Update app.py

app.py

@@ -1,3 +1,9 @@
+Okay, let's integrate gr.Examples into your script with some diverse, high-quality Stable Diffusion-style prompts and enable caching.
+
+Caching examples means that when the Space first builds (or rebuilds after changes), it will run the query function once for each example and store the resulting image. When a user clicks that example later, the cached image is shown instantly instead of running the model again.
+
+Here's the modified app.py:
+
 import gradio as gr
 import requests
 import io
@@ -7,113 +13,319 @@ import time
 from PIL import Image
 from deep_translator import GoogleTranslator
 import json
+from typing import Callable, List, Any, Literal # Added for type hinting Examples
 
 # Project by Nymbo
 
-API_URL = "https://api-inference.huggingface.co/models/black-forest-labs/FLUX.1-schnell"
-API_TOKEN = os.getenv("HF_READ_TOKEN")
-headers = {"Authorization": f"Bearer {API_TOKEN}"}
+API_URL = "https://api-inference.huggingface.co/models/black-forest-labs/FLUX.1-dev"
+API_TOKEN = os.getenv("HF_READ_TOKEN") # Your main token for primary use
+# Consider adding more read tokens if you have them for resilience
+ADDITIONAL_TOKENS = [t for t in os.getenv("HF_EXTRA_TOKENS", "").split(',') if t] # Example: HF_EXTRA_TOKENS="token1,token2"
+ALL_TOKENS = [API_TOKEN] + ADDITIONAL_TOKENS
+if not API_TOKEN:
+    print("Warning: HF_READ_TOKEN is not set. API calls may fail.")
+    # Optional: raise an error or use a dummy token if needed
+    # raise ValueError("HF_READ_TOKEN environment variable is required.")
+
 timeout = 100
 
 # Function to query the API and return the generated image
-def query(prompt, is_negative=False, steps=30, cfg_scale=7, sampler="DPM++ 2M Karras", seed=-1, strength=0.7, width=1024, height=1024):
-    if prompt == "" or prompt is None:
+# Added type hints for clarity
+def query(
+    prompt: str,
+    negative_prompt: str = "(deformed, distorted, disfigured), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, (mutated hands and fingers), disconnected limbs, mutation, mutated, ugly, disgusting, blurry, amputation, misspellings, typos",
+    steps: int = 35,
+    cfg_scale: float = 7.0,
+    sampler: str = "DPM++ 2M Karras", # Matched default in UI
+    seed: int = -1,
+    strength: float = 0.7,
+    width: int = 1024,
+    height: int = 1024
+) -> Image.Image | None:
+
+    if not prompt: # Simplified check
+        gr.Warning("Prompt cannot be empty.")
+        return None
+    if not ALL_TOKENS:
+        gr.Error("No Hugging Face API tokens available.")
         return None
 
-    key = random.randint(0, 999)
-    
-    API_TOKEN = random.choice([os.getenv("HF_READ_TOKEN")])
-    headers = {"Authorization": f"Bearer {API_TOKEN}"}
-    
-    # Translate the prompt from Russian to English if necessary
-    prompt = GoogleTranslator(source='ru', target='en').translate(prompt)
-    print(f'\033[1mGeneration {key} translation:\033[0m {prompt}')
+    key = random.randint(0, 9999)
+    start_time = time.time()
+
+    # Rotate through available tokens
+    selected_api_token = random.choice(ALL_TOKENS)
+    headers = {"Authorization": f"Bearer {selected_api_token}"}
+
+    translated_prompt = prompt
+    try:
+        # Simple check if likely Russian (Cyrillic) - adjust if needed
+        if any('\u0400' <= char <= '\u04FF' for char in prompt):
+             translated_prompt = GoogleTranslator(source='ru', target='en').translate(prompt)
+             print(f'\033[1mGeneration {key} translation:\033[0m {translated_prompt}')
+        else:
+             # Assume English or other non-Russian if no Cyrillic detected
+             pass
+    except Exception as e:
+        print(f"Translation failed: {e}. Using original prompt.")
+        # Decide if you want to raise an error or just proceed
+        # gr.Warning(f"Translation failed: {e}. Using original prompt.")
 
