enabling MCP server

app.py

@@ -4,8 +4,8 @@ import io
 import random
 import os
 import time
+from typing import Optional
 from PIL import Image
-from deep_translator import GoogleTranslator
 import json
 
 # Project by Nymbo
@@ -15,36 +15,76 @@ API_TOKEN = os.getenv("HF_READ_TOKEN")
 headers = {"Authorization": f"Bearer {API_TOKEN}"}
 timeout = 100
 
-# Function to query the API and return the generated image
-def query(prompt, is_negative=False, steps=35, cfg_scale=7, sampler="DPM++ 2M Karras", seed=-1, strength=0.7, width=1024, height=1024):
-    if prompt == "" or prompt is None:
+def query(
+    prompt: str,
+    negative_prompt: str = "",
+    steps: int = 35,
+    cfg_scale: float = 7,
+    sampler: str = "DPM++ 2M Karras",
+    seed: int = -1,
+    strength: float = 0.7,
+    width: int = 1024,
+    height: int = 1024,
+) -> Optional[Image.Image]:
+    """Generate an image from a text prompt using FLUX.1-Krea via the Hugging Face Inference API.
+
+    This function is exposed as an MCP tool via Gradio. It accepts common text-to-image
+    parameters and returns a single generated image.
+
+    Args:
+        prompt: The main text prompt describing the desired image. Required.
+        negative_prompt: Text describing what should be avoided in the image.
+        steps: Number of denoising steps (quality vs speed). Typical range 1–100.
+        cfg_scale: Classifier-free guidance scale (adherence to prompt). Typical range 1–20.
+        sampler: Sampling/scheduler method name. If unsupported by the backend, it may be ignored.
+        seed: Random seed. Use -1 for a random seed on each call.
+        strength: Image-to-image strength (0–1). Not used for pure text-to-image but included for compatibility.
+        width: Output image width in pixels. Usually multiples of 32.
+        height: Output image height in pixels. Usually multiples of 32.
+
+    Returns:
+        A PIL.Image if generation succeeds; otherwise None.
+
+    Raises:
+        gr.Error: If the backend returns a non-200 HTTP status.
+    """
+    if not prompt:
         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}')
 
-    # 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}')
-    
-    # Prepare the payload for the API call, including width and height
+    api_token = os.getenv("HF_READ_TOKEN")
+    if not api_token:
+        raise gr.Error("Missing HF_READ_TOKEN environment variable.")
+
+    headers = {"Authorization": f"Bearer {api_token}"}
+
+    # Slightly augment the prompt for higher quality
+    prompt_aug = f"{prompt} | ultra detail, ultra elaboration, ultra quality, perfect."
+    print(f'\033[1mGeneration {key}:\033[0m {prompt_aug}')
+
+    # Compute seed value
+    seed_value = seed if seed != -1 else random.randint(1, 1_000_000_000)
+
+    # Prepare payload aligned with HF Inference API conventions
+    parameters = {
+        "num_inference_steps": steps,
+        "guidance_scale": cfg_scale,
+        "width": width,
+        "height": height,
+        "seed": seed_value,
+        # Best-effort: some backends accept scheduler/sampler names; safe to send.
+        "scheduler": sampler,
+    }
+    if negative_prompt and negative_prompt.strip():
+        parameters["negative_prompt"] = negative_prompt
+    # Include strength only if meaningful
+    if 0 < strength <= 1:
+        parameters["strength"] = strength
+
     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,
-        "parameters": {
-            "width": width,  # Pass the width to the API
-            "height": height  # Pass the height to the API
-        }
+        "inputs": prompt_aug,
+        "parameters": parameters,
     }
 
     # Send the request to the API and handle the response
@@ -55,12 +95,12 @@ def query(prompt, is_negative=False, steps=35, cfg_scale=7, sampler="DPM++ 2M Ka
         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:
         # 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})')
+        print(f'\033[1mGeneration {key} completed!\033[0m ({prompt_aug})')
         return image
     except Exception as e:
         print(f"Error when trying to open the image: {e}")
@@ -110,7 +150,21 @@ with gr.Blocks(theme='Nymbo/Nymbo_Theme', css=css) as app:
             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)
+        text_button.click(
+            query,
+            inputs=[
+                text_prompt,
+                negative_prompt,
+                steps,
+                cfg,
+                method,
+                seed,
+                strength,
+                width,
+                height,
+            ],
+            outputs=image_output,
+        )
 
 # Launch the Gradio app
 app.launch(show_api=True, share=False, mcp_server=True)