Ensure image behaves predictably in flex */
+        }
+    """
+    # --- Pass css to gr.Blocks ---
+    with gr.Blocks(theme=gr.themes.Monochrome(), css=custom_css) as demo:
         with gr.Row():
             with gr.Column(scale=3):
                 gr.Markdown("# MATCHA: Museum Alt-Text for Cultural Heritage with AI 🍵 🌿")
@@ -92,279 +103,179 @@ def create_demo():
                 gr.Markdown("Developed by the Natural History Museum in Partnership with National Museums Liverpool. Funded by the DCMS Pilot Scheme")
             with gr.Column(scale=1):
                 with gr.Row():
-                    # Use gr.Image with all interactive features disabled
-                    gr.Image("images/nhm_logo.png", show_label=False, height=120, 
-                            interactive=False, show_download_button=False, 
-                            show_share_button=False, show_fullscreen_button=False,
-                            container=False)
-                    gr.Image("images/nml_logo.png", show_label=False, height=120, 
-                            interactive=False, show_download_button=False,
-                            show_share_button=False, show_fullscreen_button=False,
-                            container=False)
-        
+                    gr.Image("images/nhm_logo.png", show_label=False, height=120,
+                             interactive=False, show_download_button=False,
+                             show_share_button=False, show_fullscreen_button=False,
+                             container=False, elem_id="nhm-logo")
+                    gr.Image("images/nml_logo.png", show_label=False, height=120,
+                             interactive=False, show_download_button=False,
+                             show_share_button=False, show_fullscreen_button=False,
+                             container=False, elem_id="nml-logo")
+
         with gr.Row():
             # Left column: Controls and uploads
             with gr.Column(scale=1):
-                # Upload interface
                 upload_button = gr.UploadButton(
-                    "Click to Upload Images", 
-                    file_types=["image"], 
+                    "Click to Upload Images",
+                    file_types=["image"],
                     file_count="multiple"
                 )
-                
-                # Define choices as a list of tuples: (Display Name, Internal Value)
                 model_choices = [
-                    # Gemini
                     ("Gemini 2.0 Flash (default)", "google/gemini-2.0-flash-001"),
-                    # GPT-4.1 Series
-                    ("GPT-4.1 Nano", "gpt-4.1-nano"),
-                    ("GPT-4.1 Mini", "gpt-4.1-mini"),
-                    ("GPT-4.1", "gpt-4.1"),
-                    ("ChatGPT Latest", "openai/chatgpt-4o-latest"),
-                    # Other Models
+                    ("GPT-4.1 Nano", "gpt-4.1-nano"), ("GPT-4.1 Mini", "gpt-4.1-mini"),
+                    ("GPT-4.1", "gpt-4.1"), ("ChatGPT Latest", "openai/chatgpt-4o-latest"),
                     ("Claude 3.7 Sonnet", "anthropic/claude-3.7-sonnet"),
                     ("Llama 4 Maverick", "meta-llama/llama-4-maverick"),
-                    # Experimental Models
                     ("Gemini 2.5 Pro (Experimental, limited)", "gemini-2.5-pro-exp-03-25"),
                     ("Gemini 2.0 Flash Thinking (Experimental, limited)", "gemini-2.0-flash-thinking-exp-01-21")
                 ]
-                
-                # Find the internal value of the default choice
                 default_model_internal_value = "google/gemini-2.0-flash-001"
-                
-                # Add model selection dropdown
                 model_choice = gr.Dropdown(
-                    choices=model_choices,
-                    label="Select Model",
-                    value=default_model_internal_value, # Use the internal value for the default
-                    # info="Choose the language model to use." # Optional: Add extra info tooltip
-                    visible=True
+                    choices=model_choices, label="Select Model",
+                    value=default_model_internal_value, visible=True
                 )
-
-                
-                # Add response length selection
                 length_choice = gr.Radio(
-                    choices=["short", "medium", "long"],
-                    label="Response Length",
-                    value="medium",
-                    info="Short: max 130 chars | Medium: 250-300 chars | Long: max 450 chars"
+                    choices=["short", "medium", "long"], label="Response Length",
+                    value="medium", info="Short: max 130 chars | Medium: 250-300 chars | Long: max 450 chars"
                 )
-                
-                # Preview gallery for uploaded images
                 gr.Markdown("### Uploaded Images")
                 input_gallery = gr.Gallery(
-                    label="",
-                    columns=3, 
-                    height=150,
-                    object_fit="contain"
+                    label="Uploaded Image Previews", columns=3, height=150,
+                    object_fit="contain", show_label=False
                 )
-                
-                # Analysis button
                 analyze_button = gr.Button("Generate Alt-Text", variant="primary", size="lg")
-                
-                # Hidden state component to store image info
                 image_state = gr.State([])
                 filename_state = gr.State([])
-                
-                # CSV download component
-                csv_download = gr.File(label="CSV Results")
-            
+                csv_download = gr.File(label="Download CSV Results")
+
             # Right column: Display area
             with gr.Column(scale=2):
-                with gr.Column(elem_classes="image-container"):
-                    current_image = gr.Image(
-                        label="Current Image",
-                        height=600,         # Set the maximum desired height
-                        width=1000,
-                        type="filepath",
-                        show_fullscreen_button=True,
-                        show_download_button=False,
-                        show_share_button=False,
-                        elem_classes="image-container"
-                    )
-                
-                # Navigation row
+                current_image = gr.Image(
+                    label="Current Image",
+                    type="filepath",
+                    elem_id="current-image-display", # ADDED - for CSS targeting
+                    show_fullscreen_button=True,
+                    show_download_button=False,
+                    show_share_button=False,
+                    show_label=False
+                 )
+
                 with gr.Row():
                     prev_button = gr.Button("← Previous", size="sm")
-                    image_counter = gr.Markdown("", elem_id="image-counter")
+                    image_counter = gr.Markdown("0 of 0", elem_id="image-counter")
                     next_button = gr.Button("Next →", size="sm")
-                
-                # Alt-text heading and output
+
                 gr.Markdown("### Generated Alt-text")
-                
-                # Alt-text
                 analysis_text = gr.Textbox(
-                    label="",
-                    value="Upload images and select model to generate alt-text!",
-                    lines=6,
-                    max_lines=10,
-                    interactive=False,
-                    show_label=False
+                    label="Generated Text",
+                    value="Upload images and click 'Generate Alt-Text'.",
+                    lines=6, max_lines=10, interactive=True, show_label=False
                 )
-                
-                # Hidden state for gallery navigation
                 current_index = gr.State(0)
                 all_images = gr.State([])
                 all_results = gr.State([])
-        
-        # Handle file uploads - store files for use during analysis
-        def handle_upload(files):
+
+        # --- Functions (handle_upload, analyze_images, navigators) remain the same ---
+        # Handle file uploads
+        def handle_upload(files, current_paths, current_filenames):
             file_paths = []
             file_names = []
-            for file in files:
-                file_paths.append(file.name)
-                # Extract filename without path or extension for later use
-                file_names.append(get_base_filename(file.name))
-            return file_paths, file_paths, file_names
-        
+            if files:
+                for file in files:
+                    file_paths.append(file.name)
+                    file_names.append(get_base_filename(file.name))
+            return file_paths, file_paths, file_names, 0, None, "0 of 0", "Upload images and click 'Generate Alt-Text'."
+
         upload_button.upload(
             fn=handle_upload,
-            inputs=[upload_button],
-            outputs=[input_gallery, image_state, filename_state]
+            inputs=[upload_button, image_state, filename_state],
+            outputs=[input_gallery, image_state, filename_state,
+                     current_index, current_image, image_counter, analysis_text]
         )
-        
-        # Function to analyze images
-        # Modify the analyze_images function in your code:
 