     # Add some extra flair to the prompt
-    prompt = f"{prompt} | ultra detail, ultra elaboration, ultra quality, perfect."
-    print(f'\033[1mGeneration {key}:\033[0m {prompt}')
-    
+    enhanced_prompt = f"{translated_prompt} | ultra detail, ultra elaboration, ultra quality, perfect."
+    print(f'\033[1mGeneration {key} starting:\033[0m {enhanced_prompt}')
+
     # Prepare the payload for the API call, including width and height
     payload = {
-        "inputs": prompt,
-        "is_negative": is_negative,
-        "steps": steps,
-        "cfg_scale": cfg_scale,
-        "seed": seed if seed != -1 else random.randint(1, 1000000000),
-        "strength": strength,
+        "inputs": enhanced_prompt,
+        "negative_prompt": negative_prompt, # Use the negative prompt input
+        # Note: The underlying API might not support all parameters directly.
+        # Check the specific model card for supported parameters.
+        # We send common ones; the model API will use what it understands.
         "parameters": {
-            "width": width,  # Pass the width to the API
-            "height": height  # Pass the height to the API
+            "num_inference_steps": steps,
+            "guidance_scale": cfg_scale,
+            "seed": seed if seed != -1 else random.randint(1, 2**32 - 1), # Use larger seed range
+            "strength": strength, # Often used for img2img, may not apply here
+            "width": width,
+            "height": height,
+            # Sampler might not be directly controllable via this endpoint structure
+            # The model often uses its default or recommended sampler.
+            # Sending it anyway in case future API versions support it.
+            "scheduler": sampler,
         }
     }
+    # Clean up payload if values are None or default where API expects omission
+    if seed == -1:
+        del payload["parameters"]["seed"] # Let API choose random if -1
+    if not negative_prompt:
+         # If empty string, some APIs prefer omitting the key entirely
+        del payload["negative_prompt"]
+
+
+    print(f"Payload for {key}: {json.dumps(payload, indent=2)}")
 
     # Send the request to the API and handle the response
-    response = requests.post(API_URL, headers=headers, json=payload, timeout=timeout)
-    if response.status_code != 200:
-        print(f"Error: Failed to get image. Response status: {response.status_code}")
-        print(f"Response content: {response.text}")
-        if response.status_code == 503:
-            raise gr.Error(f"{response.status_code} : The model is being loaded")
-        raise gr.Error(f"{response.status_code}")
-    
     try:
+        response = requests.post(API_URL, headers=headers, json=payload, timeout=timeout)
+        response.raise_for_status() # Raises HTTPError for bad responses (4xx or 5xx)
+
         # Convert the response content into an image
         image_bytes = response.content
         image = Image.open(io.BytesIO(image_bytes))
-        print(f'\033[1mGeneration {key} completed!\033[0m ({prompt})')
+        end_time = time.time()
+        print(f'\033[1mGeneration {key} completed in {end_time - start_time:.2f}s!\033[0m')
         return image
-    except Exception as e:
-        print(f"Error when trying to open the image: {e}")
-        return None
+
+    except requests.exceptions.Timeout:
+        print(f"Error: Request timed out after {timeout} seconds.")
+        raise gr.Error(f"Request timed out ({timeout}s). Model might be busy. Try again later.")
+    except requests.exceptions.RequestException as e:
+        status_code = e.response.status_code if e.response is not None else "N/A"
+        error_content = e.response.text if e.response is not None else str(e)
+        print(f"Error: API request failed. Status: {status_code}, Content: {error_content}")
+        if status_code == 503:
+            # Check for specific error messages if available
+            if "is currently loading" in error_content:
+                 raise gr.Error("Model is loading. Please wait a moment and try again.")
+            else:
+                 raise gr.Error("Model service unavailable (503). It might be overloaded or down. Try again later.")
+        elif status_code == 400:
+             raise gr.Error(f"Bad Request (400). Check parameters. API response: {error_content[:200]}")
+        elif status_code == 429:
+             raise gr.Error("Too many requests (429). Rate limit hit. Please wait.")
+        else:
+            raise gr.Error(f"API Error ({status_code}). Response: {error_content[:200]}")
+    except (OSError, json.JSONDecodeError, IOError) as e:
+         # Catch potential issues with image decoding or other file errors
+         print(f"Error processing response or image: {e}")
+         raise gr.Error(f"Failed to process the image response: {e}")
+    except Exception as e: # Catch any other unexpected errors
+        print(f"An unexpected error occurred: {e}")
+        import traceback
+        traceback.print_exc() # Print full traceback to console logs
+        raise gr.Error(f"An unexpected error occurred: {e}")
+
+
+# --- Gradio UI ---
 