+        # Analyze images
         def analyze_images(image_paths, model_choice, length_choice, filenames):
             if not image_paths:
-                return [], [], 0, "", "No images", "", ""
-            
-            # Get system prompt based on length selection
+                return [], [], 0, None, "0 of 0", "No images uploaded to analyze.", None
+
             sys_prompt = get_sys_prompt(length_choice)
-            
             image_results = []
-            
-            for i, image_path in enumerate(image_paths):
-                # Use original filename as image_id if available
-                if i < len(filenames) and filenames[i]:
-                    image_id = filenames[i]
-                else:
-                    image_id = f"Image {i+1}"
-                
+            analysis_progress = gr.Progress(track_tqdm=True)
+
+            for i, image_path in enumerate(analysis_progress.tqdm(image_paths, desc="Analyzing Images")):
+                image_id = filenames[i] if i < len(filenames) and filenames[i] else f"Image_{i+1}_{os.path.basename(image_path)}"
                 try:
-                    # Open the image file for analysis
                     img = Image.open(image_path)
-                    prompt0 = prompt_new()  # Using the new prompt function
-                    
-                    # Extract the actual model name (remove any labels like "(default)")
-                    if " (" in model_choice:
-                        model_name = model_choice.split(" (")[0]
-                    else:
-                        model_name = model_choice
-                    
-                    # Check if this is one of the Gemini models that needs special handling
-                    is_gemini_model = "gemini-2.5-pro" in model_name or "gemini-2.0-flash-thinking" in model_name
-                    
-                    if is_gemini_model:
-                        try:
-                            # First try using the dedicated gemini client
-                            result = gemini.generate_caption(
-                                img, 
-                                model=model_name, 
-                                max_image_size=512,
-                                prompt=prompt0, 
-                                prompt_dev=sys_prompt, 
-                                temperature=1
-                            )
-                        except Exception as gemini_error:
-                            # If gemini client fails, fall back to standard OR client
-                            result = OR.generate_caption(
-                                img, 
-                                model=model_name, 
-                                max_image_size=512,
-                                prompt=prompt0, 
-                                prompt_dev=sys_prompt, 
-                                temperature=1
-                            )
-                    else:
-                        # For all other models, use OR client directly
-                        result = OR.generate_caption(
-                            img, 
-                            model=model_name, 
-                            max_image_size=512,
-                            prompt=prompt0, 
-                            prompt_dev=sys_prompt, 
-                            temperature=1
-                        )
-                    
-                    # Add to results
-                    image_results.append({
-                        "image_id": image_id,
-                        "content": result
-                    })
-                    
+                    prompt0 = prompt_new()
+                    model_name = model_choice
+                    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
+                    )
+                    image_results.append({"image_id": image_id, "content": result.strip()})
+                except FileNotFoundError:
+                     error_message = f"Error: File not found at path '{image_path}'"
+                     print(error_message)
+                     image_results.append({"image_id": image_id, "content": error_message})
                 except Exception as e:
-                    error_message = f"Error: {str(e)}"
-                    image_results.append({
-                        "image_id": image_id,
-                        "content": error_message
-                    })
-            
-            # Create a CSV file for download
+                    error_message = f"Error processing {image_id}: {str(e)}"
+                    print(error_message)
+                    image_results.append({"image_id": image_id, "content": error_message})
+
             csv_path = create_csv_file_simple(image_results)
-            
-            # Set up initial display with first image
-            if len(image_paths) > 0:
-                initial_image = image_paths[0]
-                initial_counter = f"{1} of {len(image_paths)}"
-                initial_text = image_results[0]["content"]
-            else:
-                initial_image = ""
-                initial_text = "No images analyzed"
-                initial_counter = "0 of 0"
-            
-            return (image_paths, image_results, 0, initial_image, initial_counter, 
+            initial_image = image_paths[0] if image_paths else None
+            initial_counter = f"1 of {len(image_paths)}" if image_paths else "0 of 0"
+            initial_text = image_results[0]["content"] if image_results else "Analysis complete, but no results generated."
+
+            return (image_paths, image_results, 0, initial_image, initial_counter,
                     initial_text, csv_path)
 
-        
-        # Function to navigate to previous image
+        # Navigate previous
         def go_to_prev(current_idx, images, results):
-            if not images or len(images) == 0:
-                return current_idx, "", "0 of 0", ""
-            
-            new_idx = (current_idx - 1) % len(images) if current_idx > 0 else len(images) - 1
-            counter_html = f"{new_idx + 1} of {len(images)}"
-            
-            return (new_idx, images[new_idx], counter_html, results[new_idx]["content"])
-        
-        # Function to navigate to next image
+            if not images or not results or len(images) == 0:
+                return current_idx, None, "0 of 0", ""
+            new_idx = (current_idx - 1 + len(images)) % len(images)
+            counter_text = f"{new_idx + 1} of {len(images)}"
+            result_content = results[new_idx]["content"] if new_idx < len(results) else "Error: Result not found"
+            return (new_idx, images[new_idx], counter_text, result_content)
+
+        # Navigate next
         def go_to_next(current_idx, images, results):
-            if not images or len(images) == 0:
-                return current_idx, "", "0 of 0", ""
-            
+            if not images or not results or len(images) == 0:
+                return current_idx, None, "0 of 0", ""
             new_idx = (current_idx + 1) % len(images)
-            counter_html = f"{new_idx + 1} of {len(images)}"
-            
-            return (new_idx, images[new_idx], counter_html, results[new_idx]["content"])
-        
-        # Connect the analyze button
+            counter_text = f"{new_idx + 1} of {len(images)}"
+            result_content = results[new_idx]["content"] if new_idx < len(results) else "Error: Result not found"
+            return (new_idx, images[new_idx], counter_text, result_content)
+
+        # Connect analyze button
         analyze_button.click(
             fn=analyze_images,
             inputs=[image_state, model_choice, length_choice, filename_state],
-            outputs=[
-                all_images, all_results, current_index, current_image, image_counter,
-                analysis_text, csv_download
-            ]
+            outputs=[all_images, all_results, current_index, current_image, image_counter,
+                     analysis_text, csv_download]
         )
-        
+
         # Connect navigation buttons
         prev_button.click(
-            fn=go_to_prev,
-            inputs=[current_index, all_images, all_results],
-            outputs=[current_index, current_image, image_counter, analysis_text]
+            fn=go_to_prev, inputs=[current_index, all_images, all_results],
+            outputs=[current_index, current_image, image_counter, analysis_text], queue=False
         )
-        
         next_button.click(
-            fn=go_to_next,
-            inputs=[current_index, all_images, all_results],
-            outputs=[current_index, current_image, image_counter, analysis_text]
+            fn=go_to_next, inputs=[current_index, all_images, all_results],
+            outputs=[current_index, current_image, image_counter, analysis_text], queue=False
         )
-        
-        # Optional: Add additional information
+
+        # About section
         with gr.Accordion("About", open=False):
-            gr.Markdown("""
-            ## About this demo
+             gr.Markdown("""
+            ## About MATCHA 🍵:
             