 # CSS to style the app
 css = """
 #app-container {
-    max-width: 800px;
+    max-width: 960px; /* Slightly wider */
     margin-left: auto;
     margin-right: auto;
 }
-textarea:focus {
-    background: #0d1117 !important;
-}
+/* Add more styling if desired */
 """
 
+# Default negative prompt can be reused
+default_negative_prompt = "(deformed, distorted, disfigured), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, (mutated hands and fingers), disconnected limbs, mutation, mutated, ugly, disgusting, blurry, amputation, misspellings, typos"
+
+# Define the examples
+# Each inner list corresponds to the inputs of the 'query' function IN ORDER:
+# [prompt, negative_prompt, steps, cfg_scale, sampler, seed, strength, width, height]
+example_list = [
+    [
+        "Epic cinematic shot of a medieval knight kneeling in a misty forest, volumetric lighting, hyperrealistic photo, 8k",
+        default_negative_prompt, 40, 7.5, "DPM++ 2M Karras", 12345, 0.7, 1024, 1024
+    ],
+    [
+        "Studio Ghibli style illustration of a cozy bakery storefront on a rainy day, warm lighting, detailed",
+        default_negative_prompt, 30, 6.0, "Euler a", 54321, 0.7, 1024, 1024
+    ],
+    [
+        "Macro photograph of a dewdrop on a spider web, intricate details, shallow depth of field, natural lighting",
+        "blurry, unfocused, cartoon", 50, 8.0, "DPM++ 2M Karras", -1, 0.7, 1024, 1024 # Random seed
+    ],
+    [
+        "Steampunk astronaut exploring an alien jungle landscape, brass and copper details, vibrant bioluminescent plants, wide angle",
+        default_negative_prompt, 35, 7.0, "DPM++ SDE Karras", 98765, 0.7, 1216, 832 # Different aspect ratio
+    ],
+    [
+        "Abstract geometric art, vibrant contrasting colors, sharp edges, minimalistic design, 4k wallpaper",
+        "photorealistic, noisy, cluttered", 25, 5.0, "Euler", -1, 0.7, 1024, 1024
+    ],
+     [
+        "Кот в очках читает книгу у камина", # Example in Russian for translation testing
+        default_negative_prompt, 35, 7.0, "DPM++ 2M Karras", 11223, 0.7, 1024, 1024
+    ]
+]
+
+
 # Build the Gradio UI with Blocks
-with gr.Blocks(theme='Nymbo/Nymbo_Theme', css=css) as app:
+# Use the custom theme if it's available, otherwise fallback to default
+try:
+    theme = gr.themes.Base.load('Nymbo/Nymbo_Theme')
+except Exception:
+    print("Could not load Nymbo/Nymbo_Theme, using default theme.")
+    theme = gr.themes.Default()
+
+with gr.Blocks(theme=theme, css=css) as app:
     # Add a title to the app
-    gr.HTML("<center><h1>FLUX.1-Schnell</h1></center>")
-    
+    gr.HTML("<center><h1>FLUX.1-Dev Image Generator</h1></center>")
+
     # Container for all the UI elements
     with gr.Column(elem_id="app-container"):
         # Add a text input for the main prompt
         with gr.Row():
-            with gr.Column(elem_id="prompt-container"):
-                with gr.Row():
-                    text_prompt = gr.Textbox(label="Prompt", placeholder="Enter a prompt here", lines=2, elem_id="prompt-text-input")
-                
-                # Accordion for advanced settings
-                with gr.Row():
-                    with gr.Accordion("Advanced Settings", open=False):