             This demo generates alternative text for images.
             
@@ -378,10 +289,11 @@ def create_demo():
             If you find any bugs/have any problems/have any suggestions please feel free to get in touch:
             chris.addis@nhm.ac.uk
             """)
-    
+
     return demo
 
 # Launch the app
 if __name__ == "__main__":
+
     app = create_demo()
     app.launch()

app.py

@@ -36,15 +36,24 @@ PREFERENCES_FILE = "data/user_preferences.csv"
 # Ensure directory exists
 os.makedirs(os.path.dirname(PREFERENCES_FILE), exist_ok=True)
 
-def get_sys_prompt(length="medium"):
+def get_sys_prompt(length="medium", photograph=False):
+
+    if photograph:
+        object_type = "wildlife photography"
+    else:
+        object_type = "museum objects"
+        
+    dev_prompt = """You are a museum curator tasked with generating alt-text (as defined by W3C) of {object_type} for visually impaired and blind users from images. Use British English and follow museum accessibility best practices. Do not start with phrases like 'The image shows' or 'This is an image of'. Be precise, concise and avoid filler and subjective statements."""
+    
     if length == "short":
-        dev_prompt = """You are a museum curator tasked with generating alt-text (as defined by W3C) of museum objects for visually impaired and blind users from images. Use British English and follow museum accessibility best practices. Do not start with phrases like 'The image shows' or 'This is an image of'. Be precise, concise and avoid filler and subjective statements. Repsonses should be a maximum of 130 characters."""
+        dev_prompt += " Repsonses should be a maximum of 130 characters."
     elif length == "medium":
-        dev_prompt = """You are a museum curator tasked with generating long descriptions (as defined in W3C) of museum objects for visually impaired and blind users from images. Use British English and follow museum accessibility best practices. Do not start with phrases like 'The image shows' or 'This is an image of'. Be precise, concise and avoid filler and subjective statements. Repsonses should be between 250-300 characters in length."""
+        dev_prompt += " Repsonses should be a maximum of 250-300 characters."
     else: # long
-        dev_prompt = """You are a museum curator tasked with generating long descriptions (as defined in W3C) of museum objects for visually impaired and blind users from images. Use British English and follow museum accessibility best practices. Do not start with phrases like 'The image shows' or 'This is an image of'. Be precise, concise and avoid filler and subjective statements. Repsonses should be a maxium of 450 characters."""
+        dev_prompt += " Repsonses should be a maximum of 450 characters."
     return dev_prompt
 