-                        negative_prompt = gr.Textbox(label="Negative Prompt", placeholder="What should not be in the image", value="(deformed, distorted, disfigured), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, (mutated hands and fingers), disconnected limbs, mutation, mutated, ugly, disgusting, blurry, amputation, misspellings, typos", lines=3, elem_id="negative-prompt-text-input")
-                        with gr.Row():
-                            width = gr.Slider(label="Width", value=1024, minimum=64, maximum=1216, step=32)
-                            height = gr.Slider(label="Height", value=1024, minimum=64, maximum=1216, step=32)
-                        steps = gr.Slider(label="Sampling steps", value=4, minimum=1, maximum=100, step=1)
-                        cfg = gr.Slider(label="CFG Scale", value=7, minimum=1, maximum=20, step=1)
-                        strength = gr.Slider(label="Strength", value=0.7, minimum=0, maximum=1, step=0.001)
-                        seed = gr.Slider(label="Seed", value=-1, minimum=-1, maximum=1000000000, step=1) # Setting the seed to -1 will make it random
-                        method = gr.Radio(label="Sampling method", value="DPM++ 2M Karras", choices=["DPM++ 2M Karras", "DPM++ SDE Karras", "Euler", "Euler a", "Heun", "DDIM"])
+            with gr.Column(scale=3): # Give prompt more width
+                text_prompt = gr.Textbox(label="Prompt", placeholder="Enter a prompt here (English or Russian)", lines=3, elem_id="prompt-text-input") # Increased lines
+
+            with gr.Column(scale=1): # Negative prompt column
+                 negative_prompt = gr.Textbox(label="Negative Prompt", placeholder="What to avoid...", value=default_negative_prompt, lines=3, elem_id="negative-prompt-text-input")
+
+        # Accordion for advanced settings
+        with gr.Accordion("Advanced Settings", open=False):
+             with gr.Row():
+                 width = gr.Slider(label="Width", value=1024, minimum=256, maximum=1216, step=64) # Adjusted steps/min
+                 height = gr.Slider(label="Height", value=1024, minimum=256, maximum=1216, step=64) # Adjusted steps/min
+             with gr.Row():
+                 steps = gr.Slider(label="Sampling steps", value=35, minimum=10, maximum=100, step=1) # Adjusted min
+                 cfg = gr.Slider(label="CFG Scale", value=7.0, minimum=1.0, maximum=20.0, step=0.5) # Added step
+                 strength = gr.Slider(label="Strength (Img2Img)", value=0.7, minimum=0.0, maximum=1.0, step=0.01, info="Primarily for Image-to-Image tasks, may have limited effect here.") # Added step and info
+             with gr.Row():
+                 seed = gr.Slider(label="Seed (-1 for random)", value=-1, minimum=-1, maximum=2**32 - 1, step=1) # Increased max seed
+                 method = gr.Radio(
+                     label="Sampling method (Scheduler)",
+                     value="DPM++ 2M Karras",
+                     choices=["DPM++ 2M Karras", "DPM++ SDE Karras", "Euler", "Euler a", "Heun", "DDIM"],
+                     info="Note: Model API might use its default scheduler regardless of selection."
+                 )
 
         # Add a button to trigger the image generation
         with gr.Row():
-            text_button = gr.Button("Run", variant='primary', elem_id="gen-button")
-        
+            text_button = gr.Button("Generate Image", variant='primary', elem_id="gen-button")
+
         # Image output area to display the generated image
         with gr.Row():
-            image_output = gr.Image(type="pil", label="Image Output", elem_id="gallery")
-        
-        # Bind the button to the query function with the added width and height inputs
-        text_button.click(query, inputs=[text_prompt, negative_prompt, steps, cfg, method, seed, strength, width, height], outputs=image_output)
+            image_output = gr.Image(type="pil", label="Generated Image", elem_id="gallery") # Changed label
+
+
+        # --- Add Examples Component Here ---
+        with gr.Row():
+             gr.Examples(
+                 examples=example_list,
+                 # List ALL components that correspond to the example list order
+                 inputs=[text_prompt, negative_prompt, steps, cfg, method, seed, strength, width, height],
+                 outputs=image_output, # The component to display the output
+                 fn=query, # The function to run TO GENERATE the examples for caching
+                 cache_examples=True, # << ENABLE CACHING HERE
+                 label="Examples (Click to Run & View Cached Result)", # Customize label
+                 examples_per_page=6 # Adjust how many show per page
+                 # run_on_click=True # Optional: Set True if you want clicking to trigger 'query' again, even if cached. Usually False when caching.
+             )
+
+        # Bind the main button to the query function
+        # Ensure the order matches the function definition (excluding self if it were a class method)
+        text_button.click(
+            query,
+            inputs=[text_prompt, negative_prompt, steps, cfg, method, seed, strength, width, height],
+            outputs=image_output,
+            api_name="generate_image" # Optional: Set API name for potential external calls
+        )
+
 
 # Launch the Gradio app
-app.launch(show_api=False, share=False)
+# share=False is correct for Hugging Face Spaces deployment
+# show_api=False is fine unless you specifically need to expose the API endpoint documentation
+# debug=True can be useful for development but remove for production/sharing
+app.launch(show_api=False, share=False)
+
+
+Key Changes:
+
+Import typing: Added Callable, List, Any, Literal for better type hinting, though not strictly required.
+
+Error Handling in query: Made error handling slightly more robust, checking for token availability, using response.raise_for_status(), catching specific request exceptions (like Timeout, 503), and providing clearer Gradio errors (gr.Error, gr.Warning). Also added basic translation error handling.
+
+API Payload: Adjusted the payload structure slightly based on common inference API patterns (e.g., num_inference_steps, guidance_scale). Added notes that the specific model might ignore some parameters. Handles -1 seed better.
+
+Default Negative Prompt: Stored the default negative prompt in a variable for reuse in examples.
+
+example_list: Defined a list of lists. Each inner list contains values for all the inputs to the query function, in the correct order. Includes diverse prompts and some parameter variations. Added a Russian example.
+
+gr.Examples Instantiation:
+
+Placed gr.Examples(...) within the gr.Blocks context, after the main input/output components.
+
+examples=example_list: Passed the defined list.
+
+inputs=[...]: Listed all the input components (gr.Textbox, gr.Slider, etc.) in the exact order corresponding to the data in example_list.
+
+outputs=image_output: Specified the output component.
+
+fn=query: Crucially, provided the query function. This tells Gradio how to generate the results for caching.
+
+cache_examples=True: This enables the caching mechanism.
+
+Added label and examples_per_page for better UI.
+
+run_on_click is typically False or omitted when cache_examples=True, as the point is to show the pre-computed result. Set it to True only if you want clicking an example to re-run the generation even if it's cached (useful if you want users to easily try variations from an example starting point).
+
+UI Tweaks: Increased prompt textbox lines, adjusted slider steps/ranges, added info text to some sliders/radios.
+
+Theme Loading: Added a try...except block for loading the custom theme to fall back gracefully if it's not found.
+
+API Token Handling: Added basic handling for multiple tokens via an environment variable HF_EXTRA_TOKENS (comma-separated) and rotation.
+
+Before Running:
+
+Update requirements.txt: Ensure gradio (version >= 4.x recommended for latest features/fixes), requests, pillow, deep-translator are listed. You likely don't need langdetect anymore if you removed it.
+
+requests
+pillow
+deep-translator
+gradio>=4.44.1 # Use the version suggested or newer
+IGNORE_WHEN_COPYING_START
+content_copy
+download
+Use code with caution.
+Txt
+IGNORE_WHEN_COPYING_END
+
+Set Environment Variables: Make sure HF_READ_TOKEN is set in your Space secrets. Optionally set HF_EXTRA_TOKENS if you have more tokens.
+
+Commit and Push: Save app.py and requirements.txt, commit, and push to your Space.
+
+The first time the Space builds after these changes, it will take longer as it runs query for each example to build the cache. Subsequent loads will be faster, and clicking examples will show results instantly.