+
 def create_csv_file_simple(results):
     """Create a CSV file from the results and return the path"""
     try:
@@ -62,7 +71,6 @@ def create_csv_file_simple(results):
         print(f"Error creating CSV: {e}")
         return None
 
-
 def get_base_filename(filepath):
     if not filepath:
         return ""
@@ -127,7 +135,8 @@ def create_demo():
                     ("Claude 3.7 Sonnet", "anthropic/claude-3.7-sonnet"),
                     ("Llama 4 Maverick", "meta-llama/llama-4-maverick"),
                     ("Gemini 2.5 Pro (Experimental, limited)", "gemini-2.5-pro-exp-03-25"),
-                    ("Gemini 2.0 Flash Thinking (Experimental, limited)", "gemini-2.0-flash-thinking-exp-01-21")
+                    ("Gemini 2.0 Flash Thinking (Experimental, limited)", "gemini-2.0-flash-thinking-exp-01-21"),
+                    ("Gemini 2.5 Flash Thinking (Preview)", "google/gemini-2.5-flash-preview:thinking")
                 ]
                 default_model_internal_value = "google/gemini-2.0-flash-001"
                 model_choice = gr.Dropdown(
@@ -138,6 +147,16 @@ def create_demo():
                     choices=["short", "medium", "long"], label="Response Length",
                     value="medium", info="Short: max 130 chars | Medium: 250-300 chars | Long: max 450 chars"
                 )
+                
+                # Advanced settings accordion
+                with gr.Accordion("Advanced Settings", open=False):
+                    content_type = gr.Radio(
+                        choices=["museum objects", "wildlife photography"], 
+                        label="Content Type",
+                        value="museum objects", 
+                        info="Choose the type of content in your images"
+                    )
+                
                 gr.Markdown("### Uploaded Images")
                 input_gallery = gr.Gallery(
                     label="Uploaded Image Previews", columns=3, height=150,
@@ -194,11 +213,12 @@ def create_demo():
         )
 
         # Analyze images
-        def analyze_images(image_paths, model_choice, length_choice, filenames):
+        def analyze_images(image_paths, model_choice, length_choice, filenames, content_type_choice):
             if not image_paths:
                 return [], [], 0, None, "0 of 0", "No images uploaded to analyze.", None
 
-            sys_prompt = get_sys_prompt(length_choice)
+            is_photography = content_type_choice == "wildlife photography"
+            sys_prompt = get_sys_prompt(length_choice, photograph=is_photography)
             image_results = []
             analysis_progress = gr.Progress(track_tqdm=True)
 
@@ -257,7 +277,7 @@ def create_demo():
         # Connect analyze button
         analyze_button.click(
             fn=analyze_images,
-            inputs=[image_state, model_choice, length_choice, filename_state],
+            inputs=[image_state, model_choice, length_choice, filename_state, content_type],
             outputs=[all_images, all_results, current_index, current_image, image_counter,
                      analysis_text, csv_download]
         )
@@ -275,9 +295,20 @@ def create_demo():
         # About section
         with gr.Accordion("About", open=False):
              gr.Markdown("""
-             ## About this demo
-             ... [content unchanged] ...
-             """)
+            ## About MATCHA 🍵:
+            
+            This demo generates alternative text for images.
+            
+            - Upload one or more images using the upload button
+            - Choose a model and response length for generation
+            - Navigate through the images with the Previous and Next buttons
+            - Download CSV with all results
+            
+            Developed by the Natural History Museum in Partnership with National Museums Liverpool.
+            
+            If you find any bugs/have any problems/have any suggestions please feel free to get in touch:
+            chris.addis@nhm.ac.uk
+            """)
 
     return demo
 

library/utils_prompt.py

@@ -11,7 +11,7 @@ def prompt_new(title=None):
         title_info = {}
     else:
         title_info = f"(titled: {title})"
-    return f"This object {title_info} is part of a museum collection."
+    return f"Generate alt-text for this object {title_info}:"
     
 def prompt_1(title=None):
     if title == None: